diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,8 +1,8 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 1.0,
-  "global_step": 109,
+  "epoch": 100.0,
+  "global_step": 10900,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -34,13 +34,13 @@
     {
       "epoch": 0.05,
       "learning_rate": 3e-06,
-      "loss": 14.1132,
+      "loss": 14.113,
       "step": 5
     },
     {
       "epoch": 0.06,
       "learning_rate": 4e-06,
-      "loss": 15.2949,
+      "loss": 15.2944,
       "step": 6
     },
     {
@@ -52,628 +52,65584 @@
     {
       "epoch": 0.07,
       "learning_rate": 6e-06,
-      "loss": 13.307,
+      "loss": 13.306,
       "step": 8
     },
     {
       "epoch": 0.08,
       "learning_rate": 6e-06,
-      "loss": 12.5476,
+      "loss": 12.5471,
       "step": 9
     },
     {
       "epoch": 0.09,
       "learning_rate": 7e-06,
-      "loss": 13.2011,
+      "loss": 13.1991,
       "step": 10
     },
     {
       "epoch": 0.1,
       "learning_rate": 8e-06,
-      "loss": 12.3008,
+      "loss": 12.3004,
       "step": 11
     },
     {
       "epoch": 0.11,
       "learning_rate": 9e-06,
-      "loss": 11.5046,
+      "loss": 11.5045,
       "step": 12
     },
     {
       "epoch": 0.12,
       "learning_rate": 1e-05,
-      "loss": 10.6207,
+      "loss": 10.6129,
       "step": 13
     },
     {
       "epoch": 0.13,
       "learning_rate": 1.1e-05,
-      "loss": 10.1038,
+      "loss": 10.1059,
       "step": 14
     },
     {
       "epoch": 0.14,
       "learning_rate": 1.2e-05,
-      "loss": 8.8277,
+      "loss": 8.8293,
       "step": 15
     },
     {
       "epoch": 0.15,
       "learning_rate": 1.3e-05,
-      "loss": 7.8204,
+      "loss": 7.8176,
       "step": 16
     },
     {
       "epoch": 0.16,
       "learning_rate": 1.4e-05,
-      "loss": 6.3586,
+      "loss": 6.3592,
       "step": 17
     },
     {
       "epoch": 0.17,
       "learning_rate": 1.5e-05,
-      "loss": 5.6274,
+      "loss": 5.6219,
       "step": 18
     },
     {
       "epoch": 0.17,
       "learning_rate": 1.6e-05,
-      "loss": 5.6395,
+      "loss": 5.6302,
       "step": 19
     },
     {
       "epoch": 0.18,
       "learning_rate": 1.7000000000000003e-05,
-      "loss": 5.0595,
+      "loss": 5.0504,
       "step": 20
     },
     {
       "epoch": 0.19,
       "learning_rate": 1.8e-05,
-      "loss": 5.2143,
+      "loss": 5.232,
       "step": 21
     },
     {
       "epoch": 0.2,
       "learning_rate": 1.9e-05,
-      "loss": 5.5995,
+      "loss": 5.7323,
       "step": 22
     },
     {
       "epoch": 0.21,
       "learning_rate": 2e-05,
-      "loss": 4.9733,
+      "loss": 4.972,
       "step": 23
     },
     {
       "epoch": 0.22,
       "learning_rate": 2.1000000000000002e-05,
-      "loss": 4.9453,
+      "loss": 4.9406,
       "step": 24
     },
     {
       "epoch": 0.23,
       "learning_rate": 2.2e-05,
-      "loss": 4.8826,
+      "loss": 4.9408,
       "step": 25
     },
     {
       "epoch": 0.24,
       "learning_rate": 2.3e-05,
-      "loss": 4.4724,
+      "loss": 4.4449,
       "step": 26
     },
     {
       "epoch": 0.25,
       "learning_rate": 2.4e-05,
-      "loss": 3.7938,
+      "loss": 3.7126,
       "step": 27
     },
     {
       "epoch": 0.26,
       "learning_rate": 2.5e-05,
-      "loss": 3.2653,
+      "loss": 3.2592,
       "step": 28
     },
     {
       "epoch": 0.27,
       "learning_rate": 2.6e-05,
-      "loss": 3.3843,
+      "loss": 3.3439,
       "step": 29
     },
     {
       "epoch": 0.28,
       "learning_rate": 2.7e-05,
-      "loss": 3.5122,
+      "loss": 3.3688,
       "step": 30
     },
     {
       "epoch": 0.28,
       "learning_rate": 2.8e-05,
-      "loss": 3.4207,
+      "loss": 3.4032,
       "step": 31
     },
     {
       "epoch": 0.29,
       "learning_rate": 2.9000000000000004e-05,
-      "loss": 3.4373,
+      "loss": 3.3115,
       "step": 32
     },
     {
       "epoch": 0.3,
       "learning_rate": 3e-05,
-      "loss": 3.2939,
+      "loss": 3.2587,
       "step": 33
     },
     {
       "epoch": 0.31,
       "learning_rate": 3.1e-05,
-      "loss": 3.2784,
+      "loss": 3.2058,
       "step": 34
     },
     {
       "epoch": 0.32,
       "learning_rate": 3.2e-05,
-      "loss": 3.2101,
+      "loss": 3.2291,
       "step": 35
     },
     {
       "epoch": 0.33,
       "learning_rate": 3.3e-05,
-      "loss": 3.1498,
+      "loss": 3.2174,
       "step": 36
     },
     {
       "epoch": 0.34,
       "learning_rate": 3.4000000000000007e-05,
-      "loss": 3.2037,
+      "loss": 3.1488,
       "step": 37
     },
     {
       "epoch": 0.35,
       "learning_rate": 3.5000000000000004e-05,
-      "loss": 3.1997,
+      "loss": 3.2666,
       "step": 38
     },
     {
       "epoch": 0.36,
       "learning_rate": 3.6e-05,
-      "loss": 3.1802,
+      "loss": 3.3063,
       "step": 39
     },
     {
       "epoch": 0.37,
       "learning_rate": 3.7e-05,
-      "loss": 3.1123,
+      "loss": 3.1416,
       "step": 40
     },
     {
       "epoch": 0.38,
       "learning_rate": 3.8e-05,
-      "loss": 3.1104,
+      "loss": 3.2415,
       "step": 41
     },
     {
       "epoch": 0.39,
       "learning_rate": 3.9e-05,
-      "loss": 3.1473,
+      "loss": 3.2565,
       "step": 42
     },
     {
       "epoch": 0.39,
       "learning_rate": 4e-05,
-      "loss": 3.1615,
+      "loss": 3.2297,
       "step": 43
     },
     {
       "epoch": 0.4,
       "learning_rate": 4.1e-05,
-      "loss": 3.1428,
+      "loss": 3.117,
       "step": 44
     },
     {
       "epoch": 0.41,
       "learning_rate": 4.2000000000000004e-05,
-      "loss": 3.0819,
+      "loss": 3.0919,
       "step": 45
     },
     {
       "epoch": 0.42,
       "learning_rate": 4.2999999999999995e-05,
-      "loss": 3.1113,
+      "loss": 3.1291,
       "step": 46
     },
     {
       "epoch": 0.43,
       "learning_rate": 4.4e-05,
-      "loss": 3.0946,
+      "loss": 3.1522,
       "step": 47
     },
     {
       "epoch": 0.44,
       "learning_rate": 4.4999999999999996e-05,
-      "loss": 3.0418,
+      "loss": 3.1181,
       "step": 48
     },
     {
       "epoch": 0.45,
       "learning_rate": 4.6e-05,
-      "loss": 3.0446,
+      "loss": 3.0412,
       "step": 49
     },
     {
       "epoch": 0.46,
       "learning_rate": 4.7000000000000004e-05,
-      "loss": 3.0657,
+      "loss": 3.0902,
       "step": 50
     },
     {
       "epoch": 0.47,
       "learning_rate": 4.8e-05,
-      "loss": 3.079,
+      "loss": 3.0485,
       "step": 51
     },
     {
       "epoch": 0.48,
       "learning_rate": 4.9000000000000005e-05,
-      "loss": 3.0337,
+      "loss": 3.0649,
       "step": 52
     },
     {
       "epoch": 0.49,
       "learning_rate": 5e-05,
-      "loss": 2.9961,
+      "loss": 3.0919,
       "step": 53
     },
     {
       "epoch": 0.5,
       "learning_rate": 5.1e-05,
-      "loss": 3.0675,
+      "loss": 3.0901,
       "step": 54
     },
     {
       "epoch": 0.5,
       "learning_rate": 5.2e-05,
-      "loss": 3.2045,
+      "loss": 2.9885,
       "step": 55
     },
     {
       "epoch": 0.51,
       "learning_rate": 5.3e-05,
-      "loss": 3.1336,
+      "loss": 2.9587,
       "step": 56
     },
     {
       "epoch": 0.52,
       "learning_rate": 5.4e-05,
-      "loss": 2.8979,
+      "loss": 2.9406,
       "step": 57
     },
     {
       "epoch": 0.53,
       "learning_rate": 5.5e-05,
-      "loss": 3.0432,
+      "loss": 2.9432,
       "step": 58
     },
     {
       "epoch": 0.54,
       "learning_rate": 5.6e-05,
-      "loss": 3.1173,
+      "loss": 2.8637,
       "step": 59
     },
     {
       "epoch": 0.55,
       "learning_rate": 5.7e-05,
-      "loss": 2.8984,
+      "loss": 2.8819,
       "step": 60
     },
     {
       "epoch": 0.56,
       "learning_rate": 5.800000000000001e-05,
-      "loss": 3.0313,
+      "loss": 2.871,
       "step": 61
     },
     {
       "epoch": 0.57,
       "learning_rate": 5.9e-05,
-      "loss": 3.0274,
+      "loss": 2.7446,
       "step": 62
     },
     {
       "epoch": 0.58,
       "learning_rate": 6e-05,
-      "loss": 2.8286,
+      "loss": 2.6881,
       "step": 63
     },
     {
       "epoch": 0.59,
       "learning_rate": 6.1e-05,
-      "loss": 2.7637,
+      "loss": 2.619,
       "step": 64
     },
     {
       "epoch": 0.6,
       "learning_rate": 6.2e-05,
-      "loss": 2.7111,
+      "loss": 2.4788,
       "step": 65
     },
     {
       "epoch": 0.61,
       "learning_rate": 6.3e-05,
-      "loss": 2.6568,
+      "loss": 2.3961,
       "step": 66
     },
     {
       "epoch": 0.61,
       "learning_rate": 6.4e-05,
-      "loss": 2.5761,
+      "loss": 2.4608,
       "step": 67
     },
     {
       "epoch": 0.62,
       "learning_rate": 6.500000000000001e-05,
-      "loss": 2.594,
+      "loss": 2.2703,
       "step": 68
     },
     {
       "epoch": 0.63,
       "learning_rate": 6.6e-05,
-      "loss": 2.5337,
+      "loss": 2.2759,
       "step": 69
     },
     {
       "epoch": 0.64,
       "learning_rate": 6.7e-05,
-      "loss": 2.328,
+      "loss": 2.1375,
       "step": 70
     },
     {
       "epoch": 0.65,
       "learning_rate": 6.800000000000001e-05,
-      "loss": 2.2416,
+      "loss": 1.8752,
       "step": 71
     },
     {
       "epoch": 0.66,
       "learning_rate": 6.900000000000001e-05,
-      "loss": 2.1625,
+      "loss": 1.8352,
       "step": 72
     },
     {
       "epoch": 0.67,
       "learning_rate": 7.000000000000001e-05,
-      "loss": 2.0447,
+      "loss": 1.7613,
       "step": 73
     },
     {
       "epoch": 0.68,
       "learning_rate": 7.099999999999999e-05,
-      "loss": 1.8691,
+      "loss": 1.6099,
       "step": 74
     },
     {
       "epoch": 0.69,
       "learning_rate": 7.2e-05,
-      "loss": 1.8868,
+      "loss": 1.5892,
       "step": 75
     },
     {
       "epoch": 0.7,
       "learning_rate": 7.3e-05,
-      "loss": 1.6314,
+      "loss": 1.5142,
       "step": 76
     },
     {
       "epoch": 0.71,
       "learning_rate": 7.4e-05,
-      "loss": 1.6226,
+      "loss": 1.415,
       "step": 77
     },
     {
       "epoch": 0.72,
       "learning_rate": 7.5e-05,
-      "loss": 1.7013,
+      "loss": 1.5443,
       "step": 78
     },
     {
       "epoch": 0.72,
       "learning_rate": 7.6e-05,
-      "loss": 1.5557,
+      "loss": 1.4341,
       "step": 79
     },
     {
       "epoch": 0.73,
       "learning_rate": 7.7e-05,
-      "loss": 1.4843,
+      "loss": 1.4534,
       "step": 80
     },
     {
       "epoch": 0.74,
       "learning_rate": 7.8e-05,
-      "loss": 1.563,
+      "loss": 1.4176,
       "step": 81
     },
     {
       "epoch": 0.75,
       "learning_rate": 7.9e-05,
-      "loss": 1.3356,
+      "loss": 1.4057,
       "step": 82
     },
     {
       "epoch": 0.76,
       "learning_rate": 8e-05,
-      "loss": 1.3477,
+      "loss": 1.28,
       "step": 83
     },
     {
       "epoch": 0.77,
       "learning_rate": 8.1e-05,
-      "loss": 1.2732,
+      "loss": 1.291,
       "step": 84
     },
     {
       "epoch": 0.78,
       "learning_rate": 8.2e-05,
-      "loss": 1.1838,
+      "loss": 1.2415,
       "step": 85
     },
     {
       "epoch": 0.79,
       "learning_rate": 8.300000000000001e-05,
-      "loss": 1.1471,
+      "loss": 1.1426,
       "step": 86
     },
     {
       "epoch": 0.8,
       "learning_rate": 8.400000000000001e-05,
-      "loss": 1.0627,
+      "loss": 1.1204,
       "step": 87
     },
     {
       "epoch": 0.81,
       "learning_rate": 8.5e-05,
-      "loss": 0.9847,
+      "loss": 1.0339,
       "step": 88
     },
     {
       "epoch": 0.82,
       "learning_rate": 8.599999999999999e-05,
-      "loss": 0.9781,
+      "loss": 1.0081,
       "step": 89
     },
     {
       "epoch": 0.83,
       "learning_rate": 8.7e-05,
-      "loss": 0.96,
+      "loss": 0.9979,
       "step": 90
     },
     {
       "epoch": 0.83,
       "learning_rate": 8.8e-05,
-      "loss": 0.9061,
+      "loss": 0.9104,
       "step": 91
     },
     {
       "epoch": 0.84,
       "learning_rate": 8.9e-05,
-      "loss": 0.9697,
+      "loss": 0.9786,
       "step": 92
     },
     {
       "epoch": 0.85,
       "learning_rate": 8.999999999999999e-05,
-      "loss": 1.0049,
+      "loss": 0.9904,
       "step": 93
     },
     {
       "epoch": 0.86,
       "learning_rate": 9.1e-05,
-      "loss": 1.1034,
+      "loss": 1.0777,
       "step": 94
     },
     {
       "epoch": 0.87,
       "learning_rate": 9.2e-05,
-      "loss": 0.9069,
+      "loss": 0.9018,
       "step": 95
     },
     {
       "epoch": 0.88,
       "learning_rate": 9.3e-05,
-      "loss": 1.026,
+      "loss": 0.9402,
       "step": 96
     },
     {
       "epoch": 0.89,
       "learning_rate": 9.400000000000001e-05,
-      "loss": 1.0734,
+      "loss": 1.0584,
       "step": 97
     },
     {
       "epoch": 0.9,
       "learning_rate": 9.5e-05,
-      "loss": 0.9626,
+      "loss": 1.0325,
       "step": 98
     },
     {
       "epoch": 0.91,
       "learning_rate": 9.6e-05,
-      "loss": 0.8731,
+      "loss": 0.7905,
       "step": 99
     },
     {
       "epoch": 0.92,
       "learning_rate": 9.7e-05,
-      "loss": 1.0878,
+      "loss": 1.1899,
       "step": 100
     },
     {
       "epoch": 0.93,
       "learning_rate": 9.800000000000001e-05,
-      "loss": 0.9423,
+      "loss": 0.8969,
       "step": 101
     },
     {
       "epoch": 0.94,
       "learning_rate": 9.900000000000001e-05,
-      "loss": 0.9755,
+      "loss": 0.9911,
       "step": 102
     },
     {
       "epoch": 0.94,
       "learning_rate": 0.0001,
-      "loss": 0.9086,
+      "loss": 0.9118,
       "step": 103
     },
     {
       "epoch": 0.95,
       "learning_rate": 0.000101,
-      "loss": 0.8551,
+      "loss": 0.8814,
       "step": 104
     },
     {
       "epoch": 0.96,
       "learning_rate": 0.000102,
-      "loss": 1.0771,
+      "loss": 0.9854,
       "step": 105
     },
     {
       "epoch": 0.97,
       "learning_rate": 0.000103,
-      "loss": 0.9983,
+      "loss": 0.9928,
       "step": 106
     },
     {
       "epoch": 0.98,
       "learning_rate": 0.000104,
-      "loss": 0.9869,
+      "loss": 1.0183,
       "step": 107
     },
     {
       "epoch": 0.99,
       "learning_rate": 0.000105,
-      "loss": 1.1931,
+      "loss": 1.0125,
       "step": 108
     },
     {
       "epoch": 1.0,
       "learning_rate": 0.000106,
-      "loss": 1.1017,
+      "loss": 1.199,
       "step": 109
     },
     {
-      "epoch": 1.0,
-      "step": 109,
-      "total_flos": 1.3366257233080689e+18,
-      "train_loss": 4.048810622560868,
-      "train_runtime": 250.3759,
-      "train_samples_per_second": 13.891,
-      "train_steps_per_second": 0.435
+      "epoch": 1.01,
+      "learning_rate": 0.000107,
+      "loss": 0.855,
+      "step": 110
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000108,
+      "loss": 0.8506,
+      "step": 111
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.000109,
+      "loss": 1.0619,
+      "step": 112
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00011,
+      "loss": 0.7957,
+      "step": 113
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.000111,
+      "loss": 0.8862,
+      "step": 114
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.000112,
+      "loss": 0.7976,
+      "step": 115
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00011300000000000001,
+      "loss": 1.049,
+      "step": 116
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.000114,
+      "loss": 0.8552,
+      "step": 117
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.000115,
+      "loss": 0.7494,
+      "step": 118
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00011600000000000001,
+      "loss": 0.7377,
+      "step": 119
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00011700000000000001,
+      "loss": 0.8836,
+      "step": 120
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.000118,
+      "loss": 0.8784,
+      "step": 121
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00011899999999999999,
+      "loss": 0.7728,
+      "step": 122
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00012,
+      "loss": 0.7766,
+      "step": 123
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.000121,
+      "loss": 0.9031,
+      "step": 124
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.000122,
+      "loss": 0.8105,
+      "step": 125
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000123,
+      "loss": 0.7673,
+      "step": 126
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000124,
+      "loss": 0.7829,
+      "step": 127
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000125,
+      "loss": 0.8379,
+      "step": 128
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.000126,
+      "loss": 1.0519,
+      "step": 129
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.000127,
+      "loss": 0.9201,
+      "step": 130
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.000128,
+      "loss": 0.8556,
+      "step": 131
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00012900000000000002,
+      "loss": 0.7174,
+      "step": 132
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 0.7999,
+      "step": 133
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.000131,
+      "loss": 0.8046,
+      "step": 134
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.000132,
+      "loss": 0.877,
+      "step": 135
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.000133,
+      "loss": 0.9048,
+      "step": 136
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.000134,
+      "loss": 0.8512,
+      "step": 137
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.000135,
+      "loss": 0.7844,
+      "step": 138
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00013600000000000003,
+      "loss": 0.8066,
+      "step": 139
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00013700000000000002,
+      "loss": 0.868,
+      "step": 140
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00013800000000000002,
+      "loss": 0.8737,
+      "step": 141
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00013900000000000002,
+      "loss": 0.7842,
+      "step": 142
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 0.7213,
+      "step": 143
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 0.6951,
+      "step": 144
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00014199999999999998,
+      "loss": 0.87,
+      "step": 145
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014299999999999998,
+      "loss": 0.7681,
+      "step": 146
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.000144,
+      "loss": 0.8044,
+      "step": 147
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.000145,
+      "loss": 0.8936,
+      "step": 148
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.000146,
+      "loss": 0.7405,
+      "step": 149
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.000147,
+      "loss": 0.9331,
+      "step": 150
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.000148,
+      "loss": 0.8906,
+      "step": 151
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.000149,
+      "loss": 0.919,
+      "step": 152
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00015,
+      "loss": 0.6376,
+      "step": 153
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.000151,
+      "loss": 0.8851,
+      "step": 154
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.000152,
+      "loss": 0.948,
+      "step": 155
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.000153,
+      "loss": 0.8521,
+      "step": 156
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.000154,
+      "loss": 0.9448,
+      "step": 157
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000155,
+      "loss": 0.8974,
+      "step": 158
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.000156,
+      "loss": 0.8137,
+      "step": 159
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.000157,
+      "loss": 0.8248,
+      "step": 160
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.000158,
+      "loss": 0.9393,
+      "step": 161
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00015900000000000002,
+      "loss": 0.9912,
+      "step": 162
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00016,
+      "loss": 0.8831,
+      "step": 163
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.000161,
+      "loss": 0.8385,
+      "step": 164
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.000162,
+      "loss": 0.9737,
+      "step": 165
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.000163,
+      "loss": 0.9007,
+      "step": 166
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.000164,
+      "loss": 0.7576,
+      "step": 167
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.000165,
+      "loss": 0.8212,
+      "step": 168
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00016600000000000002,
+      "loss": 0.9865,
+      "step": 169
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00016700000000000002,
+      "loss": 0.7642,
+      "step": 170
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00016800000000000002,
+      "loss": 0.7457,
+      "step": 171
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00016900000000000002,
+      "loss": 0.8548,
+      "step": 172
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00017,
+      "loss": 0.8308,
+      "step": 173
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.000171,
+      "loss": 0.8595,
+      "step": 174
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00017199999999999998,
+      "loss": 0.8856,
+      "step": 175
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.000173,
+      "loss": 0.7756,
+      "step": 176
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.000174,
+      "loss": 0.7542,
+      "step": 177
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.000175,
+      "loss": 0.9347,
+      "step": 178
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000176,
+      "loss": 0.8227,
+      "step": 179
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.000177,
+      "loss": 0.8607,
+      "step": 180
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.000178,
+      "loss": 0.7882,
+      "step": 181
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.000179,
+      "loss": 0.9872,
+      "step": 182
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.8981,
+      "step": 183
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.000181,
+      "loss": 0.7831,
+      "step": 184
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.000182,
+      "loss": 0.8964,
+      "step": 185
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.000183,
+      "loss": 0.8213,
+      "step": 186
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.000184,
+      "loss": 0.9628,
+      "step": 187
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.000185,
+      "loss": 0.8119,
+      "step": 188
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.000186,
+      "loss": 1.0963,
+      "step": 189
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.000187,
+      "loss": 1.015,
+      "step": 190
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00018800000000000002,
+      "loss": 0.8741,
+      "step": 191
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.000189,
+      "loss": 0.8115,
+      "step": 192
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00019,
+      "loss": 0.8478,
+      "step": 193
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.000191,
+      "loss": 0.9544,
+      "step": 194
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.000192,
+      "loss": 0.8794,
+      "step": 195
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.000193,
+      "loss": 0.8016,
+      "step": 196
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.000194,
+      "loss": 0.7901,
+      "step": 197
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 0.8481,
+      "step": 198
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00019600000000000002,
+      "loss": 0.9701,
+      "step": 199
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00019700000000000002,
+      "loss": 0.7847,
+      "step": 200
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00019800000000000002,
+      "loss": 0.8618,
+      "step": 201
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.000199,
+      "loss": 0.8816,
+      "step": 202
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002,
+      "loss": 0.8356,
+      "step": 203
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.000201,
+      "loss": 0.8051,
+      "step": 204
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.000202,
+      "loss": 0.885,
+      "step": 205
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00020300000000000003,
+      "loss": 0.8644,
+      "step": 206
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.000204,
+      "loss": 0.9245,
+      "step": 207
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.000205,
+      "loss": 0.821,
+      "step": 208
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.000206,
+      "loss": 0.8055,
+      "step": 209
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.000207,
+      "loss": 0.7631,
+      "step": 210
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.000208,
+      "loss": 0.828,
+      "step": 211
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00020899999999999998,
+      "loss": 0.8925,
+      "step": 212
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00021,
+      "loss": 0.9761,
+      "step": 213
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.000211,
+      "loss": 0.9328,
+      "step": 214
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.000212,
+      "loss": 0.9403,
+      "step": 215
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.000213,
+      "loss": 1.1516,
+      "step": 216
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.000214,
+      "loss": 0.9257,
+      "step": 217
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.000215,
+      "loss": 0.8389,
+      "step": 218
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.000216,
+      "loss": 0.9406,
+      "step": 219
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.00021700000000000002,
+      "loss": 0.7932,
+      "step": 220
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.000218,
+      "loss": 0.9084,
+      "step": 221
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.000219,
+      "loss": 0.8446,
+      "step": 222
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00022,
+      "loss": 0.7847,
+      "step": 223
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.000221,
+      "loss": 0.7572,
+      "step": 224
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.000222,
+      "loss": 0.8638,
+      "step": 225
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.000223,
+      "loss": 0.8418,
+      "step": 226
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.000224,
+      "loss": 0.7481,
+      "step": 227
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 0.826,
+      "step": 228
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00022600000000000002,
+      "loss": 0.7507,
+      "step": 229
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00022700000000000002,
+      "loss": 0.7673,
+      "step": 230
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.000228,
+      "loss": 0.7812,
+      "step": 231
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.000229,
+      "loss": 0.8013,
+      "step": 232
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.00023,
+      "loss": 0.9534,
+      "step": 233
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.000231,
+      "loss": 0.9536,
+      "step": 234
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.00023200000000000003,
+      "loss": 0.6724,
+      "step": 235
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00023300000000000003,
+      "loss": 0.5844,
+      "step": 236
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00023400000000000002,
+      "loss": 0.9228,
+      "step": 237
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.000235,
+      "loss": 0.6508,
+      "step": 238
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.000236,
+      "loss": 0.7184,
+      "step": 239
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.000237,
+      "loss": 0.8169,
+      "step": 240
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.00023799999999999998,
+      "loss": 0.9676,
+      "step": 241
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.00023899999999999998,
+      "loss": 0.7811,
+      "step": 242
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00024,
+      "loss": 0.8423,
+      "step": 243
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.000241,
+      "loss": 0.9599,
+      "step": 244
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.000242,
+      "loss": 0.8492,
+      "step": 245
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.000243,
+      "loss": 0.8892,
+      "step": 246
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.000244,
+      "loss": 0.8491,
+      "step": 247
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.000245,
+      "loss": 0.9339,
+      "step": 248
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.000246,
+      "loss": 0.9203,
+      "step": 249
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.000247,
+      "loss": 0.798,
+      "step": 250
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.000248,
+      "loss": 0.7288,
+      "step": 251
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.000249,
+      "loss": 0.9401,
+      "step": 252
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00025,
+      "loss": 0.7373,
+      "step": 253
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00025100000000000003,
+      "loss": 0.7714,
+      "step": 254
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.000252,
+      "loss": 0.7332,
+      "step": 255
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.000253,
+      "loss": 0.9206,
+      "step": 256
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.000254,
+      "loss": 0.8617,
+      "step": 257
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.000255,
+      "loss": 0.7401,
+      "step": 258
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.000256,
+      "loss": 0.8888,
+      "step": 259
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.000257,
+      "loss": 0.7575,
+      "step": 260
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00025800000000000004,
+      "loss": 0.8186,
+      "step": 261
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.000259,
+      "loss": 0.9301,
+      "step": 262
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 0.8435,
+      "step": 263
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.000261,
+      "loss": 0.9087,
+      "step": 264
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.000262,
+      "loss": 0.9546,
+      "step": 265
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.000263,
+      "loss": 0.9296,
+      "step": 266
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.000264,
+      "loss": 0.8755,
+      "step": 267
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.00026500000000000004,
+      "loss": 1.0179,
+      "step": 268
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.000266,
+      "loss": 0.9092,
+      "step": 269
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.00026700000000000004,
+      "loss": 1.0174,
+      "step": 270
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.000268,
+      "loss": 0.9106,
+      "step": 271
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.00026900000000000003,
+      "loss": 0.9254,
+      "step": 272
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.00027,
+      "loss": 0.8664,
+      "step": 273
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00027100000000000003,
+      "loss": 0.8069,
+      "step": 274
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 0.8325,
+      "step": 275
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.000273,
+      "loss": 0.8521,
+      "step": 276
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00027400000000000005,
+      "loss": 0.7486,
+      "step": 277
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.000275,
+      "loss": 0.8621,
+      "step": 278
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00027600000000000004,
+      "loss": 0.9327,
+      "step": 279
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.000277,
+      "loss": 0.7462,
+      "step": 280
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.00027800000000000004,
+      "loss": 0.7326,
+      "step": 281
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.000279,
+      "loss": 0.6892,
+      "step": 282
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 0.8842,
+      "step": 283
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00028100000000000005,
+      "loss": 0.9274,
+      "step": 284
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 0.7804,
+      "step": 285
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.000283,
+      "loss": 1.0858,
+      "step": 286
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00028399999999999996,
+      "loss": 1.0578,
+      "step": 287
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.000285,
+      "loss": 0.9239,
+      "step": 288
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.00028599999999999996,
+      "loss": 0.9414,
+      "step": 289
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.000287,
+      "loss": 0.7785,
+      "step": 290
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.000288,
+      "loss": 0.9458,
+      "step": 291
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.000289,
+      "loss": 0.8933,
+      "step": 292
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00029,
+      "loss": 0.956,
+      "step": 293
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 0.8677,
+      "step": 294
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.000292,
+      "loss": 0.8051,
+      "step": 295
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00029299999999999997,
+      "loss": 0.8529,
+      "step": 296
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.000294,
+      "loss": 1.0472,
+      "step": 297
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.000295,
+      "loss": 1.239,
+      "step": 298
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.000296,
+      "loss": 0.8557,
+      "step": 299
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.000297,
+      "loss": 1.0188,
+      "step": 300
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.000298,
+      "loss": 0.939,
+      "step": 301
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.000299,
+      "loss": 0.9302,
+      "step": 302
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0003,
+      "loss": 1.0368,
+      "step": 303
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.000301,
+      "loss": 0.9943,
+      "step": 304
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.000302,
+      "loss": 0.9268,
+      "step": 305
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.000303,
+      "loss": 0.8335,
+      "step": 306
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.000304,
+      "loss": 0.9226,
+      "step": 307
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.000305,
+      "loss": 0.8811,
+      "step": 308
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.000306,
+      "loss": 0.9947,
+      "step": 309
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.000307,
+      "loss": 0.8541,
+      "step": 310
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.000308,
+      "loss": 0.8974,
+      "step": 311
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.00030900000000000003,
+      "loss": 1.0284,
+      "step": 312
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00031,
+      "loss": 1.0208,
+      "step": 313
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.000311,
+      "loss": 1.0817,
+      "step": 314
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.000312,
+      "loss": 0.9175,
+      "step": 315
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.000313,
+      "loss": 0.9441,
+      "step": 316
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.000314,
+      "loss": 0.9489,
+      "step": 317
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.000315,
+      "loss": 1.0831,
+      "step": 318
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.000316,
+      "loss": 0.9136,
+      "step": 319
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.000317,
+      "loss": 1.0493,
+      "step": 320
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00031800000000000003,
+      "loss": 0.9049,
+      "step": 321
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.000319,
+      "loss": 0.895,
+      "step": 322
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00032,
+      "loss": 1.1026,
+      "step": 323
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.000321,
+      "loss": 0.9457,
+      "step": 324
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.000322,
+      "loss": 1.1172,
+      "step": 325
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.000323,
+      "loss": 1.138,
+      "step": 326
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.000324,
+      "loss": 1.0015,
+      "step": 327
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 0.9543,
+      "step": 328
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.000326,
+      "loss": 0.865,
+      "step": 329
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00032700000000000003,
+      "loss": 0.8718,
+      "step": 330
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.000328,
+      "loss": 0.8832,
+      "step": 331
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00032900000000000003,
+      "loss": 0.9233,
+      "step": 332
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.00033,
+      "loss": 0.9403,
+      "step": 333
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.000331,
+      "loss": 0.8338,
+      "step": 334
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00033200000000000005,
+      "loss": 0.7815,
+      "step": 335
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.000333,
+      "loss": 0.7849,
+      "step": 336
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00033400000000000004,
+      "loss": 0.8684,
+      "step": 337
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.000335,
+      "loss": 0.9105,
+      "step": 338
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00033600000000000004,
+      "loss": 0.7736,
+      "step": 339
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.000337,
+      "loss": 0.7805,
+      "step": 340
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.00033800000000000003,
+      "loss": 0.935,
+      "step": 341
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00033900000000000005,
+      "loss": 0.8801,
+      "step": 342
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00034,
+      "loss": 0.7727,
+      "step": 343
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00034100000000000005,
+      "loss": 0.9658,
+      "step": 344
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.000342,
+      "loss": 0.8092,
+      "step": 345
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00034300000000000004,
+      "loss": 0.859,
+      "step": 346
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00034399999999999996,
+      "loss": 0.9139,
+      "step": 347
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.000345,
+      "loss": 0.9811,
+      "step": 348
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.000346,
+      "loss": 1.0785,
+      "step": 349
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.000347,
+      "loss": 0.9783,
+      "step": 350
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.000348,
+      "loss": 0.8845,
+      "step": 351
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00034899999999999997,
+      "loss": 0.8462,
+      "step": 352
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00035,
+      "loss": 1.0202,
+      "step": 353
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00035099999999999997,
+      "loss": 0.9083,
+      "step": 354
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.000352,
+      "loss": 0.9835,
+      "step": 355
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.00035299999999999996,
+      "loss": 0.9465,
+      "step": 356
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.000354,
+      "loss": 1.1778,
+      "step": 357
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.000355,
+      "loss": 0.926,
+      "step": 358
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.000356,
+      "loss": 0.8306,
+      "step": 359
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.000357,
+      "loss": 0.8872,
+      "step": 360
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.000358,
+      "loss": 0.9393,
+      "step": 361
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.000359,
+      "loss": 0.7678,
+      "step": 362
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 0.8853,
+      "step": 363
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.000361,
+      "loss": 0.9414,
+      "step": 364
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.000362,
+      "loss": 0.9436,
+      "step": 365
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.000363,
+      "loss": 0.8973,
+      "step": 366
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.000364,
+      "loss": 0.8938,
+      "step": 367
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.000365,
+      "loss": 0.9352,
+      "step": 368
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.000366,
+      "loss": 0.8689,
+      "step": 369
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.000367,
+      "loss": 1.0152,
+      "step": 370
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.000368,
+      "loss": 0.9203,
+      "step": 371
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.000369,
+      "loss": 0.8105,
+      "step": 372
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00037,
+      "loss": 1.0456,
+      "step": 373
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.000371,
+      "loss": 0.8978,
+      "step": 374
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.000372,
+      "loss": 0.9478,
+      "step": 375
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.000373,
+      "loss": 1.0347,
+      "step": 376
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.000374,
+      "loss": 0.9996,
+      "step": 377
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.000375,
+      "loss": 1.1262,
+      "step": 378
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.00037600000000000003,
+      "loss": 1.1398,
+      "step": 379
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.000377,
+      "loss": 1.0633,
+      "step": 380
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.000378,
+      "loss": 0.9757,
+      "step": 381
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.000379,
+      "loss": 1.0895,
+      "step": 382
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00038,
+      "loss": 0.9751,
+      "step": 383
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.000381,
+      "loss": 1.0236,
+      "step": 384
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.000382,
+      "loss": 0.9207,
+      "step": 385
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.00038300000000000004,
+      "loss": 0.9935,
+      "step": 386
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.000384,
+      "loss": 0.8587,
+      "step": 387
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00038500000000000003,
+      "loss": 0.9348,
+      "step": 388
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.000386,
+      "loss": 0.842,
+      "step": 389
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00038700000000000003,
+      "loss": 1.0173,
+      "step": 390
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.000388,
+      "loss": 0.9139,
+      "step": 391
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.000389,
+      "loss": 1.0815,
+      "step": 392
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 0.956,
+      "step": 393
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.000391,
+      "loss": 0.9945,
+      "step": 394
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 0.9918,
+      "step": 395
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.000393,
+      "loss": 0.9994,
+      "step": 396
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.00039400000000000004,
+      "loss": 1.1402,
+      "step": 397
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.000395,
+      "loss": 1.0114,
+      "step": 398
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00039600000000000003,
+      "loss": 0.8996,
+      "step": 399
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00039700000000000005,
+      "loss": 0.9863,
+      "step": 400
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.000398,
+      "loss": 0.9065,
+      "step": 401
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00039900000000000005,
+      "loss": 1.0639,
+      "step": 402
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0004,
+      "loss": 1.071,
+      "step": 403
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.00040100000000000004,
+      "loss": 1.014,
+      "step": 404
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.000402,
+      "loss": 0.8809,
+      "step": 405
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00040300000000000004,
+      "loss": 1.0018,
+      "step": 406
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.000404,
+      "loss": 1.0392,
+      "step": 407
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00040500000000000003,
+      "loss": 1.1197,
+      "step": 408
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00040600000000000006,
+      "loss": 1.0419,
+      "step": 409
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00040699999999999997,
+      "loss": 0.9237,
+      "step": 410
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.000408,
+      "loss": 0.934,
+      "step": 411
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00040899999999999997,
+      "loss": 0.9585,
+      "step": 412
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00041,
+      "loss": 0.9803,
+      "step": 413
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00041099999999999996,
+      "loss": 0.954,
+      "step": 414
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.000412,
+      "loss": 0.9587,
+      "step": 415
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.000413,
+      "loss": 0.8709,
+      "step": 416
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.000414,
+      "loss": 0.8709,
+      "step": 417
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.000415,
+      "loss": 1.003,
+      "step": 418
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.000416,
+      "loss": 1.0172,
+      "step": 419
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.000417,
+      "loss": 0.9181,
+      "step": 420
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00041799999999999997,
+      "loss": 1.0221,
+      "step": 421
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.000419,
+      "loss": 1.0124,
+      "step": 422
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00042,
+      "loss": 0.9793,
+      "step": 423
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.000421,
+      "loss": 1.097,
+      "step": 424
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.000422,
+      "loss": 0.8788,
+      "step": 425
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.000423,
+      "loss": 1.0303,
+      "step": 426
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.000424,
+      "loss": 1.2121,
+      "step": 427
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.000425,
+      "loss": 0.9664,
+      "step": 428
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.000426,
+      "loss": 1.0561,
+      "step": 429
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.000427,
+      "loss": 0.9694,
+      "step": 430
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.000428,
+      "loss": 0.9438,
+      "step": 431
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.000429,
+      "loss": 1.036,
+      "step": 432
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00043,
+      "loss": 1.0136,
+      "step": 433
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.000431,
+      "loss": 1.1031,
+      "step": 434
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.000432,
+      "loss": 1.1379,
+      "step": 435
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.000433,
+      "loss": 1.0827,
+      "step": 436
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.00043400000000000003,
+      "loss": 0.9154,
+      "step": 437
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.000435,
+      "loss": 0.9827,
+      "step": 438
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.000436,
+      "loss": 0.9376,
+      "step": 439
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.000437,
+      "loss": 1.1454,
+      "step": 440
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.000438,
+      "loss": 0.9389,
+      "step": 441
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.000439,
+      "loss": 0.947,
+      "step": 442
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00044,
+      "loss": 0.9499,
+      "step": 443
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.000441,
+      "loss": 0.9558,
+      "step": 444
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.000442,
+      "loss": 0.8672,
+      "step": 445
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.00044300000000000003,
+      "loss": 0.8519,
+      "step": 446
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.000444,
+      "loss": 0.9398,
+      "step": 447
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.00044500000000000003,
+      "loss": 0.9419,
+      "step": 448
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.000446,
+      "loss": 0.9816,
+      "step": 449
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.000447,
+      "loss": 1.0649,
+      "step": 450
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.000448,
+      "loss": 1.0448,
+      "step": 451
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.000449,
+      "loss": 1.0459,
+      "step": 452
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 1.0103,
+      "step": 453
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.000451,
+      "loss": 0.9537,
+      "step": 454
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 1.0611,
+      "step": 455
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.000453,
+      "loss": 1.1025,
+      "step": 456
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.00045400000000000003,
+      "loss": 1.0208,
+      "step": 457
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.000455,
+      "loss": 0.9947,
+      "step": 458
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.000456,
+      "loss": 0.9338,
+      "step": 459
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.00045700000000000005,
+      "loss": 0.992,
+      "step": 460
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.000458,
+      "loss": 1.123,
+      "step": 461
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00045900000000000004,
+      "loss": 0.9532,
+      "step": 462
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.00046,
+      "loss": 1.1185,
+      "step": 463
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00046100000000000004,
+      "loss": 1.0723,
+      "step": 464
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.000462,
+      "loss": 0.95,
+      "step": 465
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00046300000000000003,
+      "loss": 1.0639,
+      "step": 466
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00046400000000000006,
+      "loss": 0.9463,
+      "step": 467
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.000465,
+      "loss": 1.0093,
+      "step": 468
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00046600000000000005,
+      "loss": 0.8967,
+      "step": 469
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.000467,
+      "loss": 0.9979,
+      "step": 470
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.00046800000000000005,
+      "loss": 0.8109,
+      "step": 471
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00046899999999999996,
+      "loss": 0.9047,
+      "step": 472
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00047,
+      "loss": 0.7607,
+      "step": 473
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.000471,
+      "loss": 0.9682,
+      "step": 474
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.000472,
+      "loss": 0.9564,
+      "step": 475
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.000473,
+      "loss": 0.8708,
+      "step": 476
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.000474,
+      "loss": 0.8592,
+      "step": 477
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.000475,
+      "loss": 0.9601,
+      "step": 478
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00047599999999999997,
+      "loss": 1.1149,
+      "step": 479
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.000477,
+      "loss": 0.9929,
+      "step": 480
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.00047799999999999996,
+      "loss": 0.8654,
+      "step": 481
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.000479,
+      "loss": 1.0139,
+      "step": 482
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.00048,
+      "loss": 0.9918,
+      "step": 483
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.000481,
+      "loss": 1.0528,
+      "step": 484
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.000482,
+      "loss": 0.9347,
+      "step": 485
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.000483,
+      "loss": 1.0451,
+      "step": 486
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.000484,
+      "loss": 1.0556,
+      "step": 487
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.00048499999999999997,
+      "loss": 0.9726,
+      "step": 488
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.000486,
+      "loss": 1.1444,
+      "step": 489
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.000487,
+      "loss": 1.2943,
+      "step": 490
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.000488,
+      "loss": 1.1617,
+      "step": 491
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.000489,
+      "loss": 1.0871,
+      "step": 492
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.00049,
+      "loss": 1.1135,
+      "step": 493
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.000491,
+      "loss": 0.9894,
+      "step": 494
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.000492,
+      "loss": 1.0445,
+      "step": 495
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.0004930000000000001,
+      "loss": 1.129,
+      "step": 496
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.000494,
+      "loss": 0.877,
+      "step": 497
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.000495,
+      "loss": 1.0675,
+      "step": 498
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.000496,
+      "loss": 1.1802,
+      "step": 499
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.000497,
+      "loss": 1.0779,
+      "step": 500
+    },
+    {
+      "epoch": 4.59,
+      "eval_cer": 0.23541582332568703,
+      "eval_loss": 0.8260353803634644,
+      "eval_runtime": 86.0896,
+      "eval_samples_per_second": 19.131,
+      "eval_steps_per_second": 2.393,
+      "eval_wer": 0.7394546011643346,
+      "step": 500
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.000498,
+      "loss": 1.0549,
+      "step": 501
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.000499,
+      "loss": 1.1182,
+      "step": 502
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0005,
+      "loss": 0.9358,
+      "step": 503
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0004999519230769231,
+      "loss": 0.9391,
+      "step": 504
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.0004999038461538461,
+      "loss": 1.0371,
+      "step": 505
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.0004998557692307693,
+      "loss": 1.1175,
+      "step": 506
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0004998076923076923,
+      "loss": 1.1461,
+      "step": 507
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.0004997596153846153,
+      "loss": 0.9591,
+      "step": 508
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0004997115384615385,
+      "loss": 1.1943,
+      "step": 509
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0004996634615384615,
+      "loss": 1.0962,
+      "step": 510
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.0004996153846153846,
+      "loss": 0.9074,
+      "step": 511
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.0004995673076923076,
+      "loss": 1.0457,
+      "step": 512
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0004995192307692308,
+      "loss": 1.0732,
+      "step": 513
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0004994711538461538,
+      "loss": 1.0267,
+      "step": 514
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0004994230769230769,
+      "loss": 1.068,
+      "step": 515
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.000499375,
+      "loss": 1.1664,
+      "step": 516
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.0004993269230769231,
+      "loss": 1.0397,
+      "step": 517
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0004992788461538461,
+      "loss": 1.2514,
+      "step": 518
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.0004992307692307693,
+      "loss": 1.1107,
+      "step": 519
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.0004991826923076923,
+      "loss": 1.1255,
+      "step": 520
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.0004991346153846154,
+      "loss": 1.1985,
+      "step": 521
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.0004990865384615384,
+      "loss": 1.0783,
+      "step": 522
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0004990384615384616,
+      "loss": 1.1404,
+      "step": 523
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.0004989903846153846,
+      "loss": 0.9647,
+      "step": 524
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0004989423076923077,
+      "loss": 1.0654,
+      "step": 525
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.0004988942307692308,
+      "loss": 1.0158,
+      "step": 526
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.0004988461538461539,
+      "loss": 0.9294,
+      "step": 527
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.0004987980769230769,
+      "loss": 0.9563,
+      "step": 528
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.0004987500000000001,
+      "loss": 1.0649,
+      "step": 529
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0004987019230769231,
+      "loss": 0.9494,
+      "step": 530
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.0004986538461538462,
+      "loss": 1.1026,
+      "step": 531
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0004986057692307692,
+      "loss": 0.9647,
+      "step": 532
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.0004985576923076924,
+      "loss": 0.9518,
+      "step": 533
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0004985096153846154,
+      "loss": 1.078,
+      "step": 534
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.0004984615384615384,
+      "loss": 0.9563,
+      "step": 535
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.0004984134615384616,
+      "loss": 0.958,
+      "step": 536
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0004983653846153846,
+      "loss": 1.1003,
+      "step": 537
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0004983173076923077,
+      "loss": 1.1324,
+      "step": 538
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0004982692307692307,
+      "loss": 0.981,
+      "step": 539
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0004982211538461539,
+      "loss": 1.09,
+      "step": 540
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.0004981730769230769,
+      "loss": 0.9591,
+      "step": 541
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.000498125,
+      "loss": 1.1162,
+      "step": 542
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.0004980769230769231,
+      "loss": 1.2226,
+      "step": 543
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.0004980288461538462,
+      "loss": 1.1487,
+      "step": 544
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.0004979807692307692,
+      "loss": 1.1372,
+      "step": 545
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0004979326923076924,
+      "loss": 1.1487,
+      "step": 546
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.0004978846153846154,
+      "loss": 1.0553,
+      "step": 547
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.0004978365384615385,
+      "loss": 1.0949,
+      "step": 548
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.0004977884615384615,
+      "loss": 0.9532,
+      "step": 549
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0004977403846153847,
+      "loss": 1.0265,
+      "step": 550
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0004976923076923077,
+      "loss": 0.8973,
+      "step": 551
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0004976442307692308,
+      "loss": 0.9966,
+      "step": 552
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.0004975961538461539,
+      "loss": 0.9501,
+      "step": 553
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.000497548076923077,
+      "loss": 0.9412,
+      "step": 554
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0004975,
+      "loss": 0.9124,
+      "step": 555
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.0004974519230769232,
+      "loss": 0.9253,
+      "step": 556
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.0004974038461538462,
+      "loss": 1.1252,
+      "step": 557
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.0004973557692307692,
+      "loss": 1.0157,
+      "step": 558
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.0004973076923076923,
+      "loss": 0.8945,
+      "step": 559
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0004972596153846153,
+      "loss": 0.9258,
+      "step": 560
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.0004972115384615385,
+      "loss": 0.8995,
+      "step": 561
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.0004971634615384615,
+      "loss": 1.0049,
+      "step": 562
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.0004971153846153846,
+      "loss": 0.9606,
+      "step": 563
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.0004970673076923076,
+      "loss": 1.0308,
+      "step": 564
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.0004970192307692308,
+      "loss": 0.9769,
+      "step": 565
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.0004969711538461538,
+      "loss": 0.98,
+      "step": 566
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.0004969230769230769,
+      "loss": 1.0529,
+      "step": 567
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.000496875,
+      "loss": 0.8042,
+      "step": 568
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0004968269230769231,
+      "loss": 0.9945,
+      "step": 569
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.0004967788461538461,
+      "loss": 0.9021,
+      "step": 570
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0004967307692307693,
+      "loss": 1.006,
+      "step": 571
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.0004966826923076923,
+      "loss": 0.9786,
+      "step": 572
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.0004966346153846154,
+      "loss": 1.0802,
+      "step": 573
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0004965865384615384,
+      "loss": 1.0395,
+      "step": 574
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.0004965384615384616,
+      "loss": 0.9447,
+      "step": 575
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.0004964903846153846,
+      "loss": 0.9693,
+      "step": 576
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.0004964423076923077,
+      "loss": 0.9556,
+      "step": 577
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.0004963942307692308,
+      "loss": 1.0065,
+      "step": 578
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0004963461538461539,
+      "loss": 1.0487,
+      "step": 579
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.0004962980769230769,
+      "loss": 0.9475,
+      "step": 580
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.0004962500000000001,
+      "loss": 0.8404,
+      "step": 581
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.0004962019230769231,
+      "loss": 0.8447,
+      "step": 582
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0004961538461538461,
+      "loss": 1.0357,
+      "step": 583
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.0004961057692307692,
+      "loss": 0.8545,
+      "step": 584
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0004960576923076923,
+      "loss": 0.9108,
+      "step": 585
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.0004960096153846154,
+      "loss": 1.0058,
+      "step": 586
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0004959615384615384,
+      "loss": 1.0374,
+      "step": 587
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0004959134615384616,
+      "loss": 1.0841,
+      "step": 588
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.0004958653846153846,
+      "loss": 0.9617,
+      "step": 589
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.0004958173076923077,
+      "loss": 0.9671,
+      "step": 590
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.0004957692307692308,
+      "loss": 1.0707,
+      "step": 591
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.0004957211538461539,
+      "loss": 1.1856,
+      "step": 592
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.0004956730769230769,
+      "loss": 1.2184,
+      "step": 593
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.000495625,
+      "loss": 1.0948,
+      "step": 594
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.0004955769230769231,
+      "loss": 0.976,
+      "step": 595
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0004955288461538462,
+      "loss": 1.0051,
+      "step": 596
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.0004954807692307692,
+      "loss": 1.0347,
+      "step": 597
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.0004954326923076924,
+      "loss": 1.0219,
+      "step": 598
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0004953846153846154,
+      "loss": 1.0722,
+      "step": 599
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0004953365384615385,
+      "loss": 1.0717,
+      "step": 600
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.0004952884615384615,
+      "loss": 1.1395,
+      "step": 601
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0004952403846153847,
+      "loss": 1.1284,
+      "step": 602
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.0004951923076923077,
+      "loss": 1.1675,
+      "step": 603
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.0004951442307692308,
+      "loss": 1.102,
+      "step": 604
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0004950961538461539,
+      "loss": 1.0845,
+      "step": 605
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.000495048076923077,
+      "loss": 0.9906,
+      "step": 606
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.000495,
+      "loss": 1.0383,
+      "step": 607
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.0004949519230769232,
+      "loss": 1.0385,
+      "step": 608
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.0004949038461538462,
+      "loss": 1.014,
+      "step": 609
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0004948557692307692,
+      "loss": 1.0348,
+      "step": 610
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0004948076923076923,
+      "loss": 0.9648,
+      "step": 611
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0004947596153846154,
+      "loss": 0.9356,
+      "step": 612
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.0004947115384615385,
+      "loss": 0.9936,
+      "step": 613
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0004946634615384615,
+      "loss": 0.8719,
+      "step": 614
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.0004946153846153847,
+      "loss": 0.8161,
+      "step": 615
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.0004945673076923076,
+      "loss": 1.0479,
+      "step": 616
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.0004945192307692308,
+      "loss": 1.0879,
+      "step": 617
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0004944711538461538,
+      "loss": 0.8995,
+      "step": 618
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0004944230769230769,
+      "loss": 1.0012,
+      "step": 619
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.000494375,
+      "loss": 1.0944,
+      "step": 620
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0004943269230769231,
+      "loss": 1.1649,
+      "step": 621
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0004942788461538461,
+      "loss": 1.1159,
+      "step": 622
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.0004942307692307693,
+      "loss": 1.0422,
+      "step": 623
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.0004941826923076923,
+      "loss": 1.0231,
+      "step": 624
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0004941346153846154,
+      "loss": 1.1603,
+      "step": 625
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.0004940865384615384,
+      "loss": 1.065,
+      "step": 626
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0004940384615384616,
+      "loss": 1.1499,
+      "step": 627
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.0004939903846153846,
+      "loss": 1.1724,
+      "step": 628
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.0004939423076923076,
+      "loss": 1.2602,
+      "step": 629
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.0004938942307692308,
+      "loss": 1.0535,
+      "step": 630
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.0004938461538461538,
+      "loss": 1.0491,
+      "step": 631
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0004937980769230769,
+      "loss": 1.076,
+      "step": 632
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00049375,
+      "loss": 0.9386,
+      "step": 633
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.0004937019230769231,
+      "loss": 0.9789,
+      "step": 634
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0004936538461538461,
+      "loss": 0.819,
+      "step": 635
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0004936057692307692,
+      "loss": 1.0355,
+      "step": 636
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.0004935576923076923,
+      "loss": 0.9974,
+      "step": 637
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.0004935096153846154,
+      "loss": 0.9693,
+      "step": 638
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0004934615384615384,
+      "loss": 1.0203,
+      "step": 639
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0004934134615384616,
+      "loss": 0.9136,
+      "step": 640
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0004933653846153846,
+      "loss": 1.1883,
+      "step": 641
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.0004933173076923077,
+      "loss": 1.15,
+      "step": 642
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.0004932692307692308,
+      "loss": 1.0432,
+      "step": 643
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0004932211538461539,
+      "loss": 1.0848,
+      "step": 644
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.0004931730769230769,
+      "loss": 0.9637,
+      "step": 645
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.000493125,
+      "loss": 1.0098,
+      "step": 646
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0004930769230769231,
+      "loss": 1.0922,
+      "step": 647
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0004930288461538462,
+      "loss": 0.9634,
+      "step": 648
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.0004929807692307692,
+      "loss": 1.0247,
+      "step": 649
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0004929326923076924,
+      "loss": 1.1232,
+      "step": 650
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.0004928846153846154,
+      "loss": 1.0288,
+      "step": 651
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.0004928365384615385,
+      "loss": 1.0783,
+      "step": 652
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0004927884615384616,
+      "loss": 1.2309,
+      "step": 653
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.0004927403846153847,
+      "loss": 1.32,
+      "step": 654
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0004926923076923077,
+      "loss": 1.0957,
+      "step": 655
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.0004926442307692307,
+      "loss": 0.9974,
+      "step": 656
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.0004925961538461539,
+      "loss": 1.0291,
+      "step": 657
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.0004925480769230769,
+      "loss": 1.0131,
+      "step": 658
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.0004925,
+      "loss": 0.9562,
+      "step": 659
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0004924519230769231,
+      "loss": 0.9094,
+      "step": 660
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0004924038461538462,
+      "loss": 1.096,
+      "step": 661
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.0004923557692307692,
+      "loss": 0.8334,
+      "step": 662
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.0004923076923076924,
+      "loss": 0.8999,
+      "step": 663
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0004922596153846154,
+      "loss": 1.0075,
+      "step": 664
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.0004922115384615385,
+      "loss": 0.9605,
+      "step": 665
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.0004921634615384615,
+      "loss": 0.972,
+      "step": 666
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.0004921153846153847,
+      "loss": 0.8798,
+      "step": 667
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.0004920673076923077,
+      "loss": 1.0736,
+      "step": 668
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0004920192307692308,
+      "loss": 0.9092,
+      "step": 669
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0004919711538461539,
+      "loss": 0.9206,
+      "step": 670
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.000491923076923077,
+      "loss": 1.0051,
+      "step": 671
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.000491875,
+      "loss": 0.7862,
+      "step": 672
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0004918269230769231,
+      "loss": 0.9342,
+      "step": 673
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.0004917788461538461,
+      "loss": 1.1219,
+      "step": 674
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.0004917307692307693,
+      "loss": 0.9447,
+      "step": 675
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.0004916826923076923,
+      "loss": 0.9354,
+      "step": 676
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0004916346153846154,
+      "loss": 1.0311,
+      "step": 677
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.0004915865384615384,
+      "loss": 1.0274,
+      "step": 678
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.0004915384615384616,
+      "loss": 1.0125,
+      "step": 679
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0004914903846153846,
+      "loss": 0.8771,
+      "step": 680
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.0004914423076923076,
+      "loss": 1.0537,
+      "step": 681
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.0004913942307692308,
+      "loss": 1.0213,
+      "step": 682
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.0004913461538461538,
+      "loss": 0.9669,
+      "step": 683
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0004912980769230769,
+      "loss": 1.0087,
+      "step": 684
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00049125,
+      "loss": 1.073,
+      "step": 685
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.0004912019230769231,
+      "loss": 0.9625,
+      "step": 686
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.0004911538461538461,
+      "loss": 0.9339,
+      "step": 687
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.0004911057692307692,
+      "loss": 0.9586,
+      "step": 688
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0004910576923076923,
+      "loss": 1.026,
+      "step": 689
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.0004910096153846154,
+      "loss": 0.975,
+      "step": 690
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.0004909615384615384,
+      "loss": 0.8453,
+      "step": 691
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.0004909134615384616,
+      "loss": 1.0263,
+      "step": 692
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.0004908653846153846,
+      "loss": 1.0013,
+      "step": 693
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0004908173076923077,
+      "loss": 0.9096,
+      "step": 694
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.0004907692307692308,
+      "loss": 0.9655,
+      "step": 695
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.0004907211538461539,
+      "loss": 0.9431,
+      "step": 696
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.0004906730769230769,
+      "loss": 1.0259,
+      "step": 697
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.000490625,
+      "loss": 1.1089,
+      "step": 698
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.0004905769230769231,
+      "loss": 0.8886,
+      "step": 699
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.0004905288461538462,
+      "loss": 0.9808,
+      "step": 700
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0004904807692307692,
+      "loss": 1.0763,
+      "step": 701
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.0004904326923076924,
+      "loss": 0.9769,
+      "step": 702
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.0004903846153846154,
+      "loss": 0.997,
+      "step": 703
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.0004903365384615384,
+      "loss": 1.0671,
+      "step": 704
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.0004902884615384616,
+      "loss": 1.0796,
+      "step": 705
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0004902403846153846,
+      "loss": 1.0294,
+      "step": 706
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.0004901923076923077,
+      "loss": 1.2051,
+      "step": 707
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0004901442307692307,
+      "loss": 1.1548,
+      "step": 708
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0004900961538461539,
+      "loss": 1.1901,
+      "step": 709
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.0004900480769230769,
+      "loss": 1.1065,
+      "step": 710
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00049,
+      "loss": 1.0188,
+      "step": 711
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.0004899519230769231,
+      "loss": 1.0164,
+      "step": 712
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.0004899038461538462,
+      "loss": 0.9798,
+      "step": 713
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.0004898557692307692,
+      "loss": 0.9768,
+      "step": 714
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0004898076923076924,
+      "loss": 0.9208,
+      "step": 715
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.0004897596153846154,
+      "loss": 0.9215,
+      "step": 716
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0004897115384615385,
+      "loss": 0.8148,
+      "step": 717
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.0004896634615384615,
+      "loss": 0.8644,
+      "step": 718
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.0004896153846153847,
+      "loss": 1.0587,
+      "step": 719
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0004895673076923077,
+      "loss": 0.8342,
+      "step": 720
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0004895192307692308,
+      "loss": 0.9029,
+      "step": 721
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.0004894711538461539,
+      "loss": 0.8114,
+      "step": 722
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.000489423076923077,
+      "loss": 1.0046,
+      "step": 723
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.000489375,
+      "loss": 1.0572,
+      "step": 724
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.0004893269230769232,
+      "loss": 1.0401,
+      "step": 725
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.0004892788461538462,
+      "loss": 1.0345,
+      "step": 726
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.0004892307692307693,
+      "loss": 0.9104,
+      "step": 727
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.0004891826923076923,
+      "loss": 0.9261,
+      "step": 728
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0004891346153846155,
+      "loss": 0.8384,
+      "step": 729
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.0004890865384615385,
+      "loss": 0.9128,
+      "step": 730
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.0004890384615384615,
+      "loss": 0.9364,
+      "step": 731
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.0004889903846153846,
+      "loss": 0.8977,
+      "step": 732
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.0004889423076923076,
+      "loss": 1.0917,
+      "step": 733
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.0004888942307692308,
+      "loss": 1.0459,
+      "step": 734
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.0004888461538461538,
+      "loss": 1.1638,
+      "step": 735
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.0004887980769230769,
+      "loss": 1.0886,
+      "step": 736
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00048875,
+      "loss": 0.9526,
+      "step": 737
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.0004887019230769231,
+      "loss": 1.012,
+      "step": 738
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.0004886538461538461,
+      "loss": 0.9046,
+      "step": 739
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.0004886057692307692,
+      "loss": 1.0787,
+      "step": 740
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0004885576923076923,
+      "loss": 1.0456,
+      "step": 741
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0004885096153846154,
+      "loss": 0.9465,
+      "step": 742
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.0004884615384615384,
+      "loss": 0.8611,
+      "step": 743
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.0004884134615384616,
+      "loss": 0.9319,
+      "step": 744
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.0004883653846153846,
+      "loss": 0.9865,
+      "step": 745
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0004883173076923077,
+      "loss": 0.9945,
+      "step": 746
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.0004882692307692308,
+      "loss": 1.0197,
+      "step": 747
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00048822115384615386,
+      "loss": 1.0225,
+      "step": 748
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00048817307692307696,
+      "loss": 1.0333,
+      "step": 749
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.000488125,
+      "loss": 1.087,
+      "step": 750
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.0004880769230769231,
+      "loss": 1.0377,
+      "step": 751
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0004880288461538462,
+      "loss": 0.9278,
+      "step": 752
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00048798076923076925,
+      "loss": 0.9399,
+      "step": 753
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.00048793269230769235,
+      "loss": 0.9802,
+      "step": 754
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.0004878846153846154,
+      "loss": 0.8783,
+      "step": 755
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0004878365384615385,
+      "loss": 0.8517,
+      "step": 756
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0004877884615384616,
+      "loss": 0.9205,
+      "step": 757
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.0004877403846153846,
+      "loss": 1.0155,
+      "step": 758
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.0004876923076923077,
+      "loss": 0.8688,
+      "step": 759
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00048764423076923073,
+      "loss": 1.0306,
+      "step": 760
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00048759615384615383,
+      "loss": 1.0195,
+      "step": 761
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.00048754807692307693,
+      "loss": 1.0036,
+      "step": 762
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.0004875,
+      "loss": 1.1348,
+      "step": 763
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0004874519230769231,
+      "loss": 1.0195,
+      "step": 764
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.0004874038461538461,
+      "loss": 0.9939,
+      "step": 765
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.0004873557692307692,
+      "loss": 1.0158,
+      "step": 766
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.0004873076923076923,
+      "loss": 0.9621,
+      "step": 767
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00048725961538461537,
+      "loss": 0.8613,
+      "step": 768
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00048721153846153847,
+      "loss": 0.9599,
+      "step": 769
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.0004871634615384615,
+      "loss": 0.8649,
+      "step": 770
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0004871153846153846,
+      "loss": 0.9128,
+      "step": 771
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.0004870673076923077,
+      "loss": 0.7396,
+      "step": 772
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00048701923076923076,
+      "loss": 0.8981,
+      "step": 773
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00048697115384615386,
+      "loss": 0.7656,
+      "step": 774
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.0004869230769230769,
+      "loss": 0.7948,
+      "step": 775
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.000486875,
+      "loss": 1.0745,
+      "step": 776
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0004868269230769231,
+      "loss": 0.8412,
+      "step": 777
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00048677884615384615,
+      "loss": 0.9844,
+      "step": 778
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00048673076923076925,
+      "loss": 0.9946,
+      "step": 779
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.0004866826923076923,
+      "loss": 0.9717,
+      "step": 780
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0004866346153846154,
+      "loss": 0.9227,
+      "step": 781
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0004865865384615385,
+      "loss": 0.8624,
+      "step": 782
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00048653846153846154,
+      "loss": 0.9639,
+      "step": 783
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00048649038461538464,
+      "loss": 0.8648,
+      "step": 784
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0004864423076923077,
+      "loss": 0.7611,
+      "step": 785
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.0004863942307692308,
+      "loss": 1.0923,
+      "step": 786
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0004863461538461539,
+      "loss": 0.8659,
+      "step": 787
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00048629807692307693,
+      "loss": 0.9415,
+      "step": 788
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00048625000000000003,
+      "loss": 0.9183,
+      "step": 789
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.0004862019230769231,
+      "loss": 0.8366,
+      "step": 790
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.0004861538461538462,
+      "loss": 1.0205,
+      "step": 791
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0004861057692307693,
+      "loss": 0.9243,
+      "step": 792
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0004860576923076923,
+      "loss": 1.0096,
+      "step": 793
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0004860096153846154,
+      "loss": 1.0063,
+      "step": 794
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00048596153846153846,
+      "loss": 0.954,
+      "step": 795
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00048591346153846156,
+      "loss": 0.9857,
+      "step": 796
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.0004858653846153846,
+      "loss": 0.9705,
+      "step": 797
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.0004858173076923077,
+      "loss": 0.9144,
+      "step": 798
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0004857692307692308,
+      "loss": 0.9595,
+      "step": 799
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00048572115384615385,
+      "loss": 0.9185,
+      "step": 800
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00048567307692307695,
+      "loss": 0.8536,
+      "step": 801
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.000485625,
+      "loss": 0.9793,
+      "step": 802
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0004855769230769231,
+      "loss": 0.8447,
+      "step": 803
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.0004855288461538462,
+      "loss": 0.8973,
+      "step": 804
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00048548076923076924,
+      "loss": 0.9132,
+      "step": 805
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00048543269230769234,
+      "loss": 0.9884,
+      "step": 806
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.0004853846153846154,
+      "loss": 0.8575,
+      "step": 807
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.0004853365384615385,
+      "loss": 0.9756,
+      "step": 808
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.0004852884615384616,
+      "loss": 1.0307,
+      "step": 809
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.00048524038461538464,
+      "loss": 0.881,
+      "step": 810
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00048519230769230773,
+      "loss": 0.9624,
+      "step": 811
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.0004851442307692308,
+      "loss": 0.9238,
+      "step": 812
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.0004850961538461539,
+      "loss": 0.8699,
+      "step": 813
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.000485048076923077,
+      "loss": 0.9119,
+      "step": 814
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00048499999999999997,
+      "loss": 0.8429,
+      "step": 815
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00048495192307692307,
+      "loss": 0.8439,
+      "step": 816
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0004849038461538461,
+      "loss": 0.987,
+      "step": 817
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0004848557692307692,
+      "loss": 1.0352,
+      "step": 818
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.0004848076923076923,
+      "loss": 0.9882,
+      "step": 819
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00048475961538461536,
+      "loss": 0.9977,
+      "step": 820
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00048471153846153846,
+      "loss": 1.2048,
+      "step": 821
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.0004846634615384615,
+      "loss": 0.9551,
+      "step": 822
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.0004846153846153846,
+      "loss": 0.8787,
+      "step": 823
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0004845673076923077,
+      "loss": 0.9426,
+      "step": 824
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00048451923076923075,
+      "loss": 0.9728,
+      "step": 825
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00048447115384615385,
+      "loss": 0.8235,
+      "step": 826
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.0004844230769230769,
+      "loss": 0.7804,
+      "step": 827
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.000484375,
+      "loss": 0.9581,
+      "step": 828
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.0004843269230769231,
+      "loss": 0.9425,
+      "step": 829
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00048427884615384614,
+      "loss": 0.8229,
+      "step": 830
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00048423076923076924,
+      "loss": 0.8518,
+      "step": 831
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0004841826923076923,
+      "loss": 0.8568,
+      "step": 832
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.0004841346153846154,
+      "loss": 1.045,
+      "step": 833
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.0004840865384615385,
+      "loss": 1.0613,
+      "step": 834
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00048403846153846153,
+      "loss": 0.8275,
+      "step": 835
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00048399038461538463,
+      "loss": 0.9121,
+      "step": 836
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.0004839423076923077,
+      "loss": 0.9866,
+      "step": 837
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0004838942307692308,
+      "loss": 0.9373,
+      "step": 838
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.0004838461538461539,
+      "loss": 0.9807,
+      "step": 839
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0004837980769230769,
+      "loss": 0.7816,
+      "step": 840
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00048375,
+      "loss": 0.9977,
+      "step": 841
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00048370192307692307,
+      "loss": 1.0011,
+      "step": 842
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00048365384615384617,
+      "loss": 1.0015,
+      "step": 843
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00048360576923076927,
+      "loss": 0.8467,
+      "step": 844
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.0004835576923076923,
+      "loss": 1.0896,
+      "step": 845
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0004835096153846154,
+      "loss": 1.0903,
+      "step": 846
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00048346153846153846,
+      "loss": 0.9567,
+      "step": 847
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00048341346153846156,
+      "loss": 1.0276,
+      "step": 848
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00048336538461538466,
+      "loss": 0.9094,
+      "step": 849
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.0004833173076923077,
+      "loss": 0.8598,
+      "step": 850
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.0004832692307692308,
+      "loss": 0.9701,
+      "step": 851
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00048322115384615385,
+      "loss": 0.9476,
+      "step": 852
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00048317307692307695,
+      "loss": 0.8701,
+      "step": 853
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00048312500000000005,
+      "loss": 0.8528,
+      "step": 854
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.0004830769230769231,
+      "loss": 0.8698,
+      "step": 855
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.0004830288461538462,
+      "loss": 0.9191,
+      "step": 856
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00048298076923076924,
+      "loss": 0.995,
+      "step": 857
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00048293269230769234,
+      "loss": 1.1388,
+      "step": 858
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00048288461538461544,
+      "loss": 1.0896,
+      "step": 859
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.0004828365384615385,
+      "loss": 1.0811,
+      "step": 860
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.0004827884615384616,
+      "loss": 0.8623,
+      "step": 861
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00048274038461538463,
+      "loss": 0.8253,
+      "step": 862
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00048269230769230773,
+      "loss": 0.9993,
+      "step": 863
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.0004826442307692308,
+      "loss": 0.9248,
+      "step": 864
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.0004825961538461539,
+      "loss": 0.9581,
+      "step": 865
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.000482548076923077,
+      "loss": 1.053,
+      "step": 866
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.0004825,
+      "loss": 0.9346,
+      "step": 867
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.0004824519230769231,
+      "loss": 1.0853,
+      "step": 868
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00048240384615384616,
+      "loss": 1.0177,
+      "step": 869
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00048235576923076926,
+      "loss": 1.057,
+      "step": 870
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0004823076923076923,
+      "loss": 0.8037,
+      "step": 871
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00048225961538461535,
+      "loss": 0.9083,
+      "step": 872
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00048221153846153845,
+      "loss": 0.9702,
+      "step": 873
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.0004821634615384615,
+      "loss": 0.9772,
+      "step": 874
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.0004821153846153846,
+      "loss": 0.9309,
+      "step": 875
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0004820673076923077,
+      "loss": 1.0356,
+      "step": 876
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00048201923076923074,
+      "loss": 0.9338,
+      "step": 877
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00048197115384615384,
+      "loss": 0.8516,
+      "step": 878
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.0004819230769230769,
+      "loss": 0.8382,
+      "step": 879
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.000481875,
+      "loss": 0.8153,
+      "step": 880
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.0004818269230769231,
+      "loss": 0.8375,
+      "step": 881
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00048177884615384614,
+      "loss": 0.9974,
+      "step": 882
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00048173076923076923,
+      "loss": 1.003,
+      "step": 883
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.0004816826923076923,
+      "loss": 0.8233,
+      "step": 884
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.0004816346153846154,
+      "loss": 0.8671,
+      "step": 885
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0004815865384615385,
+      "loss": 0.8795,
+      "step": 886
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0004815384615384615,
+      "loss": 0.9033,
+      "step": 887
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0004814903846153846,
+      "loss": 0.9132,
+      "step": 888
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00048144230769230767,
+      "loss": 0.9836,
+      "step": 889
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00048139423076923077,
+      "loss": 0.7845,
+      "step": 890
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00048134615384615387,
+      "loss": 0.8423,
+      "step": 891
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0004812980769230769,
+      "loss": 0.9392,
+      "step": 892
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00048125,
+      "loss": 1.0053,
+      "step": 893
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00048120192307692306,
+      "loss": 0.8903,
+      "step": 894
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00048115384615384616,
+      "loss": 0.8287,
+      "step": 895
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00048110576923076926,
+      "loss": 0.9673,
+      "step": 896
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.0004810576923076923,
+      "loss": 0.8354,
+      "step": 897
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.0004810096153846154,
+      "loss": 0.9301,
+      "step": 898
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00048096153846153845,
+      "loss": 0.8577,
+      "step": 899
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.00048091346153846155,
+      "loss": 1.0891,
+      "step": 900
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00048086538461538465,
+      "loss": 1.1201,
+      "step": 901
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0004808173076923077,
+      "loss": 0.9679,
+      "step": 902
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0004807692307692308,
+      "loss": 0.8441,
+      "step": 903
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00048072115384615384,
+      "loss": 0.8966,
+      "step": 904
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00048067307692307694,
+      "loss": 0.9185,
+      "step": 905
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00048062500000000004,
+      "loss": 0.9494,
+      "step": 906
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.0004805769230769231,
+      "loss": 0.854,
+      "step": 907
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.0004805288461538462,
+      "loss": 0.8011,
+      "step": 908
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00048048076923076923,
+      "loss": 0.8425,
+      "step": 909
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00048043269230769233,
+      "loss": 0.7994,
+      "step": 910
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00048038461538461543,
+      "loss": 0.7088,
+      "step": 911
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.0004803365384615385,
+      "loss": 0.9599,
+      "step": 912
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.0004802884615384616,
+      "loss": 0.8883,
+      "step": 913
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0004802403846153846,
+      "loss": 0.9907,
+      "step": 914
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0004801923076923077,
+      "loss": 0.9068,
+      "step": 915
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.0004801442307692308,
+      "loss": 0.7979,
+      "step": 916
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00048009615384615387,
+      "loss": 0.8661,
+      "step": 917
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00048004807692307697,
+      "loss": 0.8285,
+      "step": 918
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00048,
+      "loss": 0.8752,
+      "step": 919
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.0004799519230769231,
+      "loss": 0.8411,
+      "step": 920
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.0004799038461538462,
+      "loss": 0.9541,
+      "step": 921
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00047985576923076926,
+      "loss": 0.9221,
+      "step": 922
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00047980769230769236,
+      "loss": 0.8273,
+      "step": 923
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0004797596153846154,
+      "loss": 0.9095,
+      "step": 924
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.0004797115384615385,
+      "loss": 0.9286,
+      "step": 925
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00047966346153846155,
+      "loss": 0.8543,
+      "step": 926
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.0004796153846153846,
+      "loss": 1.0012,
+      "step": 927
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0004795673076923077,
+      "loss": 0.958,
+      "step": 928
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00047951923076923074,
+      "loss": 0.9774,
+      "step": 929
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00047947115384615384,
+      "loss": 0.8909,
+      "step": 930
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.0004794230769230769,
+      "loss": 0.9176,
+      "step": 931
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.000479375,
+      "loss": 0.88,
+      "step": 932
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0004793269230769231,
+      "loss": 0.9134,
+      "step": 933
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00047927884615384613,
+      "loss": 0.8352,
+      "step": 934
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00047923076923076923,
+      "loss": 0.8781,
+      "step": 935
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.0004791826923076923,
+      "loss": 0.8076,
+      "step": 936
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0004791346153846154,
+      "loss": 1.0811,
+      "step": 937
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0004790865384615385,
+      "loss": 0.9409,
+      "step": 938
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0004790384615384615,
+      "loss": 0.9051,
+      "step": 939
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.0004789903846153846,
+      "loss": 0.9354,
+      "step": 940
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00047894230769230766,
+      "loss": 0.8704,
+      "step": 941
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00047889423076923076,
+      "loss": 0.8962,
+      "step": 942
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00047884615384615386,
+      "loss": 0.8277,
+      "step": 943
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.0004787980769230769,
+      "loss": 0.911,
+      "step": 944
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00047875,
+      "loss": 0.8807,
+      "step": 945
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00047870192307692305,
+      "loss": 0.9107,
+      "step": 946
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.00047865384615384615,
+      "loss": 0.8391,
+      "step": 947
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00047860576923076925,
+      "loss": 0.8572,
+      "step": 948
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0004785576923076923,
+      "loss": 0.8652,
+      "step": 949
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0004785096153846154,
+      "loss": 0.8508,
+      "step": 950
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00047846153846153844,
+      "loss": 1.1651,
+      "step": 951
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00047841346153846154,
+      "loss": 1.0486,
+      "step": 952
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00047836538461538464,
+      "loss": 0.8043,
+      "step": 953
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0004783173076923077,
+      "loss": 0.9333,
+      "step": 954
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.0004782692307692308,
+      "loss": 0.8855,
+      "step": 955
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00047822115384615383,
+      "loss": 1.025,
+      "step": 956
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00047817307692307693,
+      "loss": 0.9252,
+      "step": 957
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00047812500000000003,
+      "loss": 0.9767,
+      "step": 958
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.0004780769230769231,
+      "loss": 0.8772,
+      "step": 959
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.0004780288461538462,
+      "loss": 0.9704,
+      "step": 960
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.0004779807692307692,
+      "loss": 0.7928,
+      "step": 961
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.0004779326923076923,
+      "loss": 0.8171,
+      "step": 962
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.0004778846153846154,
+      "loss": 0.7725,
+      "step": 963
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.00047783653846153847,
+      "loss": 0.9875,
+      "step": 964
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00047778846153846157,
+      "loss": 0.8746,
+      "step": 965
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.0004777403846153846,
+      "loss": 0.8348,
+      "step": 966
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.0004776923076923077,
+      "loss": 0.9093,
+      "step": 967
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.0004776442307692308,
+      "loss": 0.9822,
+      "step": 968
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00047759615384615386,
+      "loss": 0.8349,
+      "step": 969
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00047754807692307696,
+      "loss": 0.9161,
+      "step": 970
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.0004775,
+      "loss": 0.8311,
+      "step": 971
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.0004774519230769231,
+      "loss": 0.9991,
+      "step": 972
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.0004774038461538462,
+      "loss": 0.8149,
+      "step": 973
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00047735576923076925,
+      "loss": 0.9685,
+      "step": 974
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00047730769230769235,
+      "loss": 0.9765,
+      "step": 975
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0004772596153846154,
+      "loss": 0.9372,
+      "step": 976
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.0004772115384615385,
+      "loss": 0.8342,
+      "step": 977
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.0004771634615384616,
+      "loss": 0.9331,
+      "step": 978
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00047711538461538464,
+      "loss": 0.8902,
+      "step": 979
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00047706730769230774,
+      "loss": 1.0366,
+      "step": 980
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0004770192307692308,
+      "loss": 1.1903,
+      "step": 981
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.0004769711538461539,
+      "loss": 0.9053,
+      "step": 982
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.000476923076923077,
+      "loss": 0.9205,
+      "step": 983
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.000476875,
+      "loss": 0.9243,
+      "step": 984
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.0004768269230769231,
+      "loss": 0.9135,
+      "step": 985
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0004767788461538461,
+      "loss": 0.9613,
+      "step": 986
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.0004767307692307692,
+      "loss": 0.8171,
+      "step": 987
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.0004766826923076923,
+      "loss": 0.9233,
+      "step": 988
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00047663461538461537,
+      "loss": 0.8647,
+      "step": 989
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00047658653846153847,
+      "loss": 0.8421,
+      "step": 990
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.0004765384615384615,
+      "loss": 0.7576,
+      "step": 991
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.0004764903846153846,
+      "loss": 0.8122,
+      "step": 992
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00047644230769230766,
+      "loss": 0.7342,
+      "step": 993
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00047639423076923076,
+      "loss": 0.808,
+      "step": 994
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00047634615384615386,
+      "loss": 0.8157,
+      "step": 995
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.0004762980769230769,
+      "loss": 0.9012,
+      "step": 996
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00047625,
+      "loss": 1.0437,
+      "step": 997
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00047620192307692305,
+      "loss": 0.8284,
+      "step": 998
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00047615384615384615,
+      "loss": 0.8533,
+      "step": 999
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00047610576923076925,
+      "loss": 0.7573,
+      "step": 1000
+    },
+    {
+      "epoch": 9.17,
+      "eval_cer": 0.2100481266585706,
+      "eval_loss": 0.7544377446174622,
+      "eval_runtime": 87.0343,
+      "eval_samples_per_second": 18.924,
+      "eval_steps_per_second": 2.367,
+      "eval_wer": 0.6960473904606271,
+      "step": 1000
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.0004760576923076923,
+      "loss": 0.8104,
+      "step": 1001
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.0004760096153846154,
+      "loss": 0.6488,
+      "step": 1002
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00047596153846153844,
+      "loss": 0.9411,
+      "step": 1003
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00047591346153846154,
+      "loss": 1.0448,
+      "step": 1004
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00047586538461538464,
+      "loss": 0.9383,
+      "step": 1005
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.0004758173076923077,
+      "loss": 0.9134,
+      "step": 1006
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.0004757692307692308,
+      "loss": 0.9484,
+      "step": 1007
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00047572115384615383,
+      "loss": 0.8928,
+      "step": 1008
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00047567307692307693,
+      "loss": 1.1089,
+      "step": 1009
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00047562500000000003,
+      "loss": 0.9966,
+      "step": 1010
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0004755769230769231,
+      "loss": 0.9878,
+      "step": 1011
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0004755288461538462,
+      "loss": 0.9115,
+      "step": 1012
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.0004754807692307692,
+      "loss": 0.9527,
+      "step": 1013
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.0004754326923076923,
+      "loss": 0.8597,
+      "step": 1014
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.0004753846153846154,
+      "loss": 0.8871,
+      "step": 1015
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00047533653846153846,
+      "loss": 0.9294,
+      "step": 1016
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00047528846153846156,
+      "loss": 0.7638,
+      "step": 1017
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.0004752403846153846,
+      "loss": 0.8325,
+      "step": 1018
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.0004751923076923077,
+      "loss": 0.8751,
+      "step": 1019
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.0004751442307692308,
+      "loss": 0.7603,
+      "step": 1020
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00047509615384615385,
+      "loss": 0.6975,
+      "step": 1021
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00047504807692307695,
+      "loss": 0.9693,
+      "step": 1022
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.000475,
+      "loss": 0.8895,
+      "step": 1023
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0004749519230769231,
+      "loss": 0.8408,
+      "step": 1024
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.0004749038461538462,
+      "loss": 0.9207,
+      "step": 1025
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00047485576923076924,
+      "loss": 0.8022,
+      "step": 1026
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00047480769230769234,
+      "loss": 0.8751,
+      "step": 1027
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.0004747596153846154,
+      "loss": 0.8819,
+      "step": 1028
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.0004747115384615385,
+      "loss": 1.0177,
+      "step": 1029
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.0004746634615384616,
+      "loss": 0.8201,
+      "step": 1030
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00047461538461538463,
+      "loss": 0.8264,
+      "step": 1031
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00047456730769230773,
+      "loss": 1.0105,
+      "step": 1032
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.0004745192307692308,
+      "loss": 0.8967,
+      "step": 1033
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.0004744711538461539,
+      "loss": 0.88,
+      "step": 1034
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.000474423076923077,
+      "loss": 0.8754,
+      "step": 1035
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.000474375,
+      "loss": 0.9908,
+      "step": 1036
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.0004743269230769231,
+      "loss": 0.9444,
+      "step": 1037
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00047427884615384617,
+      "loss": 0.9656,
+      "step": 1038
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00047423076923076927,
+      "loss": 0.9321,
+      "step": 1039
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00047418269230769237,
+      "loss": 0.8737,
+      "step": 1040
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00047413461538461536,
+      "loss": 0.8675,
+      "step": 1041
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.00047408653846153846,
+      "loss": 0.8926,
+      "step": 1042
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.0004740384615384615,
+      "loss": 0.7959,
+      "step": 1043
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.0004739903846153846,
+      "loss": 0.7162,
+      "step": 1044
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.0004739423076923077,
+      "loss": 0.7964,
+      "step": 1045
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00047389423076923075,
+      "loss": 0.937,
+      "step": 1046
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00047384615384615385,
+      "loss": 0.8898,
+      "step": 1047
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.0004737980769230769,
+      "loss": 0.9285,
+      "step": 1048
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00047375,
+      "loss": 0.8709,
+      "step": 1049
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.0004737019230769231,
+      "loss": 0.8179,
+      "step": 1050
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00047365384615384614,
+      "loss": 0.9095,
+      "step": 1051
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00047360576923076924,
+      "loss": 0.8115,
+      "step": 1052
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.0004735576923076923,
+      "loss": 0.7371,
+      "step": 1053
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0004735096153846154,
+      "loss": 0.9203,
+      "step": 1054
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.0004734615384615385,
+      "loss": 1.0186,
+      "step": 1055
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00047341346153846153,
+      "loss": 0.972,
+      "step": 1056
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00047336538461538463,
+      "loss": 0.8889,
+      "step": 1057
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.0004733173076923077,
+      "loss": 0.9029,
+      "step": 1058
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0004732692307692308,
+      "loss": 0.9536,
+      "step": 1059
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0004732211538461538,
+      "loss": 0.827,
+      "step": 1060
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.0004731730769230769,
+      "loss": 0.8248,
+      "step": 1061
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.000473125,
+      "loss": 0.9776,
+      "step": 1062
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.00047307692307692307,
+      "loss": 1.0623,
+      "step": 1063
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00047302884615384617,
+      "loss": 1.0627,
+      "step": 1064
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0004729807692307692,
+      "loss": 1.0404,
+      "step": 1065
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.0004729326923076923,
+      "loss": 1.0349,
+      "step": 1066
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.0004728846153846154,
+      "loss": 0.9301,
+      "step": 1067
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00047283653846153846,
+      "loss": 0.8917,
+      "step": 1068
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00047278846153846156,
+      "loss": 0.9055,
+      "step": 1069
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.0004727403846153846,
+      "loss": 0.9203,
+      "step": 1070
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.0004726923076923077,
+      "loss": 0.8748,
+      "step": 1071
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.0004726442307692308,
+      "loss": 0.9198,
+      "step": 1072
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00047259615384615385,
+      "loss": 0.9349,
+      "step": 1073
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00047254807692307695,
+      "loss": 0.771,
+      "step": 1074
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.0004725,
+      "loss": 0.7589,
+      "step": 1075
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.0004724519230769231,
+      "loss": 1.0442,
+      "step": 1076
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0004724038461538462,
+      "loss": 0.7104,
+      "step": 1077
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00047235576923076924,
+      "loss": 0.8701,
+      "step": 1078
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00047230769230769234,
+      "loss": 0.8932,
+      "step": 1079
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.0004722596153846154,
+      "loss": 0.8652,
+      "step": 1080
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.0004722115384615385,
+      "loss": 1.0119,
+      "step": 1081
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.0004721634615384616,
+      "loss": 0.8499,
+      "step": 1082
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00047211538461538463,
+      "loss": 0.8732,
+      "step": 1083
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00047206730769230773,
+      "loss": 0.9112,
+      "step": 1084
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.0004720192307692308,
+      "loss": 0.8871,
+      "step": 1085
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00047197115384615387,
+      "loss": 0.8358,
+      "step": 1086
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00047192307692307697,
+      "loss": 0.7823,
+      "step": 1087
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.000471875,
+      "loss": 0.9063,
+      "step": 1088
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.0004718269230769231,
+      "loss": 0.9375,
+      "step": 1089
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00047177884615384616,
+      "loss": 0.884,
+      "step": 1090
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00047173076923076926,
+      "loss": 0.9372,
+      "step": 1091
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00047168269230769236,
+      "loss": 0.8695,
+      "step": 1092
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0004716346153846154,
+      "loss": 0.9267,
+      "step": 1093
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.0004715865384615385,
+      "loss": 0.8559,
+      "step": 1094
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00047153846153846155,
+      "loss": 0.8227,
+      "step": 1095
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00047149038461538465,
+      "loss": 0.907,
+      "step": 1096
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00047144230769230775,
+      "loss": 0.8104,
+      "step": 1097
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00047139423076923074,
+      "loss": 0.996,
+      "step": 1098
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00047134615384615384,
+      "loss": 0.7887,
+      "step": 1099
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0004712980769230769,
+      "loss": 0.7269,
+      "step": 1100
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00047125,
+      "loss": 0.7539,
+      "step": 1101
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.0004712019230769231,
+      "loss": 0.8473,
+      "step": 1102
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00047115384615384613,
+      "loss": 0.752,
+      "step": 1103
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00047110576923076923,
+      "loss": 0.7759,
+      "step": 1104
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.0004710576923076923,
+      "loss": 0.9246,
+      "step": 1105
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.0004710096153846154,
+      "loss": 0.8788,
+      "step": 1106
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.0004709615384615385,
+      "loss": 0.7016,
+      "step": 1107
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0004709134615384615,
+      "loss": 0.823,
+      "step": 1108
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0004708653846153846,
+      "loss": 0.7441,
+      "step": 1109
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00047081730769230767,
+      "loss": 0.7105,
+      "step": 1110
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00047076923076923077,
+      "loss": 0.6273,
+      "step": 1111
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00047072115384615387,
+      "loss": 0.8515,
+      "step": 1112
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.0004706730769230769,
+      "loss": 0.7739,
+      "step": 1113
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.000470625,
+      "loss": 0.7393,
+      "step": 1114
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00047057692307692306,
+      "loss": 0.8421,
+      "step": 1115
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.00047052884615384616,
+      "loss": 0.8037,
+      "step": 1116
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00047048076923076926,
+      "loss": 0.935,
+      "step": 1117
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.0004704326923076923,
+      "loss": 0.852,
+      "step": 1118
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.0004703846153846154,
+      "loss": 0.7975,
+      "step": 1119
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00047033653846153845,
+      "loss": 0.9102,
+      "step": 1120
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00047028846153846155,
+      "loss": 0.7795,
+      "step": 1121
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.0004702403846153846,
+      "loss": 0.823,
+      "step": 1122
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.0004701923076923077,
+      "loss": 0.7838,
+      "step": 1123
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.0004701442307692308,
+      "loss": 0.735,
+      "step": 1124
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00047009615384615384,
+      "loss": 0.73,
+      "step": 1125
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00047004807692307694,
+      "loss": 0.7768,
+      "step": 1126
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00047,
+      "loss": 0.7992,
+      "step": 1127
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.0004699519230769231,
+      "loss": 0.8024,
+      "step": 1128
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.0004699038461538462,
+      "loss": 0.79,
+      "step": 1129
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00046985576923076923,
+      "loss": 0.8038,
+      "step": 1130
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00046980769230769233,
+      "loss": 0.9063,
+      "step": 1131
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0004697596153846154,
+      "loss": 0.9104,
+      "step": 1132
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0004697115384615385,
+      "loss": 0.8647,
+      "step": 1133
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.0004696634615384616,
+      "loss": 0.8355,
+      "step": 1134
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.0004696153846153846,
+      "loss": 0.7986,
+      "step": 1135
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.0004695673076923077,
+      "loss": 0.9955,
+      "step": 1136
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00046951923076923077,
+      "loss": 0.7605,
+      "step": 1137
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00046947115384615387,
+      "loss": 0.9661,
+      "step": 1138
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00046942307692307697,
+      "loss": 0.8543,
+      "step": 1139
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.000469375,
+      "loss": 0.7999,
+      "step": 1140
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0004693269230769231,
+      "loss": 0.6733,
+      "step": 1141
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00046927884615384616,
+      "loss": 0.8674,
+      "step": 1142
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00046923076923076926,
+      "loss": 0.8242,
+      "step": 1143
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00046918269230769236,
+      "loss": 0.7961,
+      "step": 1144
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0004691346153846154,
+      "loss": 0.9071,
+      "step": 1145
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.0004690865384615385,
+      "loss": 0.9289,
+      "step": 1146
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00046903846153846155,
+      "loss": 0.8408,
+      "step": 1147
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00046899038461538465,
+      "loss": 0.9104,
+      "step": 1148
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00046894230769230775,
+      "loss": 0.9713,
+      "step": 1149
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.0004688942307692308,
+      "loss": 0.8224,
+      "step": 1150
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.0004688461538461539,
+      "loss": 0.7382,
+      "step": 1151
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00046879807692307694,
+      "loss": 0.8103,
+      "step": 1152
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00046875,
+      "loss": 0.7914,
+      "step": 1153
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.0004687019230769231,
+      "loss": 0.795,
+      "step": 1154
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00046865384615384613,
+      "loss": 0.7414,
+      "step": 1155
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00046860576923076923,
+      "loss": 0.899,
+      "step": 1156
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0004685576923076923,
+      "loss": 0.8187,
+      "step": 1157
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00046850961538461537,
+      "loss": 0.8284,
+      "step": 1158
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00046846153846153847,
+      "loss": 0.828,
+      "step": 1159
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.0004684134615384615,
+      "loss": 0.9636,
+      "step": 1160
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.0004683653846153846,
+      "loss": 0.7871,
+      "step": 1161
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00046831730769230766,
+      "loss": 0.9095,
+      "step": 1162
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00046826923076923076,
+      "loss": 1.0216,
+      "step": 1163
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00046822115384615386,
+      "loss": 0.9019,
+      "step": 1164
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.0004681730769230769,
+      "loss": 0.9203,
+      "step": 1165
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.000468125,
+      "loss": 0.8048,
+      "step": 1166
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00046807692307692305,
+      "loss": 0.8478,
+      "step": 1167
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00046802884615384615,
+      "loss": 0.9143,
+      "step": 1168
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00046798076923076925,
+      "loss": 0.8018,
+      "step": 1169
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.0004679326923076923,
+      "loss": 0.8721,
+      "step": 1170
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.0004678846153846154,
+      "loss": 0.9164,
+      "step": 1171
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00046783653846153844,
+      "loss": 1.0259,
+      "step": 1172
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00046778846153846154,
+      "loss": 0.8604,
+      "step": 1173
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00046774038461538464,
+      "loss": 1.0257,
+      "step": 1174
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.0004676923076923077,
+      "loss": 0.8709,
+      "step": 1175
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.0004676442307692308,
+      "loss": 0.814,
+      "step": 1176
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00046759615384615383,
+      "loss": 0.815,
+      "step": 1177
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00046754807692307693,
+      "loss": 0.9821,
+      "step": 1178
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00046750000000000003,
+      "loss": 0.7521,
+      "step": 1179
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0004674519230769231,
+      "loss": 0.7244,
+      "step": 1180
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0004674038461538462,
+      "loss": 0.8937,
+      "step": 1181
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.0004673557692307692,
+      "loss": 0.8759,
+      "step": 1182
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.0004673076923076923,
+      "loss": 0.8108,
+      "step": 1183
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.0004672596153846154,
+      "loss": 0.8074,
+      "step": 1184
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00046721153846153847,
+      "loss": 0.8381,
+      "step": 1185
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00046716346153846157,
+      "loss": 0.8372,
+      "step": 1186
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.0004671153846153846,
+      "loss": 0.8124,
+      "step": 1187
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0004670673076923077,
+      "loss": 0.7451,
+      "step": 1188
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00046701923076923076,
+      "loss": 0.9024,
+      "step": 1189
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00046697115384615386,
+      "loss": 0.7161,
+      "step": 1190
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00046692307692307696,
+      "loss": 0.8123,
+      "step": 1191
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.000466875,
+      "loss": 0.7904,
+      "step": 1192
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.0004668269230769231,
+      "loss": 0.7303,
+      "step": 1193
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00046677884615384615,
+      "loss": 0.792,
+      "step": 1194
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00046673076923076925,
+      "loss": 0.8881,
+      "step": 1195
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00046668269230769235,
+      "loss": 0.8234,
+      "step": 1196
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.0004666346153846154,
+      "loss": 0.9418,
+      "step": 1197
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.0004665865384615385,
+      "loss": 0.7754,
+      "step": 1198
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00046653846153846154,
+      "loss": 0.7204,
+      "step": 1199
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00046649038461538464,
+      "loss": 1.0702,
+      "step": 1200
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.00046644230769230774,
+      "loss": 0.8774,
+      "step": 1201
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.0004663942307692308,
+      "loss": 0.9527,
+      "step": 1202
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.0004663461538461539,
+      "loss": 0.7727,
+      "step": 1203
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00046629807692307693,
+      "loss": 0.7537,
+      "step": 1204
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00046625000000000003,
+      "loss": 0.7244,
+      "step": 1205
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00046620192307692313,
+      "loss": 0.7598,
+      "step": 1206
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.0004661538461538462,
+      "loss": 0.8116,
+      "step": 1207
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.0004661057692307693,
+      "loss": 0.8306,
+      "step": 1208
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00046605769230769227,
+      "loss": 0.8934,
+      "step": 1209
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00046600961538461537,
+      "loss": 0.6228,
+      "step": 1210
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00046596153846153847,
+      "loss": 0.7701,
+      "step": 1211
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.0004659134615384615,
+      "loss": 0.7976,
+      "step": 1212
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.0004658653846153846,
+      "loss": 0.8551,
+      "step": 1213
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00046581730769230766,
+      "loss": 0.7247,
+      "step": 1214
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00046576923076923076,
+      "loss": 0.8136,
+      "step": 1215
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00046572115384615386,
+      "loss": 0.7579,
+      "step": 1216
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.0004656730769230769,
+      "loss": 0.6725,
+      "step": 1217
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.000465625,
+      "loss": 0.7873,
+      "step": 1218
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00046557692307692305,
+      "loss": 0.8298,
+      "step": 1219
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00046552884615384615,
+      "loss": 0.7909,
+      "step": 1220
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00046548076923076925,
+      "loss": 0.6646,
+      "step": 1221
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.0004654326923076923,
+      "loss": 0.8131,
+      "step": 1222
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0004653846153846154,
+      "loss": 0.7828,
+      "step": 1223
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00046533653846153844,
+      "loss": 0.7382,
+      "step": 1224
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00046528846153846154,
+      "loss": 0.7096,
+      "step": 1225
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00046524038461538464,
+      "loss": 0.7514,
+      "step": 1226
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.0004651923076923077,
+      "loss": 0.931,
+      "step": 1227
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.0004651442307692308,
+      "loss": 0.7987,
+      "step": 1228
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00046509615384615383,
+      "loss": 0.8728,
+      "step": 1229
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00046504807692307693,
+      "loss": 0.9002,
+      "step": 1230
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.000465,
+      "loss": 0.7227,
+      "step": 1231
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00046495192307692307,
+      "loss": 0.8519,
+      "step": 1232
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00046490384615384617,
+      "loss": 0.7668,
+      "step": 1233
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.0004648557692307692,
+      "loss": 0.8001,
+      "step": 1234
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.0004648076923076923,
+      "loss": 0.7375,
+      "step": 1235
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.0004647596153846154,
+      "loss": 0.7579,
+      "step": 1236
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00046471153846153846,
+      "loss": 0.7104,
+      "step": 1237
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00046466346153846156,
+      "loss": 0.7717,
+      "step": 1238
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.0004646153846153846,
+      "loss": 0.7214,
+      "step": 1239
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.0004645673076923077,
+      "loss": 0.8879,
+      "step": 1240
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.0004645192307692308,
+      "loss": 0.866,
+      "step": 1241
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00046447115384615385,
+      "loss": 0.8645,
+      "step": 1242
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00046442307692307695,
+      "loss": 0.8781,
+      "step": 1243
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.000464375,
+      "loss": 0.8945,
+      "step": 1244
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.0004643269230769231,
+      "loss": 0.8547,
+      "step": 1245
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.0004642788461538462,
+      "loss": 0.8329,
+      "step": 1246
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00046423076923076924,
+      "loss": 0.8731,
+      "step": 1247
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00046418269230769234,
+      "loss": 0.8965,
+      "step": 1248
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.0004641346153846154,
+      "loss": 0.6283,
+      "step": 1249
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.0004640865384615385,
+      "loss": 0.8248,
+      "step": 1250
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00046403846153846153,
+      "loss": 0.8259,
+      "step": 1251
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00046399038461538463,
+      "loss": 0.8198,
+      "step": 1252
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00046394230769230773,
+      "loss": 0.7818,
+      "step": 1253
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.0004638942307692308,
+      "loss": 0.8757,
+      "step": 1254
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.0004638461538461539,
+      "loss": 0.9135,
+      "step": 1255
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0004637980769230769,
+      "loss": 0.8469,
+      "step": 1256
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00046375,
+      "loss": 0.9057,
+      "step": 1257
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.0004637019230769231,
+      "loss": 0.7837,
+      "step": 1258
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00046365384615384617,
+      "loss": 0.8917,
+      "step": 1259
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00046360576923076927,
+      "loss": 0.7828,
+      "step": 1260
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.0004635576923076923,
+      "loss": 0.7811,
+      "step": 1261
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.0004635096153846154,
+      "loss": 0.7412,
+      "step": 1262
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.0004634615384615385,
+      "loss": 0.812,
+      "step": 1263
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.00046341346153846156,
+      "loss": 0.7896,
+      "step": 1264
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00046336538461538466,
+      "loss": 0.7125,
+      "step": 1265
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00046331730769230765,
+      "loss": 0.7851,
+      "step": 1266
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00046326923076923075,
+      "loss": 0.7804,
+      "step": 1267
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00046322115384615385,
+      "loss": 0.795,
+      "step": 1268
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.0004631730769230769,
+      "loss": 0.8353,
+      "step": 1269
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.000463125,
+      "loss": 0.6798,
+      "step": 1270
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00046307692307692304,
+      "loss": 0.7521,
+      "step": 1271
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00046302884615384614,
+      "loss": 0.8477,
+      "step": 1272
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00046298076923076924,
+      "loss": 0.7506,
+      "step": 1273
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.0004629326923076923,
+      "loss": 0.7985,
+      "step": 1274
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.0004628846153846154,
+      "loss": 0.8261,
+      "step": 1275
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00046283653846153843,
+      "loss": 0.9711,
+      "step": 1276
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00046278846153846153,
+      "loss": 0.7221,
+      "step": 1277
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00046274038461538463,
+      "loss": 0.9538,
+      "step": 1278
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.0004626923076923077,
+      "loss": 0.8143,
+      "step": 1279
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.0004626442307692308,
+      "loss": 0.8171,
+      "step": 1280
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.0004625961538461538,
+      "loss": 0.9195,
+      "step": 1281
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.0004625480769230769,
+      "loss": 0.8082,
+      "step": 1282
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.0004625,
+      "loss": 0.9279,
+      "step": 1283
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00046245192307692307,
+      "loss": 0.7543,
+      "step": 1284
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.00046240384615384617,
+      "loss": 0.8224,
+      "step": 1285
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.0004623557692307692,
+      "loss": 0.8635,
+      "step": 1286
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.0004623076923076923,
+      "loss": 0.6975,
+      "step": 1287
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.0004622596153846154,
+      "loss": 0.8184,
+      "step": 1288
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00046221153846153846,
+      "loss": 0.8352,
+      "step": 1289
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00046216346153846156,
+      "loss": 0.6907,
+      "step": 1290
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.0004621153846153846,
+      "loss": 0.7764,
+      "step": 1291
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.0004620673076923077,
+      "loss": 0.8294,
+      "step": 1292
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.0004620192307692308,
+      "loss": 0.7689,
+      "step": 1293
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00046197115384615385,
+      "loss": 0.8379,
+      "step": 1294
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00046192307692307695,
+      "loss": 0.7524,
+      "step": 1295
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.000461875,
+      "loss": 0.9234,
+      "step": 1296
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0004618269230769231,
+      "loss": 0.8464,
+      "step": 1297
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.0004617788461538462,
+      "loss": 0.7924,
+      "step": 1298
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00046173076923076924,
+      "loss": 0.6937,
+      "step": 1299
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00046168269230769234,
+      "loss": 0.7243,
+      "step": 1300
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0004616346153846154,
+      "loss": 0.7816,
+      "step": 1301
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0004615865384615385,
+      "loss": 0.6826,
+      "step": 1302
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.0004615384615384616,
+      "loss": 0.8564,
+      "step": 1303
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.0004614903846153846,
+      "loss": 0.8723,
+      "step": 1304
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.0004614423076923077,
+      "loss": 0.8951,
+      "step": 1305
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00046139423076923077,
+      "loss": 1.0317,
+      "step": 1306
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00046134615384615387,
+      "loss": 0.8194,
+      "step": 1307
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00046129807692307697,
+      "loss": 0.8888,
+      "step": 1308
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00046125,
+      "loss": 0.8755,
+      "step": 1309
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.0004612019230769231,
+      "loss": 0.7525,
+      "step": 1310
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00046115384615384616,
+      "loss": 0.8396,
+      "step": 1311
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00046110576923076926,
+      "loss": 0.8314,
+      "step": 1312
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00046105769230769236,
+      "loss": 0.7749,
+      "step": 1313
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.0004610096153846154,
+      "loss": 0.7195,
+      "step": 1314
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.0004609615384615385,
+      "loss": 0.7799,
+      "step": 1315
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00046091346153846155,
+      "loss": 0.7764,
+      "step": 1316
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00046086538461538465,
+      "loss": 0.825,
+      "step": 1317
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.0004608173076923077,
+      "loss": 0.7338,
+      "step": 1318
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0004607692307692308,
+      "loss": 0.7784,
+      "step": 1319
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0004607211538461539,
+      "loss": 0.7813,
+      "step": 1320
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00046067307692307694,
+      "loss": 0.7962,
+      "step": 1321
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00046062500000000004,
+      "loss": 0.7485,
+      "step": 1322
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00046057692307692303,
+      "loss": 0.8142,
+      "step": 1323
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00046052884615384613,
+      "loss": 0.6693,
+      "step": 1324
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00046048076923076923,
+      "loss": 0.715,
+      "step": 1325
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0004604326923076923,
+      "loss": 0.6032,
+      "step": 1326
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0004603846153846154,
+      "loss": 0.7911,
+      "step": 1327
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.0004603365384615384,
+      "loss": 0.7353,
+      "step": 1328
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.0004602884615384615,
+      "loss": 0.7908,
+      "step": 1329
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.0004602403846153846,
+      "loss": 0.8029,
+      "step": 1330
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00046019230769230767,
+      "loss": 0.8293,
+      "step": 1331
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00046014423076923077,
+      "loss": 0.8502,
+      "step": 1332
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.0004600961538461538,
+      "loss": 0.8142,
+      "step": 1333
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.0004600480769230769,
+      "loss": 0.7715,
+      "step": 1334
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00046,
+      "loss": 0.6238,
+      "step": 1335
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00045995192307692306,
+      "loss": 0.8611,
+      "step": 1336
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00045990384615384616,
+      "loss": 1.0082,
+      "step": 1337
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.0004598557692307692,
+      "loss": 0.8657,
+      "step": 1338
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.0004598076923076923,
+      "loss": 0.7995,
+      "step": 1339
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.0004597596153846154,
+      "loss": 0.8178,
+      "step": 1340
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00045971153846153845,
+      "loss": 0.6303,
+      "step": 1341
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00045966346153846155,
+      "loss": 0.8828,
+      "step": 1342
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.0004596153846153846,
+      "loss": 0.8326,
+      "step": 1343
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.0004595673076923077,
+      "loss": 0.7601,
+      "step": 1344
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.0004595192307692308,
+      "loss": 0.6824,
+      "step": 1345
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00045947115384615384,
+      "loss": 0.7298,
+      "step": 1346
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00045942307692307694,
+      "loss": 0.7277,
+      "step": 1347
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.000459375,
+      "loss": 0.9409,
+      "step": 1348
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.0004593269230769231,
+      "loss": 0.8132,
+      "step": 1349
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.0004592788461538462,
+      "loss": 0.7382,
+      "step": 1350
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00045923076923076923,
+      "loss": 0.8463,
+      "step": 1351
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00045918269230769233,
+      "loss": 0.6956,
+      "step": 1352
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.0004591346153846154,
+      "loss": 0.7462,
+      "step": 1353
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.0004590865384615385,
+      "loss": 0.8574,
+      "step": 1354
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.0004590384615384616,
+      "loss": 0.8643,
+      "step": 1355
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.0004589903846153846,
+      "loss": 0.7485,
+      "step": 1356
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.0004589423076923077,
+      "loss": 0.776,
+      "step": 1357
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00045889423076923077,
+      "loss": 0.8113,
+      "step": 1358
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00045884615384615387,
+      "loss": 0.6954,
+      "step": 1359
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00045879807692307697,
+      "loss": 0.6924,
+      "step": 1360
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00045875,
+      "loss": 0.7656,
+      "step": 1361
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.0004587019230769231,
+      "loss": 0.6345,
+      "step": 1362
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00045865384615384616,
+      "loss": 0.9357,
+      "step": 1363
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00045860576923076926,
+      "loss": 0.9322,
+      "step": 1364
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00045855769230769236,
+      "loss": 0.8337,
+      "step": 1365
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.0004585096153846154,
+      "loss": 0.9719,
+      "step": 1366
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.0004584615384615385,
+      "loss": 0.8007,
+      "step": 1367
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00045841346153846155,
+      "loss": 0.8268,
+      "step": 1368
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00045836538461538465,
+      "loss": 0.8355,
+      "step": 1369
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00045831730769230775,
+      "loss": 0.7946,
+      "step": 1370
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.0004582692307692308,
+      "loss": 0.5743,
+      "step": 1371
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.0004582211538461539,
+      "loss": 0.7196,
+      "step": 1372
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00045817307692307694,
+      "loss": 0.7838,
+      "step": 1373
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00045812500000000004,
+      "loss": 0.6426,
+      "step": 1374
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00045807692307692314,
+      "loss": 0.6822,
+      "step": 1375
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.0004580288461538462,
+      "loss": 0.7512,
+      "step": 1376
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.0004579807692307693,
+      "loss": 0.8389,
+      "step": 1377
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.0004579326923076923,
+      "loss": 0.7779,
+      "step": 1378
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.0004578846153846154,
+      "loss": 0.7139,
+      "step": 1379
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00045783653846153847,
+      "loss": 0.7346,
+      "step": 1380
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.0004577884615384615,
+      "loss": 0.7056,
+      "step": 1381
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.0004577403846153846,
+      "loss": 0.7955,
+      "step": 1382
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00045769230769230766,
+      "loss": 0.7453,
+      "step": 1383
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00045764423076923076,
+      "loss": 0.8673,
+      "step": 1384
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.0004575961538461538,
+      "loss": 0.8308,
+      "step": 1385
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.0004575480769230769,
+      "loss": 0.844,
+      "step": 1386
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.0004575,
+      "loss": 0.7526,
+      "step": 1387
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00045745192307692305,
+      "loss": 0.9423,
+      "step": 1388
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00045740384615384615,
+      "loss": 0.7434,
+      "step": 1389
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.0004573557692307692,
+      "loss": 0.9249,
+      "step": 1390
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.0004573076923076923,
+      "loss": 0.8719,
+      "step": 1391
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.0004572596153846154,
+      "loss": 0.861,
+      "step": 1392
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00045721153846153844,
+      "loss": 0.7933,
+      "step": 1393
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00045716346153846154,
+      "loss": 0.7372,
+      "step": 1394
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.0004571153846153846,
+      "loss": 0.8492,
+      "step": 1395
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.0004570673076923077,
+      "loss": 0.8081,
+      "step": 1396
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.0004570192307692308,
+      "loss": 0.9025,
+      "step": 1397
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00045697115384615383,
+      "loss": 0.7786,
+      "step": 1398
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00045692307692307693,
+      "loss": 0.7056,
+      "step": 1399
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.000456875,
+      "loss": 0.7067,
+      "step": 1400
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.0004568269230769231,
+      "loss": 0.8268,
+      "step": 1401
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.0004567788461538462,
+      "loss": 0.7587,
+      "step": 1402
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.0004567307692307692,
+      "loss": 0.7618,
+      "step": 1403
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.0004566826923076923,
+      "loss": 0.7485,
+      "step": 1404
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00045663461538461537,
+      "loss": 0.7717,
+      "step": 1405
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00045658653846153847,
+      "loss": 0.8241,
+      "step": 1406
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00045653846153846157,
+      "loss": 0.8798,
+      "step": 1407
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.0004564903846153846,
+      "loss": 0.7723,
+      "step": 1408
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.0004564423076923077,
+      "loss": 0.7263,
+      "step": 1409
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00045639423076923076,
+      "loss": 0.7656,
+      "step": 1410
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00045634615384615386,
+      "loss": 0.74,
+      "step": 1411
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00045629807692307696,
+      "loss": 0.6236,
+      "step": 1412
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00045625,
+      "loss": 0.6307,
+      "step": 1413
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.0004562019230769231,
+      "loss": 0.7748,
+      "step": 1414
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00045615384615384615,
+      "loss": 0.9258,
+      "step": 1415
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00045610576923076925,
+      "loss": 0.9127,
+      "step": 1416
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00045605769230769235,
+      "loss": 0.9336,
+      "step": 1417
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.0004560096153846154,
+      "loss": 0.8812,
+      "step": 1418
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.0004559615384615385,
+      "loss": 0.8298,
+      "step": 1419
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00045591346153846154,
+      "loss": 0.756,
+      "step": 1420
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00045586538461538464,
+      "loss": 0.7048,
+      "step": 1421
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00045581730769230774,
+      "loss": 0.7416,
+      "step": 1422
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.0004557692307692308,
+      "loss": 0.7336,
+      "step": 1423
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.0004557211538461539,
+      "loss": 0.716,
+      "step": 1424
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00045567307692307693,
+      "loss": 0.5796,
+      "step": 1425
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00045562500000000003,
+      "loss": 0.6163,
+      "step": 1426
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00045557692307692313,
+      "loss": 0.6038,
+      "step": 1427
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.0004555288461538462,
+      "loss": 0.6107,
+      "step": 1428
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.0004554807692307693,
+      "loss": 0.6874,
+      "step": 1429
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.0004554326923076923,
+      "loss": 0.617,
+      "step": 1430
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.0004553846153846154,
+      "loss": 0.6435,
+      "step": 1431
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.0004553365384615385,
+      "loss": 0.7562,
+      "step": 1432
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00045528846153846157,
+      "loss": 0.697,
+      "step": 1433
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00045524038461538466,
+      "loss": 0.6948,
+      "step": 1434
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.0004551923076923077,
+      "loss": 0.5972,
+      "step": 1435
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00045514423076923076,
+      "loss": 0.838,
+      "step": 1436
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00045509615384615386,
+      "loss": 0.59,
+      "step": 1437
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.0004550480769230769,
+      "loss": 0.7382,
+      "step": 1438
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.000455,
+      "loss": 0.8308,
+      "step": 1439
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00045495192307692305,
+      "loss": 0.6061,
+      "step": 1440
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00045490384615384615,
+      "loss": 0.8054,
+      "step": 1441
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00045485576923076925,
+      "loss": 0.704,
+      "step": 1442
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.0004548076923076923,
+      "loss": 0.7807,
+      "step": 1443
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.0004547596153846154,
+      "loss": 0.8625,
+      "step": 1444
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00045471153846153844,
+      "loss": 0.8693,
+      "step": 1445
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00045466346153846154,
+      "loss": 0.8456,
+      "step": 1446
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.0004546153846153846,
+      "loss": 0.8152,
+      "step": 1447
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.0004545673076923077,
+      "loss": 0.6702,
+      "step": 1448
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.0004545192307692308,
+      "loss": 0.7359,
+      "step": 1449
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.0004544711538461538,
+      "loss": 0.8005,
+      "step": 1450
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.0004544230769230769,
+      "loss": 0.6293,
+      "step": 1451
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00045437499999999997,
+      "loss": 0.7381,
+      "step": 1452
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00045432692307692307,
+      "loss": 0.6389,
+      "step": 1453
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00045427884615384617,
+      "loss": 0.7207,
+      "step": 1454
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.0004542307692307692,
+      "loss": 0.675,
+      "step": 1455
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.0004541826923076923,
+      "loss": 0.6756,
+      "step": 1456
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00045413461538461536,
+      "loss": 0.701,
+      "step": 1457
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00045408653846153846,
+      "loss": 0.6693,
+      "step": 1458
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00045403846153846156,
+      "loss": 0.7749,
+      "step": 1459
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.0004539903846153846,
+      "loss": 0.5799,
+      "step": 1460
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.0004539423076923077,
+      "loss": 0.8139,
+      "step": 1461
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00045389423076923075,
+      "loss": 0.6281,
+      "step": 1462
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00045384615384615385,
+      "loss": 0.6224,
+      "step": 1463
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00045379807692307695,
+      "loss": 0.7126,
+      "step": 1464
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00045375,
+      "loss": 0.7097,
+      "step": 1465
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.0004537019230769231,
+      "loss": 0.7584,
+      "step": 1466
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00045365384615384614,
+      "loss": 0.6973,
+      "step": 1467
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00045360576923076924,
+      "loss": 0.8123,
+      "step": 1468
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00045355769230769234,
+      "loss": 0.6595,
+      "step": 1469
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.0004535096153846154,
+      "loss": 0.7558,
+      "step": 1470
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.0004534615384615385,
+      "loss": 0.6884,
+      "step": 1471
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00045341346153846153,
+      "loss": 0.8109,
+      "step": 1472
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00045336538461538463,
+      "loss": 0.8625,
+      "step": 1473
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00045331730769230773,
+      "loss": 0.7259,
+      "step": 1474
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.0004532692307692308,
+      "loss": 0.746,
+      "step": 1475
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.0004532211538461539,
+      "loss": 0.6808,
+      "step": 1476
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.0004531730769230769,
+      "loss": 0.8307,
+      "step": 1477
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.000453125,
+      "loss": 0.8094,
+      "step": 1478
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.0004530769230769231,
+      "loss": 0.6648,
+      "step": 1479
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00045302884615384617,
+      "loss": 0.7375,
+      "step": 1480
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00045298076923076927,
+      "loss": 0.6929,
+      "step": 1481
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.0004529326923076923,
+      "loss": 0.8839,
+      "step": 1482
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.0004528846153846154,
+      "loss": 0.7944,
+      "step": 1483
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.0004528365384615385,
+      "loss": 0.752,
+      "step": 1484
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00045278846153846156,
+      "loss": 0.7412,
+      "step": 1485
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00045274038461538466,
+      "loss": 0.9078,
+      "step": 1486
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.0004526923076923077,
+      "loss": 0.707,
+      "step": 1487
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.0004526442307692308,
+      "loss": 0.6114,
+      "step": 1488
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.0004525961538461539,
+      "loss": 0.9019,
+      "step": 1489
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 0.00045254807692307695,
+      "loss": 0.7453,
+      "step": 1490
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 0.00045250000000000005,
+      "loss": 0.6947,
+      "step": 1491
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 0.00045245192307692304,
+      "loss": 0.6857,
+      "step": 1492
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 0.00045240384615384614,
+      "loss": 0.7273,
+      "step": 1493
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 0.00045235576923076924,
+      "loss": 0.8002,
+      "step": 1494
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.0004523076923076923,
+      "loss": 0.7278,
+      "step": 1495
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.0004522596153846154,
+      "loss": 0.7987,
+      "step": 1496
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 0.00045221153846153843,
+      "loss": 0.7305,
+      "step": 1497
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 0.00045216346153846153,
+      "loss": 0.7215,
+      "step": 1498
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 0.00045211538461538463,
+      "loss": 0.8059,
+      "step": 1499
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 0.0004520673076923077,
+      "loss": 0.8225,
+      "step": 1500
+    },
+    {
+      "epoch": 13.76,
+      "eval_cer": 0.20205700235385837,
+      "eval_loss": 0.6867184638977051,
+      "eval_runtime": 87.3624,
+      "eval_samples_per_second": 18.853,
+      "eval_steps_per_second": 2.358,
+      "eval_wer": 0.6672454294760495,
+      "step": 1500
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 0.0004520192307692308,
+      "loss": 0.8961,
+      "step": 1501
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 0.0004519711538461538,
+      "loss": 0.8002,
+      "step": 1502
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 0.0004519230769230769,
+      "loss": 0.7057,
+      "step": 1503
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 0.000451875,
+      "loss": 0.792,
+      "step": 1504
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 0.00045182692307692307,
+      "loss": 0.6318,
+      "step": 1505
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 0.00045177884615384616,
+      "loss": 0.7239,
+      "step": 1506
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 0.0004517307692307692,
+      "loss": 0.5739,
+      "step": 1507
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 0.0004516826923076923,
+      "loss": 0.6703,
+      "step": 1508
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 0.0004516346153846154,
+      "loss": 0.8205,
+      "step": 1509
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 0.00045158653846153846,
+      "loss": 0.6822,
+      "step": 1510
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 0.00045153846153846156,
+      "loss": 0.6428,
+      "step": 1511
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 0.0004514903846153846,
+      "loss": 0.753,
+      "step": 1512
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 0.0004514423076923077,
+      "loss": 0.6914,
+      "step": 1513
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 0.00045139423076923075,
+      "loss": 0.7549,
+      "step": 1514
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 0.00045134615384615385,
+      "loss": 0.5862,
+      "step": 1515
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 0.00045129807692307695,
+      "loss": 0.7832,
+      "step": 1516
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 0.00045125,
+      "loss": 0.7019,
+      "step": 1517
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 0.0004512019230769231,
+      "loss": 0.7391,
+      "step": 1518
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.00045115384615384614,
+      "loss": 0.8751,
+      "step": 1519
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.00045110576923076924,
+      "loss": 0.7198,
+      "step": 1520
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.00045105769230769234,
+      "loss": 0.6233,
+      "step": 1521
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 0.0004510096153846154,
+      "loss": 0.6398,
+      "step": 1522
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 0.0004509615384615385,
+      "loss": 0.7433,
+      "step": 1523
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 0.0004509134615384615,
+      "loss": 0.9008,
+      "step": 1524
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 0.0004508653846153846,
+      "loss": 0.7451,
+      "step": 1525
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 0.0004508173076923077,
+      "loss": 0.7771,
+      "step": 1526
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 0.00045076923076923077,
+      "loss": 0.8463,
+      "step": 1527
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 0.00045072115384615387,
+      "loss": 0.8217,
+      "step": 1528
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 0.0004506730769230769,
+      "loss": 0.7891,
+      "step": 1529
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 0.000450625,
+      "loss": 0.6963,
+      "step": 1530
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 0.0004505769230769231,
+      "loss": 0.8044,
+      "step": 1531
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.00045052884615384616,
+      "loss": 0.7576,
+      "step": 1532
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.00045048076923076926,
+      "loss": 0.7293,
+      "step": 1533
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 0.0004504326923076923,
+      "loss": 0.6363,
+      "step": 1534
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 0.0004503846153846154,
+      "loss": 0.6248,
+      "step": 1535
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 0.0004503365384615385,
+      "loss": 0.6265,
+      "step": 1536
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 0.00045028846153846155,
+      "loss": 0.689,
+      "step": 1537
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 0.00045024038461538465,
+      "loss": 0.6814,
+      "step": 1538
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 0.0004501923076923077,
+      "loss": 0.6959,
+      "step": 1539
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 0.0004501442307692308,
+      "loss": 0.6264,
+      "step": 1540
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.0004500961538461539,
+      "loss": 0.6893,
+      "step": 1541
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 0.00045004807692307694,
+      "loss": 0.689,
+      "step": 1542
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 0.7692,
+      "step": 1543
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.0004499519230769231,
+      "loss": 0.6165,
+      "step": 1544
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.0004499038461538462,
+      "loss": 0.7808,
+      "step": 1545
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 0.0004498557692307693,
+      "loss": 0.7954,
+      "step": 1546
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 0.00044980769230769233,
+      "loss": 0.6552,
+      "step": 1547
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 0.00044975961538461543,
+      "loss": 0.8069,
+      "step": 1548
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 0.0004497115384615384,
+      "loss": 0.8248,
+      "step": 1549
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 0.0004496634615384615,
+      "loss": 0.58,
+      "step": 1550
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 0.0004496153846153846,
+      "loss": 0.6577,
+      "step": 1551
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 0.00044956730769230767,
+      "loss": 0.7671,
+      "step": 1552
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 0.00044951923076923077,
+      "loss": 0.6828,
+      "step": 1553
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 0.0004494711538461538,
+      "loss": 0.7891,
+      "step": 1554
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 0.0004494230769230769,
+      "loss": 0.7714,
+      "step": 1555
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.000449375,
+      "loss": 0.7471,
+      "step": 1556
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.00044932692307692306,
+      "loss": 0.7422,
+      "step": 1557
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 0.00044927884615384616,
+      "loss": 0.6876,
+      "step": 1558
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 0.0004492307692307692,
+      "loss": 0.7204,
+      "step": 1559
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 0.0004491826923076923,
+      "loss": 0.5998,
+      "step": 1560
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 0.0004491346153846154,
+      "loss": 0.7713,
+      "step": 1561
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 0.00044908653846153845,
+      "loss": 0.6062,
+      "step": 1562
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 0.00044903846153846155,
+      "loss": 0.6756,
+      "step": 1563
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 0.0004489903846153846,
+      "loss": 0.7225,
+      "step": 1564
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 0.0004489423076923077,
+      "loss": 0.6555,
+      "step": 1565
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 0.0004488942307692308,
+      "loss": 0.7615,
+      "step": 1566
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 0.00044884615384615384,
+      "loss": 0.7675,
+      "step": 1567
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 0.00044879807692307694,
+      "loss": 0.9288,
+      "step": 1568
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 0.00044875,
+      "loss": 0.7582,
+      "step": 1569
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 0.0004487019230769231,
+      "loss": 0.8542,
+      "step": 1570
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 0.0004486538461538462,
+      "loss": 0.8254,
+      "step": 1571
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 0.00044860576923076923,
+      "loss": 0.7587,
+      "step": 1572
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 0.00044855769230769233,
+      "loss": 0.7487,
+      "step": 1573
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 0.0004485096153846154,
+      "loss": 0.7836,
+      "step": 1574
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 0.0004484615384615385,
+      "loss": 0.6371,
+      "step": 1575
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 0.0004484134615384615,
+      "loss": 0.7162,
+      "step": 1576
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 0.0004483653846153846,
+      "loss": 0.6344,
+      "step": 1577
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 0.0004483173076923077,
+      "loss": 0.7354,
+      "step": 1578
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 0.00044826923076923076,
+      "loss": 0.8861,
+      "step": 1579
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.00044822115384615386,
+      "loss": 0.886,
+      "step": 1580
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.0004481730769230769,
+      "loss": 0.894,
+      "step": 1581
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 0.000448125,
+      "loss": 0.8231,
+      "step": 1582
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 0.0004480769230769231,
+      "loss": 0.8045,
+      "step": 1583
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 0.00044802884615384616,
+      "loss": 0.7949,
+      "step": 1584
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 0.00044798076923076925,
+      "loss": 0.7278,
+      "step": 1585
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 0.0004479326923076923,
+      "loss": 0.8566,
+      "step": 1586
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 0.0004478846153846154,
+      "loss": 0.859,
+      "step": 1587
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 0.0004478365384615385,
+      "loss": 0.7778,
+      "step": 1588
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 0.00044778846153846155,
+      "loss": 0.6737,
+      "step": 1589
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 0.00044774038461538465,
+      "loss": 0.7255,
+      "step": 1590
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 0.0004476923076923077,
+      "loss": 0.758,
+      "step": 1591
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.0004476442307692308,
+      "loss": 0.6352,
+      "step": 1592
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.0004475961538461539,
+      "loss": 0.5953,
+      "step": 1593
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 0.00044754807692307694,
+      "loss": 0.7525,
+      "step": 1594
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 0.00044750000000000004,
+      "loss": 0.821,
+      "step": 1595
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 0.0004474519230769231,
+      "loss": 0.7952,
+      "step": 1596
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 0.0004474038461538462,
+      "loss": 0.7641,
+      "step": 1597
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 0.0004473557692307693,
+      "loss": 0.8011,
+      "step": 1598
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 0.0004473076923076923,
+      "loss": 0.839,
+      "step": 1599
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 0.0004472596153846154,
+      "loss": 0.7526,
+      "step": 1600
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 0.00044721153846153847,
+      "loss": 0.7097,
+      "step": 1601
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 0.00044716346153846157,
+      "loss": 0.8238,
+      "step": 1602
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 0.00044711538461538467,
+      "loss": 0.6494,
+      "step": 1603
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 0.0004470673076923077,
+      "loss": 0.7702,
+      "step": 1604
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 0.0004470192307692308,
+      "loss": 0.7047,
+      "step": 1605
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 0.0004469711538461538,
+      "loss": 0.78,
+      "step": 1606
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 0.0004469230769230769,
+      "loss": 0.8144,
+      "step": 1607
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 0.000446875,
+      "loss": 0.8516,
+      "step": 1608
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 0.00044682692307692305,
+      "loss": 0.8334,
+      "step": 1609
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 0.00044677884615384615,
+      "loss": 0.8369,
+      "step": 1610
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 0.0004467307692307692,
+      "loss": 0.8042,
+      "step": 1611
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 0.0004466826923076923,
+      "loss": 0.7973,
+      "step": 1612
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 0.0004466346153846154,
+      "loss": 0.6708,
+      "step": 1613
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 0.00044658653846153844,
+      "loss": 0.7989,
+      "step": 1614
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 0.00044653846153846154,
+      "loss": 0.711,
+      "step": 1615
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.0004464903846153846,
+      "loss": 0.6435,
+      "step": 1616
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.0004464423076923077,
+      "loss": 0.7147,
+      "step": 1617
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 0.0004463942307692308,
+      "loss": 0.7548,
+      "step": 1618
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 0.00044634615384615383,
+      "loss": 0.6738,
+      "step": 1619
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 0.00044629807692307693,
+      "loss": 0.7285,
+      "step": 1620
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 0.00044625,
+      "loss": 0.7439,
+      "step": 1621
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 0.0004462019230769231,
+      "loss": 0.7441,
+      "step": 1622
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 0.0004461538461538462,
+      "loss": 0.7742,
+      "step": 1623
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 0.0004461057692307692,
+      "loss": 0.6163,
+      "step": 1624
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 0.0004460576923076923,
+      "loss": 0.6908,
+      "step": 1625
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 0.00044600961538461537,
+      "loss": 0.8546,
+      "step": 1626
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 0.00044596153846153847,
+      "loss": 0.7869,
+      "step": 1627
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.00044591346153846157,
+      "loss": 0.7212,
+      "step": 1628
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.0004458653846153846,
+      "loss": 0.7731,
+      "step": 1629
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 0.0004458173076923077,
+      "loss": 0.6444,
+      "step": 1630
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 0.00044576923076923076,
+      "loss": 0.8877,
+      "step": 1631
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 0.00044572115384615386,
+      "loss": 0.8718,
+      "step": 1632
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 0.00044567307692307696,
+      "loss": 0.8241,
+      "step": 1633
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 0.000445625,
+      "loss": 0.6742,
+      "step": 1634
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.0004455769230769231,
+      "loss": 0.8742,
+      "step": 1635
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 0.00044552884615384615,
+      "loss": 0.8498,
+      "step": 1636
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 0.00044548076923076925,
+      "loss": 0.7519,
+      "step": 1637
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 0.00044543269230769235,
+      "loss": 0.8734,
+      "step": 1638
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 0.0004453846153846154,
+      "loss": 0.8106,
+      "step": 1639
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 0.0004453365384615385,
+      "loss": 0.6862,
+      "step": 1640
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.00044528846153846154,
+      "loss": 0.6955,
+      "step": 1641
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.00044524038461538464,
+      "loss": 0.7147,
+      "step": 1642
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 0.0004451923076923077,
+      "loss": 0.6847,
+      "step": 1643
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 0.0004451442307692308,
+      "loss": 0.7805,
+      "step": 1644
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 0.0004450961538461539,
+      "loss": 0.6643,
+      "step": 1645
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 0.00044504807692307693,
+      "loss": 0.8088,
+      "step": 1646
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 0.00044500000000000003,
+      "loss": 0.5885,
+      "step": 1647
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 0.0004449519230769231,
+      "loss": 0.6709,
+      "step": 1648
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 0.0004449038461538462,
+      "loss": 0.8452,
+      "step": 1649
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 0.0004448557692307693,
+      "loss": 0.6668,
+      "step": 1650
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 0.0004448076923076923,
+      "loss": 0.5638,
+      "step": 1651
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 0.0004447596153846154,
+      "loss": 0.5573,
+      "step": 1652
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.00044471153846153846,
+      "loss": 0.7107,
+      "step": 1653
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.00044466346153846156,
+      "loss": 0.6424,
+      "step": 1654
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 0.00044461538461538466,
+      "loss": 0.7435,
+      "step": 1655
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 0.0004445673076923077,
+      "loss": 0.7065,
+      "step": 1656
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 0.0004445192307692308,
+      "loss": 0.6479,
+      "step": 1657
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 0.00044447115384615385,
+      "loss": 0.703,
+      "step": 1658
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 0.00044442307692307695,
+      "loss": 0.4978,
+      "step": 1659
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 0.00044437500000000005,
+      "loss": 0.8836,
+      "step": 1660
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 0.0004443269230769231,
+      "loss": 0.5898,
+      "step": 1661
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 0.00044427884615384615,
+      "loss": 0.7053,
+      "step": 1662
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 0.0004442307692307692,
+      "loss": 0.9121,
+      "step": 1663
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 0.0004441826923076923,
+      "loss": 0.8241,
+      "step": 1664
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 0.0004441346153846154,
+      "loss": 0.6532,
+      "step": 1665
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 0.00044408653846153844,
+      "loss": 0.7321,
+      "step": 1666
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 0.00044403846153846154,
+      "loss": 0.8355,
+      "step": 1667
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 0.0004439903846153846,
+      "loss": 0.7695,
+      "step": 1668
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 0.0004439423076923077,
+      "loss": 0.7468,
+      "step": 1669
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 0.0004438942307692308,
+      "loss": 0.6798,
+      "step": 1670
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 0.0004438461538461538,
+      "loss": 0.7441,
+      "step": 1671
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 0.0004437980769230769,
+      "loss": 0.8506,
+      "step": 1672
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 0.00044374999999999997,
+      "loss": 0.7371,
+      "step": 1673
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 0.00044370192307692307,
+      "loss": 0.7139,
+      "step": 1674
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 0.00044365384615384617,
+      "loss": 0.7214,
+      "step": 1675
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 0.0004436057692307692,
+      "loss": 0.7767,
+      "step": 1676
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.0004435576923076923,
+      "loss": 0.7306,
+      "step": 1677
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.00044350961538461536,
+      "loss": 0.7551,
+      "step": 1678
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 0.00044346153846153846,
+      "loss": 0.67,
+      "step": 1679
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 0.00044341346153846156,
+      "loss": 0.7447,
+      "step": 1680
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 0.0004433653846153846,
+      "loss": 0.7754,
+      "step": 1681
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 0.0004433173076923077,
+      "loss": 0.729,
+      "step": 1682
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 0.00044326923076923075,
+      "loss": 0.784,
+      "step": 1683
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 0.00044322115384615385,
+      "loss": 0.7587,
+      "step": 1684
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 0.00044317307692307695,
+      "loss": 0.6613,
+      "step": 1685
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 0.000443125,
+      "loss": 0.8335,
+      "step": 1686
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 0.0004430769230769231,
+      "loss": 0.8728,
+      "step": 1687
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 0.00044302884615384614,
+      "loss": 0.6537,
+      "step": 1688
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.00044298076923076924,
+      "loss": 0.686,
+      "step": 1689
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.00044293269230769234,
+      "loss": 0.7718,
+      "step": 1690
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 0.0004428846153846154,
+      "loss": 0.7117,
+      "step": 1691
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 0.0004428365384615385,
+      "loss": 0.8117,
+      "step": 1692
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 0.00044278846153846153,
+      "loss": 0.726,
+      "step": 1693
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 0.00044274038461538463,
+      "loss": 0.7307,
+      "step": 1694
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 0.00044269230769230773,
+      "loss": 0.7527,
+      "step": 1695
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 0.0004426442307692308,
+      "loss": 0.7269,
+      "step": 1696
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 0.0004425961538461539,
+      "loss": 0.7592,
+      "step": 1697
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 0.0004425480769230769,
+      "loss": 0.7388,
+      "step": 1698
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 0.0004425,
+      "loss": 0.672,
+      "step": 1699
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 0.0004424519230769231,
+      "loss": 0.6629,
+      "step": 1700
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 0.00044240384615384617,
+      "loss": 0.7393,
+      "step": 1701
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 0.00044235576923076927,
+      "loss": 0.6466,
+      "step": 1702
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 0.0004423076923076923,
+      "loss": 0.7462,
+      "step": 1703
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 0.0004422596153846154,
+      "loss": 0.6442,
+      "step": 1704
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 0.0004422115384615385,
+      "loss": 0.7221,
+      "step": 1705
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 0.00044216346153846156,
+      "loss": 0.7663,
+      "step": 1706
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 0.00044211538461538466,
+      "loss": 0.7101,
+      "step": 1707
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 0.0004420673076923077,
+      "loss": 0.6749,
+      "step": 1708
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 0.0004420192307692308,
+      "loss": 0.7038,
+      "step": 1709
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 0.00044197115384615385,
+      "loss": 0.7537,
+      "step": 1710
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 0.00044192307692307695,
+      "loss": 0.6633,
+      "step": 1711
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 0.00044187500000000005,
+      "loss": 0.7448,
+      "step": 1712
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.0004418269230769231,
+      "loss": 0.7114,
+      "step": 1713
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.0004417788461538462,
+      "loss": 0.6901,
+      "step": 1714
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 0.00044173076923076924,
+      "loss": 0.7694,
+      "step": 1715
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 0.00044168269230769234,
+      "loss": 0.6023,
+      "step": 1716
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 0.00044163461538461544,
+      "loss": 0.7568,
+      "step": 1717
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 0.00044158653846153843,
+      "loss": 0.8226,
+      "step": 1718
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 0.00044153846153846153,
+      "loss": 0.7683,
+      "step": 1719
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 0.0004414903846153846,
+      "loss": 0.7837,
+      "step": 1720
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 0.0004414423076923077,
+      "loss": 0.7052,
+      "step": 1721
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 0.0004413942307692308,
+      "loss": 0.8421,
+      "step": 1722
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 0.0004413461538461538,
+      "loss": 0.7666,
+      "step": 1723
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 0.0004412980769230769,
+      "loss": 0.725,
+      "step": 1724
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.00044124999999999996,
+      "loss": 0.7128,
+      "step": 1725
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.00044120192307692306,
+      "loss": 0.6366,
+      "step": 1726
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 0.00044115384615384616,
+      "loss": 0.6149,
+      "step": 1727
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 0.0004411057692307692,
+      "loss": 0.6841,
+      "step": 1728
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.0004410576923076923,
+      "loss": 0.6895,
+      "step": 1729
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 0.00044100961538461535,
+      "loss": 0.7508,
+      "step": 1730
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 0.00044096153846153845,
+      "loss": 0.8152,
+      "step": 1731
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 0.00044091346153846155,
+      "loss": 0.8885,
+      "step": 1732
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 0.0004408653846153846,
+      "loss": 0.7304,
+      "step": 1733
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 0.0004408173076923077,
+      "loss": 0.5762,
+      "step": 1734
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 0.00044076923076923075,
+      "loss": 0.78,
+      "step": 1735
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 0.00044072115384615384,
+      "loss": 0.6862,
+      "step": 1736
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 0.00044067307692307694,
+      "loss": 0.7178,
+      "step": 1737
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 0.000440625,
+      "loss": 0.7177,
+      "step": 1738
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 0.0004405769230769231,
+      "loss": 0.713,
+      "step": 1739
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 0.00044052884615384614,
+      "loss": 0.6755,
+      "step": 1740
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 0.00044048076923076924,
+      "loss": 0.5891,
+      "step": 1741
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 0.00044043269230769233,
+      "loss": 0.7649,
+      "step": 1742
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 0.0004403846153846154,
+      "loss": 0.7549,
+      "step": 1743
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 0.0004403365384615385,
+      "loss": 0.6971,
+      "step": 1744
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 0.0004402884615384615,
+      "loss": 0.7546,
+      "step": 1745
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 0.0004402403846153846,
+      "loss": 0.6932,
+      "step": 1746
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 0.0004401923076923077,
+      "loss": 0.6922,
+      "step": 1747
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 0.00044014423076923077,
+      "loss": 0.6849,
+      "step": 1748
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 0.00044009615384615387,
+      "loss": 0.7421,
+      "step": 1749
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.0004400480769230769,
+      "loss": 0.6847,
+      "step": 1750
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.00044,
+      "loss": 0.7409,
+      "step": 1751
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 0.0004399519230769231,
+      "loss": 0.6656,
+      "step": 1752
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 0.00043990384615384616,
+      "loss": 0.6169,
+      "step": 1753
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 0.00043985576923076926,
+      "loss": 0.6443,
+      "step": 1754
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 0.0004398076923076923,
+      "loss": 0.6255,
+      "step": 1755
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 0.0004397596153846154,
+      "loss": 0.6855,
+      "step": 1756
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 0.0004397115384615385,
+      "loss": 0.7278,
+      "step": 1757
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 0.00043966346153846155,
+      "loss": 0.6855,
+      "step": 1758
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 0.00043961538461538465,
+      "loss": 0.7043,
+      "step": 1759
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 0.0004395673076923077,
+      "loss": 0.6287,
+      "step": 1760
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 0.0004395192307692308,
+      "loss": 0.7275,
+      "step": 1761
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 0.0004394711538461539,
+      "loss": 0.6199,
+      "step": 1762
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 0.00043942307692307694,
+      "loss": 0.6607,
+      "step": 1763
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 0.00043937500000000004,
+      "loss": 0.8053,
+      "step": 1764
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 0.0004393269230769231,
+      "loss": 0.6391,
+      "step": 1765
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 0.0004392788461538462,
+      "loss": 0.5894,
+      "step": 1766
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 0.0004392307692307693,
+      "loss": 0.5825,
+      "step": 1767
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 0.00043918269230769233,
+      "loss": 0.7377,
+      "step": 1768
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 0.00043913461538461543,
+      "loss": 0.5808,
+      "step": 1769
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 0.0004390865384615385,
+      "loss": 0.6499,
+      "step": 1770
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 0.0004390384615384616,
+      "loss": 0.6795,
+      "step": 1771
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 0.0004389903846153846,
+      "loss": 0.8477,
+      "step": 1772
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 0.0004389423076923077,
+      "loss": 0.7902,
+      "step": 1773
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.0004388942307692308,
+      "loss": 0.6881,
+      "step": 1774
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.0004388461538461538,
+      "loss": 0.7236,
+      "step": 1775
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 0.0004387980769230769,
+      "loss": 0.6231,
+      "step": 1776
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 0.00043874999999999996,
+      "loss": 0.6899,
+      "step": 1777
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 0.00043870192307692306,
+      "loss": 0.6283,
+      "step": 1778
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 0.00043865384615384616,
+      "loss": 0.7183,
+      "step": 1779
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 0.0004386057692307692,
+      "loss": 0.6016,
+      "step": 1780
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 0.0004385576923076923,
+      "loss": 0.7712,
+      "step": 1781
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 0.00043850961538461535,
+      "loss": 0.6325,
+      "step": 1782
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 0.00043846153846153845,
+      "loss": 0.6646,
+      "step": 1783
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 0.00043841346153846155,
+      "loss": 0.5337,
+      "step": 1784
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 0.0004383653846153846,
+      "loss": 0.7186,
+      "step": 1785
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.0004383173076923077,
+      "loss": 0.71,
+      "step": 1786
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.00043826923076923074,
+      "loss": 0.6464,
+      "step": 1787
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 0.00043822115384615384,
+      "loss": 0.6409,
+      "step": 1788
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 0.00043817307692307694,
+      "loss": 0.6019,
+      "step": 1789
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 0.000438125,
+      "loss": 0.7723,
+      "step": 1790
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 0.0004380769230769231,
+      "loss": 0.6327,
+      "step": 1791
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 0.00043802884615384613,
+      "loss": 0.6868,
+      "step": 1792
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 0.00043798076923076923,
+      "loss": 0.7222,
+      "step": 1793
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 0.00043793269230769233,
+      "loss": 0.6638,
+      "step": 1794
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 0.0004378846153846154,
+      "loss": 0.6,
+      "step": 1795
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 0.0004378365384615385,
+      "loss": 0.6447,
+      "step": 1796
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 0.0004377884615384615,
+      "loss": 0.6335,
+      "step": 1797
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 0.0004377403846153846,
+      "loss": 0.7869,
+      "step": 1798
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 0.0004376923076923077,
+      "loss": 0.837,
+      "step": 1799
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 0.00043764423076923076,
+      "loss": 0.7482,
+      "step": 1800
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 0.00043759615384615386,
+      "loss": 0.646,
+      "step": 1801
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 0.0004375480769230769,
+      "loss": 0.7736,
+      "step": 1802
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 0.0004375,
+      "loss": 0.711,
+      "step": 1803
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 0.0004374519230769231,
+      "loss": 0.8033,
+      "step": 1804
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 0.00043740384615384615,
+      "loss": 0.7245,
+      "step": 1805
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 0.00043735576923076925,
+      "loss": 0.5373,
+      "step": 1806
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 0.0004373076923076923,
+      "loss": 0.6469,
+      "step": 1807
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 0.0004372596153846154,
+      "loss": 0.7062,
+      "step": 1808
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 0.0004372115384615385,
+      "loss": 0.7209,
+      "step": 1809
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.00043716346153846154,
+      "loss": 0.7462,
+      "step": 1810
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.00043711538461538464,
+      "loss": 0.6681,
+      "step": 1811
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 0.0004370673076923077,
+      "loss": 0.6935,
+      "step": 1812
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 0.0004370192307692308,
+      "loss": 0.6258,
+      "step": 1813
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 0.0004369711538461539,
+      "loss": 0.7119,
+      "step": 1814
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 0.00043692307692307693,
+      "loss": 0.6642,
+      "step": 1815
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 0.00043687500000000003,
+      "loss": 0.7628,
+      "step": 1816
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 0.0004368269230769231,
+      "loss": 0.63,
+      "step": 1817
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 0.0004367788461538462,
+      "loss": 0.7131,
+      "step": 1818
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 0.0004367307692307693,
+      "loss": 0.7551,
+      "step": 1819
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 0.0004366826923076923,
+      "loss": 0.7056,
+      "step": 1820
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 0.0004366346153846154,
+      "loss": 0.6633,
+      "step": 1821
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.00043658653846153847,
+      "loss": 0.8084,
+      "step": 1822
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.00043653846153846157,
+      "loss": 0.6787,
+      "step": 1823
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 0.00043649038461538467,
+      "loss": 0.6797,
+      "step": 1824
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 0.0004364423076923077,
+      "loss": 0.6318,
+      "step": 1825
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 0.0004363942307692308,
+      "loss": 0.8276,
+      "step": 1826
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 0.00043634615384615386,
+      "loss": 0.7554,
+      "step": 1827
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 0.00043629807692307696,
+      "loss": 0.8407,
+      "step": 1828
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 0.00043625000000000006,
+      "loss": 0.8,
+      "step": 1829
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 0.0004362019230769231,
+      "loss": 0.8053,
+      "step": 1830
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 0.0004361538461538462,
+      "loss": 0.6877,
+      "step": 1831
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 0.0004361057692307692,
+      "loss": 0.7385,
+      "step": 1832
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 0.0004360576923076923,
+      "loss": 0.6774,
+      "step": 1833
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.0004360096153846154,
+      "loss": 0.6622,
+      "step": 1834
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.00043596153846153844,
+      "loss": 0.785,
+      "step": 1835
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 0.00043591346153846154,
+      "loss": 0.6206,
+      "step": 1836
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 0.0004358653846153846,
+      "loss": 0.6374,
+      "step": 1837
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 0.0004358173076923077,
+      "loss": 0.6213,
+      "step": 1838
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 0.00043576923076923073,
+      "loss": 0.7158,
+      "step": 1839
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 0.00043572115384615383,
+      "loss": 0.665,
+      "step": 1840
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 0.00043567307692307693,
+      "loss": 0.6208,
+      "step": 1841
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 0.000435625,
+      "loss": 0.7167,
+      "step": 1842
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 0.0004355769230769231,
+      "loss": 0.6972,
+      "step": 1843
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 0.0004355288461538461,
+      "loss": 0.6014,
+      "step": 1844
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 0.0004354807692307692,
+      "loss": 0.7326,
+      "step": 1845
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.0004354326923076923,
+      "loss": 0.6821,
+      "step": 1846
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.00043538461538461537,
+      "loss": 0.6059,
+      "step": 1847
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 0.00043533653846153847,
+      "loss": 0.6204,
+      "step": 1848
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 0.0004352884615384615,
+      "loss": 0.757,
+      "step": 1849
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 0.0004352403846153846,
+      "loss": 0.5981,
+      "step": 1850
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 0.0004351923076923077,
+      "loss": 0.6469,
+      "step": 1851
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 0.00043514423076923076,
+      "loss": 0.5484,
+      "step": 1852
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 0.00043509615384615386,
+      "loss": 0.7692,
+      "step": 1853
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 0.0004350480769230769,
+      "loss": 0.7159,
+      "step": 1854
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 0.000435,
+      "loss": 0.6718,
+      "step": 1855
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 0.0004349519230769231,
+      "loss": 0.7354,
+      "step": 1856
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 0.00043490384615384615,
+      "loss": 0.6643,
+      "step": 1857
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 0.00043485576923076925,
+      "loss": 0.6377,
+      "step": 1858
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 0.0004348076923076923,
+      "loss": 0.7441,
+      "step": 1859
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 0.0004347596153846154,
+      "loss": 0.6043,
+      "step": 1860
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 0.0004347115384615385,
+      "loss": 0.557,
+      "step": 1861
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 0.00043466346153846154,
+      "loss": 0.6954,
+      "step": 1862
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 0.00043461538461538464,
+      "loss": 0.66,
+      "step": 1863
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 0.0004345673076923077,
+      "loss": 0.6767,
+      "step": 1864
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 0.0004345192307692308,
+      "loss": 0.6144,
+      "step": 1865
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 0.0004344711538461539,
+      "loss": 0.6235,
+      "step": 1866
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 0.00043442307692307693,
+      "loss": 0.6687,
+      "step": 1867
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 0.00043437500000000003,
+      "loss": 0.7673,
+      "step": 1868
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 0.0004343269230769231,
+      "loss": 0.6595,
+      "step": 1869
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 0.0004342788461538462,
+      "loss": 0.5276,
+      "step": 1870
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 0.0004342307692307693,
+      "loss": 0.6267,
+      "step": 1871
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 0.0004341826923076923,
+      "loss": 0.6127,
+      "step": 1872
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 0.0004341346153846154,
+      "loss": 0.6246,
+      "step": 1873
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 0.00043408653846153846,
+      "loss": 0.7635,
+      "step": 1874
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 0.00043403846153846156,
+      "loss": 0.6703,
+      "step": 1875
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 0.00043399038461538466,
+      "loss": 0.6735,
+      "step": 1876
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 0.0004339423076923077,
+      "loss": 0.6771,
+      "step": 1877
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 0.0004338942307692308,
+      "loss": 0.6648,
+      "step": 1878
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 0.00043384615384615385,
+      "loss": 0.7665,
+      "step": 1879
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 0.00043379807692307695,
+      "loss": 0.6297,
+      "step": 1880
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 0.00043375000000000005,
+      "loss": 0.8255,
+      "step": 1881
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 0.0004337019230769231,
+      "loss": 0.7989,
+      "step": 1882
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 0.0004336538461538462,
+      "loss": 0.7502,
+      "step": 1883
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 0.00043360576923076924,
+      "loss": 0.7415,
+      "step": 1884
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 0.00043355769230769234,
+      "loss": 0.7193,
+      "step": 1885
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 0.00043350961538461544,
+      "loss": 0.6539,
+      "step": 1886
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 0.0004334615384615385,
+      "loss": 0.5909,
+      "step": 1887
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 0.0004334134615384616,
+      "loss": 0.5294,
+      "step": 1888
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 0.0004333653846153846,
+      "loss": 0.7206,
+      "step": 1889
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 0.0004333173076923077,
+      "loss": 0.6572,
+      "step": 1890
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 0.0004332692307692308,
+      "loss": 0.6339,
+      "step": 1891
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 0.0004332211538461538,
+      "loss": 0.5458,
+      "step": 1892
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 0.0004331730769230769,
+      "loss": 0.6295,
+      "step": 1893
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 0.00043312499999999997,
+      "loss": 0.6665,
+      "step": 1894
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 0.00043307692307692307,
+      "loss": 0.8051,
+      "step": 1895
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 0.00043302884615384617,
+      "loss": 0.6863,
+      "step": 1896
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 0.0004329807692307692,
+      "loss": 0.6844,
+      "step": 1897
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 0.0004329326923076923,
+      "loss": 0.7048,
+      "step": 1898
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 0.00043288461538461536,
+      "loss": 0.764,
+      "step": 1899
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 0.00043283653846153846,
+      "loss": 0.5901,
+      "step": 1900
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 0.0004327884615384615,
+      "loss": 0.7418,
+      "step": 1901
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 0.0004327403846153846,
+      "loss": 0.5911,
+      "step": 1902
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 0.0004326923076923077,
+      "loss": 0.606,
+      "step": 1903
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 0.00043264423076923075,
+      "loss": 0.7788,
+      "step": 1904
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 0.00043259615384615385,
+      "loss": 0.6924,
+      "step": 1905
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 0.0004325480769230769,
+      "loss": 0.7193,
+      "step": 1906
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 0.0004325,
+      "loss": 0.6717,
+      "step": 1907
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 0.0004324519230769231,
+      "loss": 0.7206,
+      "step": 1908
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 0.00043240384615384614,
+      "loss": 0.7007,
+      "step": 1909
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 0.00043235576923076924,
+      "loss": 0.726,
+      "step": 1910
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 0.0004323076923076923,
+      "loss": 0.6898,
+      "step": 1911
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 0.0004322596153846154,
+      "loss": 0.6781,
+      "step": 1912
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 0.0004322115384615385,
+      "loss": 0.7143,
+      "step": 1913
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 0.00043216346153846153,
+      "loss": 0.6935,
+      "step": 1914
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 0.00043211538461538463,
+      "loss": 0.6016,
+      "step": 1915
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 0.0004320673076923077,
+      "loss": 0.5595,
+      "step": 1916
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 0.0004320192307692308,
+      "loss": 0.7,
+      "step": 1917
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 0.0004319711538461539,
+      "loss": 0.6265,
+      "step": 1918
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 0.0004319230769230769,
+      "loss": 0.6496,
+      "step": 1919
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 0.000431875,
+      "loss": 0.6143,
+      "step": 1920
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 0.00043182692307692307,
+      "loss": 0.5895,
+      "step": 1921
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 0.00043177884615384617,
+      "loss": 0.715,
+      "step": 1922
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 0.00043173076923076927,
+      "loss": 0.6791,
+      "step": 1923
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 0.0004316826923076923,
+      "loss": 0.6326,
+      "step": 1924
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 0.0004316346153846154,
+      "loss": 0.6391,
+      "step": 1925
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 0.00043158653846153846,
+      "loss": 0.6837,
+      "step": 1926
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 0.00043153846153846156,
+      "loss": 0.6943,
+      "step": 1927
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 0.00043149038461538466,
+      "loss": 0.7084,
+      "step": 1928
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 0.0004314423076923077,
+      "loss": 0.6707,
+      "step": 1929
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 0.0004313942307692308,
+      "loss": 0.618,
+      "step": 1930
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 0.00043134615384615385,
+      "loss": 0.5509,
+      "step": 1931
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 0.00043129807692307695,
+      "loss": 0.5774,
+      "step": 1932
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 0.00043125000000000005,
+      "loss": 0.6362,
+      "step": 1933
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 0.0004312019230769231,
+      "loss": 0.5938,
+      "step": 1934
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 0.0004311538461538462,
+      "loss": 0.8928,
+      "step": 1935
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 0.00043110576923076924,
+      "loss": 0.7489,
+      "step": 1936
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 0.00043105769230769234,
+      "loss": 0.7857,
+      "step": 1937
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 0.00043100961538461544,
+      "loss": 0.7626,
+      "step": 1938
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 0.0004309615384615385,
+      "loss": 0.715,
+      "step": 1939
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 0.0004309134615384616,
+      "loss": 0.6984,
+      "step": 1940
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 0.00043086538461538463,
+      "loss": 0.6929,
+      "step": 1941
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 0.00043081730769230773,
+      "loss": 0.7767,
+      "step": 1942
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 0.00043076923076923083,
+      "loss": 0.671,
+      "step": 1943
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 0.0004307211538461539,
+      "loss": 0.6422,
+      "step": 1944
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 0.0004306730769230769,
+      "loss": 0.6943,
+      "step": 1945
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 0.00043062499999999996,
+      "loss": 0.6256,
+      "step": 1946
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 0.00043057692307692306,
+      "loss": 0.6433,
+      "step": 1947
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 0.00043052884615384616,
+      "loss": 0.7198,
+      "step": 1948
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 0.0004304807692307692,
+      "loss": 0.6318,
+      "step": 1949
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 0.0004304326923076923,
+      "loss": 0.7408,
+      "step": 1950
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 0.00043038461538461535,
+      "loss": 0.562,
+      "step": 1951
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 0.00043033653846153845,
+      "loss": 0.6407,
+      "step": 1952
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 0.00043028846153846155,
+      "loss": 0.6255,
+      "step": 1953
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 0.0004302403846153846,
+      "loss": 0.6711,
+      "step": 1954
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 0.0004301923076923077,
+      "loss": 0.6463,
+      "step": 1955
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 0.00043014423076923074,
+      "loss": 0.5802,
+      "step": 1956
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 0.00043009615384615384,
+      "loss": 0.7142,
+      "step": 1957
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 0.00043004807692307694,
+      "loss": 0.6249,
+      "step": 1958
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 0.00043,
+      "loss": 0.686,
+      "step": 1959
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 0.0004299519230769231,
+      "loss": 0.7025,
+      "step": 1960
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 0.00042990384615384613,
+      "loss": 0.7281,
+      "step": 1961
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 0.00042985576923076923,
+      "loss": 0.6354,
+      "step": 1962
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 0.00042980769230769233,
+      "loss": 0.7903,
+      "step": 1963
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 0.0004297596153846154,
+      "loss": 0.7327,
+      "step": 1964
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 0.0004297115384615385,
+      "loss": 0.6209,
+      "step": 1965
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 0.0004296634615384615,
+      "loss": 0.6262,
+      "step": 1966
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 0.0004296153846153846,
+      "loss": 0.6857,
+      "step": 1967
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 0.00042956730769230767,
+      "loss": 0.6021,
+      "step": 1968
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 0.00042951923076923077,
+      "loss": 0.6389,
+      "step": 1969
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 0.00042947115384615387,
+      "loss": 0.7415,
+      "step": 1970
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 0.0004294230769230769,
+      "loss": 0.5873,
+      "step": 1971
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 0.000429375,
+      "loss": 0.6858,
+      "step": 1972
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 0.00042932692307692306,
+      "loss": 0.7534,
+      "step": 1973
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 0.00042927884615384616,
+      "loss": 0.5709,
+      "step": 1974
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 0.00042923076923076926,
+      "loss": 0.643,
+      "step": 1975
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 0.0004291826923076923,
+      "loss": 0.6656,
+      "step": 1976
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 0.0004291346153846154,
+      "loss": 0.7137,
+      "step": 1977
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 0.00042908653846153845,
+      "loss": 0.6868,
+      "step": 1978
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 0.00042903846153846155,
+      "loss": 0.5324,
+      "step": 1979
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 0.00042899038461538465,
+      "loss": 0.5804,
+      "step": 1980
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 0.0004289423076923077,
+      "loss": 0.6557,
+      "step": 1981
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 0.0004288942307692308,
+      "loss": 0.6404,
+      "step": 1982
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 0.00042884615384615384,
+      "loss": 0.6665,
+      "step": 1983
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 0.00042879807692307694,
+      "loss": 0.626,
+      "step": 1984
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 0.00042875000000000004,
+      "loss": 0.7288,
+      "step": 1985
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 0.0004287019230769231,
+      "loss": 0.6424,
+      "step": 1986
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 0.0004286538461538462,
+      "loss": 0.6113,
+      "step": 1987
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 0.00042860576923076923,
+      "loss": 0.6198,
+      "step": 1988
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 0.00042855769230769233,
+      "loss": 0.5726,
+      "step": 1989
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 0.00042850961538461543,
+      "loss": 0.821,
+      "step": 1990
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 0.0004284615384615385,
+      "loss": 0.6638,
+      "step": 1991
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 0.0004284134615384616,
+      "loss": 0.7864,
+      "step": 1992
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 0.0004283653846153846,
+      "loss": 0.5934,
+      "step": 1993
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 0.0004283173076923077,
+      "loss": 0.6825,
+      "step": 1994
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 0.0004282692307692308,
+      "loss": 0.5285,
+      "step": 1995
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 0.00042822115384615387,
+      "loss": 0.7485,
+      "step": 1996
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 0.00042817307692307697,
+      "loss": 0.6285,
+      "step": 1997
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 0.000428125,
+      "loss": 0.535,
+      "step": 1998
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 0.0004280769230769231,
+      "loss": 0.6646,
+      "step": 1999
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 0.0004280288461538462,
+      "loss": 0.621,
+      "step": 2000
+    },
+    {
+      "epoch": 18.35,
+      "eval_cer": 0.1874091065833071,
+      "eval_loss": 0.6824229955673218,
+      "eval_runtime": 87.0196,
+      "eval_samples_per_second": 18.927,
+      "eval_steps_per_second": 2.367,
+      "eval_wer": 0.6208763149831478,
+      "step": 2000
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 0.0004279807692307692,
+      "loss": 0.5749,
+      "step": 2001
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 0.0004279326923076923,
+      "loss": 0.6325,
+      "step": 2002
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 0.00042788461538461535,
+      "loss": 0.6195,
+      "step": 2003
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 0.00042783653846153845,
+      "loss": 0.7133,
+      "step": 2004
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 0.00042778846153846155,
+      "loss": 0.7276,
+      "step": 2005
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 0.0004277403846153846,
+      "loss": 0.5721,
+      "step": 2006
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 0.0004276923076923077,
+      "loss": 0.7687,
+      "step": 2007
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 0.00042764423076923074,
+      "loss": 0.6296,
+      "step": 2008
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 0.00042759615384615384,
+      "loss": 0.6557,
+      "step": 2009
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 0.00042754807692307694,
+      "loss": 0.5959,
+      "step": 2010
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 0.0004275,
+      "loss": 0.5957,
+      "step": 2011
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 0.0004274519230769231,
+      "loss": 0.4781,
+      "step": 2012
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 0.00042740384615384613,
+      "loss": 0.6061,
+      "step": 2013
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 0.00042735576923076923,
+      "loss": 0.6583,
+      "step": 2014
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 0.00042730769230769233,
+      "loss": 0.6876,
+      "step": 2015
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 0.0004272596153846154,
+      "loss": 0.531,
+      "step": 2016
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 0.00042721153846153847,
+      "loss": 0.8049,
+      "step": 2017
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 0.0004271634615384615,
+      "loss": 0.797,
+      "step": 2018
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 0.0004271153846153846,
+      "loss": 0.6958,
+      "step": 2019
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 0.0004270673076923077,
+      "loss": 0.679,
+      "step": 2020
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 0.00042701923076923076,
+      "loss": 0.8486,
+      "step": 2021
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 0.00042697115384615386,
+      "loss": 0.657,
+      "step": 2022
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 0.0004269230769230769,
+      "loss": 0.7311,
+      "step": 2023
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 0.000426875,
+      "loss": 0.7322,
+      "step": 2024
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 0.0004268269230769231,
+      "loss": 0.7014,
+      "step": 2025
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 0.00042677884615384615,
+      "loss": 0.6058,
+      "step": 2026
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 0.00042673076923076925,
+      "loss": 0.6903,
+      "step": 2027
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 0.0004266826923076923,
+      "loss": 0.6891,
+      "step": 2028
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 0.0004266346153846154,
+      "loss": 0.7323,
+      "step": 2029
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 0.0004265865384615385,
+      "loss": 0.6732,
+      "step": 2030
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 0.00042653846153846154,
+      "loss": 0.6518,
+      "step": 2031
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 0.00042649038461538464,
+      "loss": 0.708,
+      "step": 2032
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 0.0004264423076923077,
+      "loss": 0.6028,
+      "step": 2033
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 0.0004263942307692308,
+      "loss": 0.5772,
+      "step": 2034
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 0.00042634615384615383,
+      "loss": 0.6944,
+      "step": 2035
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 0.00042629807692307693,
+      "loss": 0.6348,
+      "step": 2036
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 0.00042625000000000003,
+      "loss": 0.6641,
+      "step": 2037
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 0.0004262019230769231,
+      "loss": 0.6627,
+      "step": 2038
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 0.0004261538461538462,
+      "loss": 0.6641,
+      "step": 2039
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 0.0004261057692307692,
+      "loss": 0.72,
+      "step": 2040
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 0.0004260576923076923,
+      "loss": 0.6281,
+      "step": 2041
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 0.0004260096153846154,
+      "loss": 0.6512,
+      "step": 2042
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 0.00042596153846153847,
+      "loss": 0.7382,
+      "step": 2043
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 0.00042591346153846157,
+      "loss": 0.871,
+      "step": 2044
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 0.0004258653846153846,
+      "loss": 0.7857,
+      "step": 2045
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 0.0004258173076923077,
+      "loss": 0.7645,
+      "step": 2046
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 0.0004257692307692308,
+      "loss": 0.758,
+      "step": 2047
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 0.00042572115384615386,
+      "loss": 0.7655,
+      "step": 2048
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 0.00042567307692307696,
+      "loss": 0.6972,
+      "step": 2049
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 0.000425625,
+      "loss": 0.742,
+      "step": 2050
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 0.0004255769230769231,
+      "loss": 0.6374,
+      "step": 2051
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 0.0004255288461538462,
+      "loss": 0.6226,
+      "step": 2052
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 0.00042548076923076925,
+      "loss": 0.6478,
+      "step": 2053
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 0.00042543269230769235,
+      "loss": 0.6152,
+      "step": 2054
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 0.0004253846153846154,
+      "loss": 0.668,
+      "step": 2055
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 0.0004253365384615385,
+      "loss": 0.5831,
+      "step": 2056
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 0.0004252884615384616,
+      "loss": 0.5807,
+      "step": 2057
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 0.0004252403846153846,
+      "loss": 0.8148,
+      "step": 2058
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 0.0004251923076923077,
+      "loss": 0.7586,
+      "step": 2059
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 0.00042514423076923073,
+      "loss": 0.629,
+      "step": 2060
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 0.00042509615384615383,
+      "loss": 0.6339,
+      "step": 2061
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 0.00042504807692307693,
+      "loss": 0.6726,
+      "step": 2062
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 0.000425,
+      "loss": 0.6759,
+      "step": 2063
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 0.0004249519230769231,
+      "loss": 0.6305,
+      "step": 2064
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 0.0004249038461538461,
+      "loss": 0.5393,
+      "step": 2065
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 0.0004248557692307692,
+      "loss": 0.7955,
+      "step": 2066
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 0.0004248076923076923,
+      "loss": 0.6969,
+      "step": 2067
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 0.00042475961538461537,
+      "loss": 0.8068,
+      "step": 2068
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 0.00042471153846153847,
+      "loss": 0.7292,
+      "step": 2069
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 0.0004246634615384615,
+      "loss": 0.5819,
+      "step": 2070
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 0.0004246153846153846,
+      "loss": 0.6712,
+      "step": 2071
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 0.0004245673076923077,
+      "loss": 0.8376,
+      "step": 2072
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 0.00042451923076923076,
+      "loss": 0.6711,
+      "step": 2073
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 0.00042447115384615386,
+      "loss": 0.7918,
+      "step": 2074
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 0.0004244230769230769,
+      "loss": 0.6272,
+      "step": 2075
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 0.000424375,
+      "loss": 0.547,
+      "step": 2076
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 0.0004243269230769231,
+      "loss": 0.5965,
+      "step": 2077
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 0.00042427884615384615,
+      "loss": 0.5609,
+      "step": 2078
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 0.00042423076923076925,
+      "loss": 0.7156,
+      "step": 2079
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 0.0004241826923076923,
+      "loss": 0.541,
+      "step": 2080
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 0.0004241346153846154,
+      "loss": 0.5378,
+      "step": 2081
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 0.0004240865384615385,
+      "loss": 0.6731,
+      "step": 2082
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 0.00042403846153846154,
+      "loss": 0.569,
+      "step": 2083
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 0.00042399038461538464,
+      "loss": 0.5746,
+      "step": 2084
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 0.0004239423076923077,
+      "loss": 0.6421,
+      "step": 2085
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 0.0004238942307692308,
+      "loss": 0.7285,
+      "step": 2086
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 0.0004238461538461539,
+      "loss": 0.5346,
+      "step": 2087
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 0.00042379807692307693,
+      "loss": 0.6284,
+      "step": 2088
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 0.00042375000000000003,
+      "loss": 0.6541,
+      "step": 2089
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 0.00042370192307692307,
+      "loss": 0.5731,
+      "step": 2090
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 0.00042365384615384617,
+      "loss": 0.6029,
+      "step": 2091
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 0.00042360576923076927,
+      "loss": 0.6231,
+      "step": 2092
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 0.0004235576923076923,
+      "loss": 0.5881,
+      "step": 2093
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 0.0004235096153846154,
+      "loss": 0.4753,
+      "step": 2094
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 0.00042346153846153846,
+      "loss": 0.604,
+      "step": 2095
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 0.00042341346153846156,
+      "loss": 0.6302,
+      "step": 2096
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 0.0004233653846153846,
+      "loss": 0.693,
+      "step": 2097
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 0.0004233173076923077,
+      "loss": 0.7329,
+      "step": 2098
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 0.0004232692307692308,
+      "loss": 0.6609,
+      "step": 2099
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 0.00042322115384615385,
+      "loss": 0.7092,
+      "step": 2100
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 0.00042317307692307695,
+      "loss": 0.666,
+      "step": 2101
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 0.000423125,
+      "loss": 0.7067,
+      "step": 2102
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 0.0004230769230769231,
+      "loss": 0.6597,
+      "step": 2103
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 0.0004230288461538462,
+      "loss": 0.6479,
+      "step": 2104
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 0.00042298076923076924,
+      "loss": 0.6342,
+      "step": 2105
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 0.00042293269230769234,
+      "loss": 0.6029,
+      "step": 2106
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 0.0004228846153846154,
+      "loss": 0.5958,
+      "step": 2107
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 0.0004228365384615385,
+      "loss": 0.5375,
+      "step": 2108
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 0.0004227884615384616,
+      "loss": 0.5542,
+      "step": 2109
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 0.00042274038461538463,
+      "loss": 0.635,
+      "step": 2110
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 0.00042269230769230773,
+      "loss": 0.5836,
+      "step": 2111
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 0.0004226442307692308,
+      "loss": 0.6728,
+      "step": 2112
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 0.0004225961538461539,
+      "loss": 0.5204,
+      "step": 2113
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 0.000422548076923077,
+      "loss": 0.7352,
+      "step": 2114
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 0.00042249999999999997,
+      "loss": 0.5756,
+      "step": 2115
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 0.00042245192307692307,
+      "loss": 0.5574,
+      "step": 2116
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 0.0004224038461538461,
+      "loss": 0.6595,
+      "step": 2117
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 0.0004223557692307692,
+      "loss": 0.621,
+      "step": 2118
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 0.0004223076923076923,
+      "loss": 0.6162,
+      "step": 2119
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 0.00042225961538461536,
+      "loss": 0.624,
+      "step": 2120
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 0.00042221153846153846,
+      "loss": 0.5913,
+      "step": 2121
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 0.0004221634615384615,
+      "loss": 0.6874,
+      "step": 2122
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 0.0004221153846153846,
+      "loss": 0.5855,
+      "step": 2123
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 0.0004220673076923077,
+      "loss": 0.6807,
+      "step": 2124
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 0.00042201923076923075,
+      "loss": 0.6597,
+      "step": 2125
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 0.00042197115384615385,
+      "loss": 0.7181,
+      "step": 2126
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 0.0004219230769230769,
+      "loss": 0.6921,
+      "step": 2127
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 0.000421875,
+      "loss": 0.6543,
+      "step": 2128
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 0.0004218269230769231,
+      "loss": 0.6778,
+      "step": 2129
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 0.00042177884615384614,
+      "loss": 0.7526,
+      "step": 2130
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 0.00042173076923076924,
+      "loss": 0.657,
+      "step": 2131
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 0.0004216826923076923,
+      "loss": 0.6276,
+      "step": 2132
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 0.0004216346153846154,
+      "loss": 0.6847,
+      "step": 2133
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 0.0004215865384615385,
+      "loss": 0.6685,
+      "step": 2134
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 0.00042153846153846153,
+      "loss": 0.556,
+      "step": 2135
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 0.00042149038461538463,
+      "loss": 0.764,
+      "step": 2136
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 0.0004214423076923077,
+      "loss": 0.649,
+      "step": 2137
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 0.0004213942307692308,
+      "loss": 0.6968,
+      "step": 2138
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 0.0004213461538461539,
+      "loss": 0.7208,
+      "step": 2139
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 0.0004212980769230769,
+      "loss": 0.5623,
+      "step": 2140
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 0.00042125,
+      "loss": 0.637,
+      "step": 2141
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 0.00042120192307692307,
+      "loss": 0.6035,
+      "step": 2142
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 0.00042115384615384617,
+      "loss": 0.66,
+      "step": 2143
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 0.00042110576923076927,
+      "loss": 0.7416,
+      "step": 2144
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 0.0004210576923076923,
+      "loss": 0.6021,
+      "step": 2145
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 0.0004210096153846154,
+      "loss": 0.579,
+      "step": 2146
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 0.00042096153846153846,
+      "loss": 0.7478,
+      "step": 2147
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 0.00042091346153846156,
+      "loss": 0.7311,
+      "step": 2148
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 0.00042086538461538466,
+      "loss": 0.5782,
+      "step": 2149
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 0.0004208173076923077,
+      "loss": 0.6014,
+      "step": 2150
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 0.0004207692307692308,
+      "loss": 0.5911,
+      "step": 2151
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 0.00042072115384615385,
+      "loss": 0.5621,
+      "step": 2152
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 0.00042067307692307695,
+      "loss": 0.8325,
+      "step": 2153
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 0.00042062500000000005,
+      "loss": 0.7552,
+      "step": 2154
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 0.0004205769230769231,
+      "loss": 0.7166,
+      "step": 2155
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 0.0004205288461538462,
+      "loss": 0.6504,
+      "step": 2156
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 0.00042048076923076924,
+      "loss": 0.6885,
+      "step": 2157
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 0.00042043269230769234,
+      "loss": 0.6515,
+      "step": 2158
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 0.00042038461538461544,
+      "loss": 0.7167,
+      "step": 2159
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 0.0004203365384615385,
+      "loss": 0.6275,
+      "step": 2160
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 0.0004202884615384616,
+      "loss": 0.5982,
+      "step": 2161
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 0.00042024038461538463,
+      "loss": 0.6507,
+      "step": 2162
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 0.0004201923076923077,
+      "loss": 0.6157,
+      "step": 2163
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 0.00042014423076923077,
+      "loss": 0.6713,
+      "step": 2164
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 0.00042009615384615387,
+      "loss": 0.5452,
+      "step": 2165
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 0.00042004807692307697,
+      "loss": 0.4895,
+      "step": 2166
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 0.00042,
+      "loss": 0.7759,
+      "step": 2167
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 0.0004199519230769231,
+      "loss": 0.6635,
+      "step": 2168
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 0.00041990384615384616,
+      "loss": 0.5963,
+      "step": 2169
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 0.00041985576923076926,
+      "loss": 0.5653,
+      "step": 2170
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 0.0004198076923076923,
+      "loss": 0.6725,
+      "step": 2171
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 0.00041975961538461535,
+      "loss": 0.7042,
+      "step": 2172
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 0.00041971153846153845,
+      "loss": 0.6146,
+      "step": 2173
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 0.0004196634615384615,
+      "loss": 0.6416,
+      "step": 2174
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 0.0004196153846153846,
+      "loss": 0.4647,
+      "step": 2175
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 0.0004195673076923077,
+      "loss": 0.6841,
+      "step": 2176
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 0.00041951923076923074,
+      "loss": 0.4873,
+      "step": 2177
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 0.00041947115384615384,
+      "loss": 0.6677,
+      "step": 2178
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 0.0004194230769230769,
+      "loss": 0.6507,
+      "step": 2179
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.000419375,
+      "loss": 0.6285,
+      "step": 2180
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 0.0004193269230769231,
+      "loss": 0.7279,
+      "step": 2181
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 0.00041927884615384613,
+      "loss": 0.6738,
+      "step": 2182
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 0.00041923076923076923,
+      "loss": 0.7254,
+      "step": 2183
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 0.0004191826923076923,
+      "loss": 0.5804,
+      "step": 2184
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 0.0004191346153846154,
+      "loss": 0.6516,
+      "step": 2185
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 0.0004190865384615385,
+      "loss": 0.5369,
+      "step": 2186
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 0.0004190384615384615,
+      "loss": 0.73,
+      "step": 2187
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 0.0004189903846153846,
+      "loss": 0.6171,
+      "step": 2188
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 0.00041894230769230767,
+      "loss": 0.6715,
+      "step": 2189
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 0.00041889423076923077,
+      "loss": 0.6207,
+      "step": 2190
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 0.00041884615384615387,
+      "loss": 0.6708,
+      "step": 2191
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 0.0004187980769230769,
+      "loss": 0.5705,
+      "step": 2192
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 0.00041875,
+      "loss": 0.5582,
+      "step": 2193
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 0.00041870192307692306,
+      "loss": 0.5565,
+      "step": 2194
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 0.00041865384615384616,
+      "loss": 0.5745,
+      "step": 2195
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 0.00041860576923076926,
+      "loss": 0.7841,
+      "step": 2196
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 0.0004185576923076923,
+      "loss": 0.6246,
+      "step": 2197
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 0.0004185096153846154,
+      "loss": 0.5502,
+      "step": 2198
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 0.00041846153846153845,
+      "loss": 0.7065,
+      "step": 2199
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 0.00041841346153846155,
+      "loss": 0.7133,
+      "step": 2200
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 0.00041836538461538465,
+      "loss": 0.6778,
+      "step": 2201
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 0.0004183173076923077,
+      "loss": 0.6987,
+      "step": 2202
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 0.0004182692307692308,
+      "loss": 0.6237,
+      "step": 2203
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 0.00041822115384615384,
+      "loss": 0.5484,
+      "step": 2204
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 0.00041817307692307694,
+      "loss": 0.7789,
+      "step": 2205
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 0.00041812500000000004,
+      "loss": 0.7521,
+      "step": 2206
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 0.0004180769230769231,
+      "loss": 0.6603,
+      "step": 2207
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 0.0004180288461538462,
+      "loss": 0.7348,
+      "step": 2208
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 0.00041798076923076923,
+      "loss": 0.7676,
+      "step": 2209
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 0.00041793269230769233,
+      "loss": 0.6425,
+      "step": 2210
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 0.00041788461538461543,
+      "loss": 0.6639,
+      "step": 2211
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 0.0004178365384615385,
+      "loss": 0.6234,
+      "step": 2212
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 0.0004177884615384616,
+      "loss": 0.6327,
+      "step": 2213
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 0.0004177403846153846,
+      "loss": 0.6304,
+      "step": 2214
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 0.0004176923076923077,
+      "loss": 0.6015,
+      "step": 2215
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 0.0004176442307692308,
+      "loss": 0.5731,
+      "step": 2216
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 0.00041759615384615387,
+      "loss": 0.5698,
+      "step": 2217
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 0.00041754807692307697,
+      "loss": 0.5303,
+      "step": 2218
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 0.0004175,
+      "loss": 0.6087,
+      "step": 2219
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 0.0004174519230769231,
+      "loss": 0.5741,
+      "step": 2220
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 0.0004174038461538462,
+      "loss": 0.6861,
+      "step": 2221
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 0.00041735576923076926,
+      "loss": 0.6262,
+      "step": 2222
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 0.00041730769230769236,
+      "loss": 0.5626,
+      "step": 2223
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 0.0004172596153846154,
+      "loss": 0.5854,
+      "step": 2224
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 0.0004172115384615385,
+      "loss": 0.6047,
+      "step": 2225
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 0.00041716346153846155,
+      "loss": 0.701,
+      "step": 2226
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 0.0004171153846153846,
+      "loss": 0.6063,
+      "step": 2227
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 0.0004170673076923077,
+      "loss": 0.57,
+      "step": 2228
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 0.00041701923076923074,
+      "loss": 0.6091,
+      "step": 2229
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 0.00041697115384615384,
+      "loss": 0.6498,
+      "step": 2230
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 0.0004169230769230769,
+      "loss": 0.5957,
+      "step": 2231
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 0.000416875,
+      "loss": 0.6192,
+      "step": 2232
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 0.0004168269230769231,
+      "loss": 0.604,
+      "step": 2233
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 0.00041677884615384613,
+      "loss": 0.5713,
+      "step": 2234
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 0.0004167307692307692,
+      "loss": 0.7389,
+      "step": 2235
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 0.00041668269230769227,
+      "loss": 0.6842,
+      "step": 2236
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 0.00041663461538461537,
+      "loss": 0.6984,
+      "step": 2237
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 0.00041658653846153847,
+      "loss": 0.6635,
+      "step": 2238
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 0.0004165384615384615,
+      "loss": 0.7024,
+      "step": 2239
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 0.0004164903846153846,
+      "loss": 0.6437,
+      "step": 2240
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 0.00041644230769230766,
+      "loss": 0.5787,
+      "step": 2241
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 0.00041639423076923076,
+      "loss": 0.5363,
+      "step": 2242
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 0.00041634615384615386,
+      "loss": 0.542,
+      "step": 2243
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 0.0004162980769230769,
+      "loss": 0.5545,
+      "step": 2244
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 0.00041625,
+      "loss": 0.5844,
+      "step": 2245
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 0.00041620192307692305,
+      "loss": 0.5979,
+      "step": 2246
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 0.00041615384615384615,
+      "loss": 0.6315,
+      "step": 2247
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 0.00041610576923076925,
+      "loss": 0.6543,
+      "step": 2248
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 0.0004160576923076923,
+      "loss": 0.5437,
+      "step": 2249
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 0.0004160096153846154,
+      "loss": 0.6237,
+      "step": 2250
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 0.00041596153846153844,
+      "loss": 0.6116,
+      "step": 2251
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 0.00041591346153846154,
+      "loss": 0.5608,
+      "step": 2252
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 0.00041586538461538464,
+      "loss": 0.5821,
+      "step": 2253
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 0.0004158173076923077,
+      "loss": 0.5265,
+      "step": 2254
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 0.0004157692307692308,
+      "loss": 0.581,
+      "step": 2255
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 0.00041572115384615383,
+      "loss": 0.6659,
+      "step": 2256
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 0.00041567307692307693,
+      "loss": 0.6021,
+      "step": 2257
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 0.00041562500000000003,
+      "loss": 0.5863,
+      "step": 2258
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 0.0004155769230769231,
+      "loss": 0.5849,
+      "step": 2259
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 0.0004155288461538462,
+      "loss": 0.7164,
+      "step": 2260
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 0.0004154807692307692,
+      "loss": 0.5691,
+      "step": 2261
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 0.0004154326923076923,
+      "loss": 0.7955,
+      "step": 2262
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 0.0004153846153846154,
+      "loss": 0.6719,
+      "step": 2263
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 0.00041533653846153847,
+      "loss": 0.7393,
+      "step": 2264
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 0.00041528846153846157,
+      "loss": 0.6939,
+      "step": 2265
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 0.0004152403846153846,
+      "loss": 0.6586,
+      "step": 2266
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 0.0004151923076923077,
+      "loss": 0.6444,
+      "step": 2267
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 0.0004151442307692308,
+      "loss": 0.7176,
+      "step": 2268
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 0.00041509615384615386,
+      "loss": 0.5954,
+      "step": 2269
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 0.00041504807692307696,
+      "loss": 0.6128,
+      "step": 2270
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 0.000415,
+      "loss": 0.6686,
+      "step": 2271
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 0.0004149519230769231,
+      "loss": 0.7924,
+      "step": 2272
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 0.0004149038461538462,
+      "loss": 0.6358,
+      "step": 2273
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 0.00041485576923076925,
+      "loss": 0.5613,
+      "step": 2274
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 0.00041480769230769235,
+      "loss": 0.6457,
+      "step": 2275
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 0.0004147596153846154,
+      "loss": 0.6766,
+      "step": 2276
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 0.0004147115384615385,
+      "loss": 0.5547,
+      "step": 2277
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 0.0004146634615384616,
+      "loss": 0.6563,
+      "step": 2278
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 0.00041461538461538464,
+      "loss": 0.625,
+      "step": 2279
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 0.00041456730769230774,
+      "loss": 0.7792,
+      "step": 2280
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 0.0004145192307692308,
+      "loss": 0.5927,
+      "step": 2281
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 0.0004144711538461539,
+      "loss": 0.644,
+      "step": 2282
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 0.000414423076923077,
+      "loss": 0.4905,
+      "step": 2283
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 0.000414375,
+      "loss": 0.5691,
+      "step": 2284
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 0.0004143269230769231,
+      "loss": 0.6916,
+      "step": 2285
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 0.0004142788461538461,
+      "loss": 0.72,
+      "step": 2286
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 0.0004142307692307692,
+      "loss": 0.5728,
+      "step": 2287
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 0.0004141826923076923,
+      "loss": 0.6409,
+      "step": 2288
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 0.00041413461538461537,
+      "loss": 0.8334,
+      "step": 2289
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 0.00041408653846153847,
+      "loss": 0.7647,
+      "step": 2290
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 0.0004140384615384615,
+      "loss": 0.617,
+      "step": 2291
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 0.0004139903846153846,
+      "loss": 0.7906,
+      "step": 2292
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 0.00041394230769230766,
+      "loss": 0.6039,
+      "step": 2293
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 0.00041389423076923076,
+      "loss": 0.7042,
+      "step": 2294
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 0.00041384615384615386,
+      "loss": 0.6389,
+      "step": 2295
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 0.0004137980769230769,
+      "loss": 0.7,
+      "step": 2296
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 0.00041375,
+      "loss": 0.635,
+      "step": 2297
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 0.00041370192307692305,
+      "loss": 0.6308,
+      "step": 2298
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 0.00041365384615384615,
+      "loss": 0.5347,
+      "step": 2299
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 0.00041360576923076925,
+      "loss": 0.4976,
+      "step": 2300
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 0.0004135576923076923,
+      "loss": 0.6606,
+      "step": 2301
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 0.0004135096153846154,
+      "loss": 0.5859,
+      "step": 2302
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 0.00041346153846153844,
+      "loss": 0.6253,
+      "step": 2303
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 0.00041341346153846154,
+      "loss": 0.6121,
+      "step": 2304
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 0.00041336538461538464,
+      "loss": 0.562,
+      "step": 2305
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 0.0004133173076923077,
+      "loss": 0.5924,
+      "step": 2306
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 0.0004132692307692308,
+      "loss": 0.5208,
+      "step": 2307
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 0.0004132211538461538,
+      "loss": 0.6549,
+      "step": 2308
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 0.0004131730769230769,
+      "loss": 0.6203,
+      "step": 2309
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 0.000413125,
+      "loss": 0.593,
+      "step": 2310
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 0.00041307692307692307,
+      "loss": 0.6859,
+      "step": 2311
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 0.00041302884615384617,
+      "loss": 0.6395,
+      "step": 2312
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 0.0004129807692307692,
+      "loss": 0.5732,
+      "step": 2313
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 0.0004129326923076923,
+      "loss": 0.6275,
+      "step": 2314
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 0.0004128846153846154,
+      "loss": 0.6857,
+      "step": 2315
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 0.00041283653846153846,
+      "loss": 0.4196,
+      "step": 2316
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 0.00041278846153846156,
+      "loss": 0.7965,
+      "step": 2317
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 0.0004127403846153846,
+      "loss": 0.7018,
+      "step": 2318
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 0.0004126923076923077,
+      "loss": 0.742,
+      "step": 2319
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 0.0004126442307692308,
+      "loss": 0.6722,
+      "step": 2320
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 0.00041259615384615385,
+      "loss": 0.7017,
+      "step": 2321
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 0.00041254807692307695,
+      "loss": 0.647,
+      "step": 2322
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 0.0004125,
+      "loss": 0.7778,
+      "step": 2323
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 0.0004124519230769231,
+      "loss": 0.5693,
+      "step": 2324
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 0.0004124038461538462,
+      "loss": 0.558,
+      "step": 2325
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 0.00041235576923076924,
+      "loss": 0.691,
+      "step": 2326
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 0.00041230769230769234,
+      "loss": 0.6975,
+      "step": 2327
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 0.0004122596153846154,
+      "loss": 0.5381,
+      "step": 2328
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 0.0004122115384615385,
+      "loss": 0.6315,
+      "step": 2329
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 0.0004121634615384616,
+      "loss": 0.6927,
+      "step": 2330
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 0.00041211538461538463,
+      "loss": 0.7531,
+      "step": 2331
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 0.00041206730769230773,
+      "loss": 0.6998,
+      "step": 2332
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 0.0004120192307692308,
+      "loss": 0.5822,
+      "step": 2333
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 0.0004119711538461539,
+      "loss": 0.6536,
+      "step": 2334
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 0.000411923076923077,
+      "loss": 0.6536,
+      "step": 2335
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 0.000411875,
+      "loss": 0.6032,
+      "step": 2336
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 0.0004118269230769231,
+      "loss": 0.7267,
+      "step": 2337
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 0.00041177884615384617,
+      "loss": 0.8107,
+      "step": 2338
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 0.00041173076923076927,
+      "loss": 0.6149,
+      "step": 2339
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 0.00041168269230769237,
+      "loss": 0.6447,
+      "step": 2340
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 0.00041163461538461536,
+      "loss": 0.553,
+      "step": 2341
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 0.00041158653846153846,
+      "loss": 0.6956,
+      "step": 2342
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 0.0004115384615384615,
+      "loss": 0.5076,
+      "step": 2343
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 0.0004114903846153846,
+      "loss": 0.7666,
+      "step": 2344
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 0.0004114423076923077,
+      "loss": 0.7988,
+      "step": 2345
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 0.00041139423076923075,
+      "loss": 0.7585,
+      "step": 2346
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 0.00041134615384615385,
+      "loss": 0.6245,
+      "step": 2347
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 0.0004112980769230769,
+      "loss": 0.6086,
+      "step": 2348
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 0.00041125,
+      "loss": 0.6861,
+      "step": 2349
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 0.0004112019230769231,
+      "loss": 0.6621,
+      "step": 2350
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 0.00041115384615384614,
+      "loss": 0.5115,
+      "step": 2351
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 0.00041110576923076924,
+      "loss": 0.6627,
+      "step": 2352
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 0.0004110576923076923,
+      "loss": 0.6201,
+      "step": 2353
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 0.0004110096153846154,
+      "loss": 0.7068,
+      "step": 2354
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 0.0004109615384615385,
+      "loss": 0.5368,
+      "step": 2355
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 0.00041091346153846153,
+      "loss": 0.6584,
+      "step": 2356
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 0.00041086538461538463,
+      "loss": 0.604,
+      "step": 2357
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 0.0004108173076923077,
+      "loss": 0.6977,
+      "step": 2358
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 0.0004107692307692308,
+      "loss": 0.574,
+      "step": 2359
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 0.0004107211538461538,
+      "loss": 0.6392,
+      "step": 2360
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 0.0004106730769230769,
+      "loss": 0.6708,
+      "step": 2361
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 0.000410625,
+      "loss": 0.6854,
+      "step": 2362
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 0.00041057692307692307,
+      "loss": 0.7262,
+      "step": 2363
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 0.00041052884615384617,
+      "loss": 0.7265,
+      "step": 2364
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 0.0004104807692307692,
+      "loss": 0.6212,
+      "step": 2365
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 0.0004104326923076923,
+      "loss": 0.5712,
+      "step": 2366
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 0.0004103846153846154,
+      "loss": 0.7014,
+      "step": 2367
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 0.00041033653846153846,
+      "loss": 0.6114,
+      "step": 2368
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 0.00041028846153846156,
+      "loss": 0.6119,
+      "step": 2369
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 0.0004102403846153846,
+      "loss": 0.7524,
+      "step": 2370
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 0.0004101923076923077,
+      "loss": 0.715,
+      "step": 2371
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 0.0004101442307692308,
+      "loss": 0.6893,
+      "step": 2372
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 0.00041009615384615385,
+      "loss": 0.6978,
+      "step": 2373
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 0.00041004807692307695,
+      "loss": 0.7265,
+      "step": 2374
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 0.00041,
+      "loss": 0.5048,
+      "step": 2375
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 0.0004099519230769231,
+      "loss": 0.637,
+      "step": 2376
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 0.0004099038461538462,
+      "loss": 0.6283,
+      "step": 2377
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 0.00040985576923076924,
+      "loss": 0.6233,
+      "step": 2378
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 0.00040980769230769234,
+      "loss": 0.6247,
+      "step": 2379
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 0.0004097596153846154,
+      "loss": 0.6593,
+      "step": 2380
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 0.0004097115384615385,
+      "loss": 0.8381,
+      "step": 2381
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 0.0004096634615384616,
+      "loss": 0.659,
+      "step": 2382
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 0.0004096153846153846,
+      "loss": 0.6238,
+      "step": 2383
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 0.0004095673076923077,
+      "loss": 0.64,
+      "step": 2384
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 0.00040951923076923077,
+      "loss": 0.6246,
+      "step": 2385
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 0.00040947115384615387,
+      "loss": 0.738,
+      "step": 2386
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 0.00040942307692307697,
+      "loss": 0.6711,
+      "step": 2387
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 0.000409375,
+      "loss": 0.5579,
+      "step": 2388
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 0.0004093269230769231,
+      "loss": 0.6555,
+      "step": 2389
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 0.00040927884615384616,
+      "loss": 0.6301,
+      "step": 2390
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 0.00040923076923076926,
+      "loss": 0.6384,
+      "step": 2391
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 0.00040918269230769236,
+      "loss": 0.648,
+      "step": 2392
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 0.0004091346153846154,
+      "loss": 0.5751,
+      "step": 2393
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 0.0004090865384615385,
+      "loss": 0.4822,
+      "step": 2394
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 0.00040903846153846155,
+      "loss": 0.5804,
+      "step": 2395
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 0.00040899038461538465,
+      "loss": 0.6963,
+      "step": 2396
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 0.00040894230769230775,
+      "loss": 0.5481,
+      "step": 2397
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 0.00040889423076923074,
+      "loss": 0.5925,
+      "step": 2398
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 0.00040884615384615384,
+      "loss": 0.6671,
+      "step": 2399
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 0.0004087980769230769,
+      "loss": 0.6049,
+      "step": 2400
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 0.00040875,
+      "loss": 0.563,
+      "step": 2401
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 0.0004087019230769231,
+      "loss": 0.4995,
+      "step": 2402
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 0.00040865384615384613,
+      "loss": 0.569,
+      "step": 2403
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 0.00040860576923076923,
+      "loss": 0.6242,
+      "step": 2404
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 0.0004085576923076923,
+      "loss": 0.5605,
+      "step": 2405
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 0.0004085096153846154,
+      "loss": 0.4483,
+      "step": 2406
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 0.0004084615384615385,
+      "loss": 0.5282,
+      "step": 2407
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 0.0004084134615384615,
+      "loss": 0.5141,
+      "step": 2408
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 0.0004083653846153846,
+      "loss": 0.61,
+      "step": 2409
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 0.00040831730769230767,
+      "loss": 0.5963,
+      "step": 2410
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 0.00040826923076923077,
+      "loss": 0.5246,
+      "step": 2411
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 0.00040822115384615387,
+      "loss": 0.5635,
+      "step": 2412
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 0.0004081730769230769,
+      "loss": 0.5906,
+      "step": 2413
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 0.000408125,
+      "loss": 0.6005,
+      "step": 2414
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 0.00040807692307692306,
+      "loss": 0.625,
+      "step": 2415
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 0.00040802884615384616,
+      "loss": 0.7521,
+      "step": 2416
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 0.00040798076923076926,
+      "loss": 0.618,
+      "step": 2417
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 0.0004079326923076923,
+      "loss": 0.6471,
+      "step": 2418
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 0.0004078846153846154,
+      "loss": 0.6801,
+      "step": 2419
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 0.00040783653846153845,
+      "loss": 0.4951,
+      "step": 2420
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 0.00040778846153846155,
+      "loss": 0.6661,
+      "step": 2421
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 0.0004077403846153846,
+      "loss": 0.5425,
+      "step": 2422
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 0.0004076923076923077,
+      "loss": 0.5784,
+      "step": 2423
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 0.0004076442307692308,
+      "loss": 0.5345,
+      "step": 2424
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 0.00040759615384615384,
+      "loss": 0.5325,
+      "step": 2425
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 0.00040754807692307694,
+      "loss": 0.715,
+      "step": 2426
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 0.0004075,
+      "loss": 0.7201,
+      "step": 2427
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 0.0004074519230769231,
+      "loss": 0.6508,
+      "step": 2428
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 0.0004074038461538462,
+      "loss": 0.7085,
+      "step": 2429
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 0.00040735576923076923,
+      "loss": 0.6457,
+      "step": 2430
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 0.00040730769230769233,
+      "loss": 0.6353,
+      "step": 2431
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 0.0004072596153846154,
+      "loss": 0.5534,
+      "step": 2432
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 0.0004072115384615385,
+      "loss": 0.495,
+      "step": 2433
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 0.0004071634615384616,
+      "loss": 0.6046,
+      "step": 2434
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 0.0004071153846153846,
+      "loss": 0.5637,
+      "step": 2435
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 0.0004070673076923077,
+      "loss": 0.692,
+      "step": 2436
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 0.00040701923076923077,
+      "loss": 0.603,
+      "step": 2437
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 0.00040697115384615386,
+      "loss": 0.6392,
+      "step": 2438
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 0.00040692307692307696,
+      "loss": 0.6084,
+      "step": 2439
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 0.000406875,
+      "loss": 0.6388,
+      "step": 2440
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 0.0004068269230769231,
+      "loss": 0.5063,
+      "step": 2441
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 0.00040677884615384616,
+      "loss": 0.5159,
+      "step": 2442
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 0.00040673076923076926,
+      "loss": 0.5419,
+      "step": 2443
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 0.00040668269230769235,
+      "loss": 0.5446,
+      "step": 2444
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 0.0004066346153846154,
+      "loss": 0.6283,
+      "step": 2445
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 0.0004065865384615385,
+      "loss": 0.6097,
+      "step": 2446
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 0.00040653846153846155,
+      "loss": 0.6378,
+      "step": 2447
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 0.00040649038461538465,
+      "loss": 0.5701,
+      "step": 2448
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 0.00040644230769230775,
+      "loss": 0.5574,
+      "step": 2449
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 0.0004063942307692308,
+      "loss": 0.5354,
+      "step": 2450
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 0.0004063461538461539,
+      "loss": 0.5473,
+      "step": 2451
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 0.00040629807692307694,
+      "loss": 0.7374,
+      "step": 2452
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 0.00040625000000000004,
+      "loss": 0.7412,
+      "step": 2453
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 0.0004062019230769231,
+      "loss": 0.7378,
+      "step": 2454
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 0.0004061538461538461,
+      "loss": 0.7237,
+      "step": 2455
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 0.0004061057692307692,
+      "loss": 0.7131,
+      "step": 2456
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 0.00040605769230769227,
+      "loss": 0.641,
+      "step": 2457
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 0.00040600961538461537,
+      "loss": 0.5829,
+      "step": 2458
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 0.00040596153846153847,
+      "loss": 0.7844,
+      "step": 2459
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 0.0004059134615384615,
+      "loss": 0.5166,
+      "step": 2460
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 0.0004058653846153846,
+      "loss": 0.6949,
+      "step": 2461
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 0.00040581730769230766,
+      "loss": 0.6187,
+      "step": 2462
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 0.00040576923076923076,
+      "loss": 0.5819,
+      "step": 2463
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 0.00040572115384615386,
+      "loss": 0.6026,
+      "step": 2464
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 0.0004056730769230769,
+      "loss": 0.6133,
+      "step": 2465
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 0.000405625,
+      "loss": 0.731,
+      "step": 2466
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 0.00040557692307692305,
+      "loss": 0.6801,
+      "step": 2467
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 0.00040552884615384615,
+      "loss": 0.6664,
+      "step": 2468
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 0.00040548076923076925,
+      "loss": 0.5825,
+      "step": 2469
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 0.0004054326923076923,
+      "loss": 0.5518,
+      "step": 2470
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 0.0004053846153846154,
+      "loss": 0.5989,
+      "step": 2471
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 0.00040533653846153844,
+      "loss": 0.5781,
+      "step": 2472
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 0.00040528846153846154,
+      "loss": 0.6229,
+      "step": 2473
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 0.00040524038461538464,
+      "loss": 0.5047,
+      "step": 2474
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 0.0004051923076923077,
+      "loss": 0.5928,
+      "step": 2475
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 0.0004051442307692308,
+      "loss": 0.5742,
+      "step": 2476
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 0.00040509615384615383,
+      "loss": 0.6139,
+      "step": 2477
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 0.00040504807692307693,
+      "loss": 0.5844,
+      "step": 2478
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 0.00040500000000000003,
+      "loss": 0.6983,
+      "step": 2479
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 0.0004049519230769231,
+      "loss": 0.757,
+      "step": 2480
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 0.0004049038461538462,
+      "loss": 0.5849,
+      "step": 2481
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 0.0004048557692307692,
+      "loss": 0.6595,
+      "step": 2482
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 0.0004048076923076923,
+      "loss": 0.6374,
+      "step": 2483
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 0.0004047596153846154,
+      "loss": 0.6128,
+      "step": 2484
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 0.00040471153846153847,
+      "loss": 0.6461,
+      "step": 2485
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 0.00040466346153846157,
+      "loss": 0.5459,
+      "step": 2486
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 0.0004046153846153846,
+      "loss": 0.5233,
+      "step": 2487
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 0.0004045673076923077,
+      "loss": 0.6274,
+      "step": 2488
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 0.00040451923076923076,
+      "loss": 0.5767,
+      "step": 2489
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 0.00040447115384615386,
+      "loss": 0.4906,
+      "step": 2490
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 0.00040442307692307696,
+      "loss": 0.5396,
+      "step": 2491
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 0.000404375,
+      "loss": 0.5644,
+      "step": 2492
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 0.0004043269230769231,
+      "loss": 0.5352,
+      "step": 2493
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 0.00040427884615384615,
+      "loss": 0.5256,
+      "step": 2494
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 0.00040423076923076925,
+      "loss": 0.6134,
+      "step": 2495
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 0.00040418269230769235,
+      "loss": 0.5666,
+      "step": 2496
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 0.0004041346153846154,
+      "loss": 0.6097,
+      "step": 2497
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 0.0004040865384615385,
+      "loss": 0.542,
+      "step": 2498
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 0.00040403846153846154,
+      "loss": 0.7909,
+      "step": 2499
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 0.00040399038461538464,
+      "loss": 0.6362,
+      "step": 2500
+    },
+    {
+      "epoch": 22.94,
+      "eval_cer": 0.1903776668315867,
+      "eval_loss": 0.6711505651473999,
+      "eval_runtime": 87.2908,
+      "eval_samples_per_second": 18.868,
+      "eval_steps_per_second": 2.36,
+      "eval_wer": 0.6286385456031048,
+      "step": 2500
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 0.00040394230769230774,
+      "loss": 0.6435,
+      "step": 2501
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 0.0004038942307692308,
+      "loss": 0.5815,
+      "step": 2502
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 0.0004038461538461539,
+      "loss": 0.6485,
+      "step": 2503
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 0.00040379807692307693,
+      "loss": 0.5987,
+      "step": 2504
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 0.00040375000000000003,
+      "loss": 0.6681,
+      "step": 2505
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 0.00040370192307692313,
+      "loss": 0.563,
+      "step": 2506
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 0.0004036538461538462,
+      "loss": 0.6973,
+      "step": 2507
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 0.0004036057692307693,
+      "loss": 0.6924,
+      "step": 2508
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 0.00040355769230769227,
+      "loss": 0.6864,
+      "step": 2509
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 0.00040350961538461537,
+      "loss": 0.7023,
+      "step": 2510
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 0.00040346153846153846,
+      "loss": 0.6094,
+      "step": 2511
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 0.0004034134615384615,
+      "loss": 0.5644,
+      "step": 2512
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 0.0004033653846153846,
+      "loss": 0.5265,
+      "step": 2513
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 0.00040331730769230766,
+      "loss": 0.7186,
+      "step": 2514
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 0.00040326923076923076,
+      "loss": 0.5284,
+      "step": 2515
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 0.00040322115384615386,
+      "loss": 0.6476,
+      "step": 2516
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 0.0004031730769230769,
+      "loss": 0.7136,
+      "step": 2517
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 0.000403125,
+      "loss": 0.5536,
+      "step": 2518
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 0.00040307692307692305,
+      "loss": 0.548,
+      "step": 2519
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 0.00040302884615384615,
+      "loss": 0.4677,
+      "step": 2520
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 0.00040298076923076925,
+      "loss": 0.6041,
+      "step": 2521
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 0.0004029326923076923,
+      "loss": 0.6434,
+      "step": 2522
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 0.0004028846153846154,
+      "loss": 0.5136,
+      "step": 2523
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 0.00040283653846153844,
+      "loss": 0.5375,
+      "step": 2524
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 0.00040278846153846154,
+      "loss": 0.6322,
+      "step": 2525
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 0.00040274038461538464,
+      "loss": 0.6418,
+      "step": 2526
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 0.0004026923076923077,
+      "loss": 0.533,
+      "step": 2527
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 0.0004026442307692308,
+      "loss": 0.4749,
+      "step": 2528
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 0.0004025961538461538,
+      "loss": 0.6237,
+      "step": 2529
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 0.0004025480769230769,
+      "loss": 0.6001,
+      "step": 2530
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 0.0004025,
+      "loss": 0.6885,
+      "step": 2531
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 0.00040245192307692307,
+      "loss": 0.5099,
+      "step": 2532
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 0.00040240384615384617,
+      "loss": 0.5789,
+      "step": 2533
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 0.0004023557692307692,
+      "loss": 0.4927,
+      "step": 2534
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 0.0004023076923076923,
+      "loss": 0.6659,
+      "step": 2535
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 0.0004022596153846154,
+      "loss": 0.6273,
+      "step": 2536
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 0.00040221153846153846,
+      "loss": 0.6628,
+      "step": 2537
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 0.00040216346153846156,
+      "loss": 0.5448,
+      "step": 2538
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 0.0004021153846153846,
+      "loss": 0.5634,
+      "step": 2539
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 0.0004020673076923077,
+      "loss": 0.6985,
+      "step": 2540
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 0.0004020192307692308,
+      "loss": 0.5617,
+      "step": 2541
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 0.00040197115384615385,
+      "loss": 0.5918,
+      "step": 2542
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 0.00040192307692307695,
+      "loss": 0.556,
+      "step": 2543
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 0.000401875,
+      "loss": 0.5034,
+      "step": 2544
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 0.0004018269230769231,
+      "loss": 0.5632,
+      "step": 2545
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 0.0004017788461538462,
+      "loss": 0.5338,
+      "step": 2546
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 0.00040173076923076924,
+      "loss": 0.5272,
+      "step": 2547
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 0.00040168269230769234,
+      "loss": 0.6161,
+      "step": 2548
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 0.0004016346153846154,
+      "loss": 0.5835,
+      "step": 2549
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 0.0004015865384615385,
+      "loss": 0.4702,
+      "step": 2550
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 0.00040153846153846153,
+      "loss": 0.628,
+      "step": 2551
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 0.00040149038461538463,
+      "loss": 0.4178,
+      "step": 2552
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 0.00040144230769230773,
+      "loss": 0.5076,
+      "step": 2553
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 0.0004013942307692308,
+      "loss": 0.5546,
+      "step": 2554
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 0.0004013461538461539,
+      "loss": 0.6627,
+      "step": 2555
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 0.0004012980769230769,
+      "loss": 0.5273,
+      "step": 2556
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 0.00040125,
+      "loss": 0.5379,
+      "step": 2557
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 0.0004012019230769231,
+      "loss": 0.7066,
+      "step": 2558
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 0.00040115384615384617,
+      "loss": 0.7838,
+      "step": 2559
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 0.00040110576923076927,
+      "loss": 0.466,
+      "step": 2560
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 0.0004010576923076923,
+      "loss": 0.3259,
+      "step": 2561
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 0.0004010096153846154,
+      "loss": 0.7197,
+      "step": 2562
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 0.0004009615384615385,
+      "loss": 0.7218,
+      "step": 2563
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 0.00040091346153846156,
+      "loss": 0.6559,
+      "step": 2564
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 0.00040086538461538466,
+      "loss": 0.6319,
+      "step": 2565
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 0.00040081730769230765,
+      "loss": 0.5508,
+      "step": 2566
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 0.00040076923076923075,
+      "loss": 0.6462,
+      "step": 2567
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 0.00040072115384615385,
+      "loss": 0.5999,
+      "step": 2568
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 0.0004006730769230769,
+      "loss": 0.5025,
+      "step": 2569
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 0.000400625,
+      "loss": 0.6801,
+      "step": 2570
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 0.00040057692307692304,
+      "loss": 0.5891,
+      "step": 2571
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 0.00040052884615384614,
+      "loss": 0.5537,
+      "step": 2572
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 0.00040048076923076924,
+      "loss": 0.6295,
+      "step": 2573
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 0.0004004326923076923,
+      "loss": 0.5818,
+      "step": 2574
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 0.0004003846153846154,
+      "loss": 0.5319,
+      "step": 2575
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 0.00040033653846153843,
+      "loss": 0.617,
+      "step": 2576
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 0.00040028846153846153,
+      "loss": 0.7261,
+      "step": 2577
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 0.00040024038461538463,
+      "loss": 0.5957,
+      "step": 2578
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 0.0004001923076923077,
+      "loss": 0.6239,
+      "step": 2579
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 0.0004001442307692308,
+      "loss": 0.5904,
+      "step": 2580
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 0.0004000961538461538,
+      "loss": 0.5802,
+      "step": 2581
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 0.0004000480769230769,
+      "loss": 0.5735,
+      "step": 2582
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 0.0004,
+      "loss": 0.602,
+      "step": 2583
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 0.00039995192307692306,
+      "loss": 0.4809,
+      "step": 2584
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 0.00039990384615384616,
+      "loss": 0.604,
+      "step": 2585
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 0.0003998557692307692,
+      "loss": 0.5225,
+      "step": 2586
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 0.0003998076923076923,
+      "loss": 0.6012,
+      "step": 2587
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 0.0003997596153846154,
+      "loss": 0.6639,
+      "step": 2588
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 0.00039971153846153845,
+      "loss": 0.7062,
+      "step": 2589
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 0.00039966346153846155,
+      "loss": 0.6435,
+      "step": 2590
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 0.0003996153846153846,
+      "loss": 0.7665,
+      "step": 2591
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 0.0003995673076923077,
+      "loss": 0.6656,
+      "step": 2592
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 0.0003995192307692308,
+      "loss": 0.5465,
+      "step": 2593
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 0.00039947115384615385,
+      "loss": 0.6097,
+      "step": 2594
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 0.00039942307692307694,
+      "loss": 0.6653,
+      "step": 2595
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 0.000399375,
+      "loss": 0.5856,
+      "step": 2596
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 0.0003993269230769231,
+      "loss": 0.5446,
+      "step": 2597
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 0.0003992788461538462,
+      "loss": 0.5657,
+      "step": 2598
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 0.00039923076923076924,
+      "loss": 0.4993,
+      "step": 2599
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 0.00039918269230769234,
+      "loss": 0.5601,
+      "step": 2600
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 0.0003991346153846154,
+      "loss": 0.6306,
+      "step": 2601
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 0.0003990865384615385,
+      "loss": 0.5267,
+      "step": 2602
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 0.0003990384615384616,
+      "loss": 0.6331,
+      "step": 2603
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 0.0003989903846153846,
+      "loss": 0.663,
+      "step": 2604
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 0.0003989423076923077,
+      "loss": 0.6326,
+      "step": 2605
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 0.00039889423076923077,
+      "loss": 0.4939,
+      "step": 2606
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 0.00039884615384615387,
+      "loss": 0.6057,
+      "step": 2607
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 0.00039879807692307697,
+      "loss": 0.6071,
+      "step": 2608
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 0.00039875,
+      "loss": 0.6099,
+      "step": 2609
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 0.0003987019230769231,
+      "loss": 0.4841,
+      "step": 2610
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 0.00039865384615384616,
+      "loss": 0.7512,
+      "step": 2611
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 0.00039860576923076926,
+      "loss": 0.5402,
+      "step": 2612
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 0.00039855769230769236,
+      "loss": 0.5362,
+      "step": 2613
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 0.0003985096153846154,
+      "loss": 0.6413,
+      "step": 2614
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 0.0003984615384615385,
+      "loss": 0.5457,
+      "step": 2615
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 0.00039841346153846155,
+      "loss": 0.7901,
+      "step": 2616
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 0.00039836538461538465,
+      "loss": 0.7028,
+      "step": 2617
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 0.0003983173076923077,
+      "loss": 0.576,
+      "step": 2618
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 0.0003982692307692308,
+      "loss": 0.6419,
+      "step": 2619
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 0.0003982211538461539,
+      "loss": 0.59,
+      "step": 2620
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 0.00039817307692307694,
+      "loss": 0.6174,
+      "step": 2621
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 0.00039812500000000004,
+      "loss": 0.6408,
+      "step": 2622
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 0.00039807692307692303,
+      "loss": 0.7042,
+      "step": 2623
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 0.00039802884615384613,
+      "loss": 0.5401,
+      "step": 2624
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 0.00039798076923076923,
+      "loss": 0.6064,
+      "step": 2625
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 0.0003979326923076923,
+      "loss": 0.6382,
+      "step": 2626
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 0.0003978846153846154,
+      "loss": 0.5971,
+      "step": 2627
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 0.0003978365384615384,
+      "loss": 0.5826,
+      "step": 2628
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 0.0003977884615384615,
+      "loss": 0.4817,
+      "step": 2629
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 0.0003977403846153846,
+      "loss": 0.5141,
+      "step": 2630
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 0.00039769230769230767,
+      "loss": 0.5638,
+      "step": 2631
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 0.00039764423076923077,
+      "loss": 0.5316,
+      "step": 2632
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 0.0003975961538461538,
+      "loss": 0.6595,
+      "step": 2633
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 0.0003975480769230769,
+      "loss": 0.5743,
+      "step": 2634
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 0.0003975,
+      "loss": 0.5384,
+      "step": 2635
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 0.00039745192307692306,
+      "loss": 0.6102,
+      "step": 2636
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 0.00039740384615384616,
+      "loss": 0.6595,
+      "step": 2637
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 0.0003973557692307692,
+      "loss": 0.6905,
+      "step": 2638
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 0.0003973076923076923,
+      "loss": 0.6514,
+      "step": 2639
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 0.0003972596153846154,
+      "loss": 0.6227,
+      "step": 2640
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 0.00039721153846153845,
+      "loss": 0.6056,
+      "step": 2641
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 0.00039716346153846155,
+      "loss": 0.5983,
+      "step": 2642
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 0.0003971153846153846,
+      "loss": 0.5395,
+      "step": 2643
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 0.0003970673076923077,
+      "loss": 0.7403,
+      "step": 2644
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 0.0003970192307692308,
+      "loss": 0.6311,
+      "step": 2645
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 0.00039697115384615384,
+      "loss": 0.6751,
+      "step": 2646
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 0.00039692307692307694,
+      "loss": 0.6546,
+      "step": 2647
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 0.000396875,
+      "loss": 0.5337,
+      "step": 2648
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 0.0003968269230769231,
+      "loss": 0.6255,
+      "step": 2649
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 0.0003967788461538462,
+      "loss": 0.6853,
+      "step": 2650
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 0.00039673076923076923,
+      "loss": 0.5435,
+      "step": 2651
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 0.00039668269230769233,
+      "loss": 0.5561,
+      "step": 2652
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 0.0003966346153846154,
+      "loss": 0.6882,
+      "step": 2653
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 0.0003965865384615385,
+      "loss": 0.6547,
+      "step": 2654
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 0.0003965384615384616,
+      "loss": 0.6084,
+      "step": 2655
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 0.0003964903846153846,
+      "loss": 0.5745,
+      "step": 2656
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 0.0003964423076923077,
+      "loss": 0.598,
+      "step": 2657
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 0.00039639423076923076,
+      "loss": 0.491,
+      "step": 2658
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 0.00039634615384615386,
+      "loss": 0.5754,
+      "step": 2659
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 0.00039629807692307696,
+      "loss": 0.5467,
+      "step": 2660
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 0.00039625,
+      "loss": 0.724,
+      "step": 2661
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 0.0003962019230769231,
+      "loss": 0.5875,
+      "step": 2662
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 0.00039615384615384615,
+      "loss": 0.5553,
+      "step": 2663
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 0.00039610576923076925,
+      "loss": 0.5654,
+      "step": 2664
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 0.00039605769230769235,
+      "loss": 0.7039,
+      "step": 2665
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 0.0003960096153846154,
+      "loss": 0.672,
+      "step": 2666
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 0.0003959615384615385,
+      "loss": 0.6385,
+      "step": 2667
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 0.00039591346153846154,
+      "loss": 0.6082,
+      "step": 2668
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 0.00039586538461538464,
+      "loss": 0.6721,
+      "step": 2669
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 0.00039581730769230774,
+      "loss": 0.5467,
+      "step": 2670
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 0.0003957692307692308,
+      "loss": 0.6547,
+      "step": 2671
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 0.0003957211538461539,
+      "loss": 0.5873,
+      "step": 2672
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 0.00039567307692307694,
+      "loss": 0.6568,
+      "step": 2673
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 0.00039562500000000003,
+      "loss": 0.7079,
+      "step": 2674
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 0.00039557692307692313,
+      "loss": 0.4988,
+      "step": 2675
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 0.0003955288461538462,
+      "loss": 0.571,
+      "step": 2676
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 0.0003954807692307693,
+      "loss": 0.5686,
+      "step": 2677
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 0.0003954326923076923,
+      "loss": 0.5767,
+      "step": 2678
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 0.0003953846153846154,
+      "loss": 0.4596,
+      "step": 2679
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 0.00039533653846153847,
+      "loss": 0.4791,
+      "step": 2680
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 0.0003952884615384615,
+      "loss": 0.4916,
+      "step": 2681
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 0.0003952403846153846,
+      "loss": 0.6213,
+      "step": 2682
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 0.00039519230769230766,
+      "loss": 0.5447,
+      "step": 2683
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 0.00039514423076923076,
+      "loss": 0.6327,
+      "step": 2684
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 0.0003950961538461538,
+      "loss": 0.5978,
+      "step": 2685
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 0.0003950480769230769,
+      "loss": 0.7014,
+      "step": 2686
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 0.000395,
+      "loss": 0.5016,
+      "step": 2687
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 0.00039495192307692305,
+      "loss": 0.5417,
+      "step": 2688
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 0.00039490384615384615,
+      "loss": 0.5263,
+      "step": 2689
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 0.0003948557692307692,
+      "loss": 0.743,
+      "step": 2690
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 0.0003948076923076923,
+      "loss": 0.5133,
+      "step": 2691
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 0.0003947596153846154,
+      "loss": 0.5012,
+      "step": 2692
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 0.00039471153846153844,
+      "loss": 0.6606,
+      "step": 2693
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 0.00039466346153846154,
+      "loss": 0.4796,
+      "step": 2694
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 0.0003946153846153846,
+      "loss": 0.4839,
+      "step": 2695
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 0.0003945673076923077,
+      "loss": 0.475,
+      "step": 2696
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 0.0003945192307692308,
+      "loss": 0.5534,
+      "step": 2697
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 0.00039447115384615383,
+      "loss": 0.7056,
+      "step": 2698
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 0.00039442307692307693,
+      "loss": 0.7464,
+      "step": 2699
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 0.000394375,
+      "loss": 0.6527,
+      "step": 2700
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 0.0003943269230769231,
+      "loss": 0.5971,
+      "step": 2701
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 0.0003942788461538462,
+      "loss": 0.5614,
+      "step": 2702
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 0.0003942307692307692,
+      "loss": 0.5737,
+      "step": 2703
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 0.0003941826923076923,
+      "loss": 0.5819,
+      "step": 2704
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 0.00039413461538461537,
+      "loss": 0.5384,
+      "step": 2705
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 0.00039408653846153847,
+      "loss": 0.6053,
+      "step": 2706
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 0.00039403846153846157,
+      "loss": 0.5138,
+      "step": 2707
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 0.0003939903846153846,
+      "loss": 0.4887,
+      "step": 2708
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 0.0003939423076923077,
+      "loss": 0.7036,
+      "step": 2709
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 0.00039389423076923076,
+      "loss": 0.5589,
+      "step": 2710
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 0.00039384615384615386,
+      "loss": 0.6495,
+      "step": 2711
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 0.00039379807692307696,
+      "loss": 0.5717,
+      "step": 2712
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 0.00039375,
+      "loss": 0.6799,
+      "step": 2713
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 0.0003937019230769231,
+      "loss": 0.457,
+      "step": 2714
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 0.00039365384615384615,
+      "loss": 0.6685,
+      "step": 2715
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 0.00039360576923076925,
+      "loss": 0.6746,
+      "step": 2716
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 0.00039355769230769235,
+      "loss": 0.5711,
+      "step": 2717
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 0.0003935096153846154,
+      "loss": 0.588,
+      "step": 2718
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 0.0003934615384615385,
+      "loss": 0.5624,
+      "step": 2719
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 0.00039341346153846154,
+      "loss": 0.6303,
+      "step": 2720
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 0.00039336538461538464,
+      "loss": 0.6741,
+      "step": 2721
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 0.00039331730769230774,
+      "loss": 0.5627,
+      "step": 2722
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 0.0003932692307692308,
+      "loss": 0.7558,
+      "step": 2723
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 0.0003932211538461539,
+      "loss": 0.4034,
+      "step": 2724
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 0.00039317307692307693,
+      "loss": 0.5525,
+      "step": 2725
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 0.00039312500000000003,
+      "loss": 0.6705,
+      "step": 2726
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 0.00039307692307692313,
+      "loss": 0.6069,
+      "step": 2727
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 0.0003930288461538462,
+      "loss": 0.6252,
+      "step": 2728
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 0.0003929807692307693,
+      "loss": 0.5429,
+      "step": 2729
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 0.0003929326923076923,
+      "loss": 0.6017,
+      "step": 2730
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 0.0003928846153846154,
+      "loss": 0.5528,
+      "step": 2731
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 0.0003928365384615385,
+      "loss": 0.545,
+      "step": 2732
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 0.00039278846153846156,
+      "loss": 0.6067,
+      "step": 2733
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 0.00039274038461538466,
+      "loss": 0.56,
+      "step": 2734
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 0.0003926923076923077,
+      "loss": 0.5249,
+      "step": 2735
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 0.00039264423076923075,
+      "loss": 0.6001,
+      "step": 2736
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 0.00039259615384615385,
+      "loss": 0.5058,
+      "step": 2737
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 0.0003925480769230769,
+      "loss": 0.529,
+      "step": 2738
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 0.0003925,
+      "loss": 0.5878,
+      "step": 2739
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 0.00039245192307692305,
+      "loss": 0.5992,
+      "step": 2740
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 0.00039240384615384614,
+      "loss": 0.6132,
+      "step": 2741
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 0.00039235576923076924,
+      "loss": 0.4081,
+      "step": 2742
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 0.0003923076923076923,
+      "loss": 0.5283,
+      "step": 2743
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 0.0003922596153846154,
+      "loss": 0.5814,
+      "step": 2744
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 0.00039221153846153844,
+      "loss": 0.5613,
+      "step": 2745
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 0.00039216346153846153,
+      "loss": 0.5268,
+      "step": 2746
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 0.0003921153846153846,
+      "loss": 0.5279,
+      "step": 2747
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 0.0003920673076923077,
+      "loss": 0.6086,
+      "step": 2748
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 0.0003920192307692308,
+      "loss": 0.4477,
+      "step": 2749
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 0.0003919711538461538,
+      "loss": 0.588,
+      "step": 2750
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 0.0003919230769230769,
+      "loss": 0.6362,
+      "step": 2751
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 0.00039187499999999997,
+      "loss": 0.6096,
+      "step": 2752
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 0.00039182692307692307,
+      "loss": 0.6817,
+      "step": 2753
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 0.00039177884615384617,
+      "loss": 0.5214,
+      "step": 2754
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 0.0003917307692307692,
+      "loss": 0.6108,
+      "step": 2755
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 0.0003916826923076923,
+      "loss": 0.595,
+      "step": 2756
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 0.00039163461538461536,
+      "loss": 0.589,
+      "step": 2757
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 0.00039158653846153846,
+      "loss": 0.5737,
+      "step": 2758
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 0.00039153846153846156,
+      "loss": 0.5714,
+      "step": 2759
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 0.0003914903846153846,
+      "loss": 0.5663,
+      "step": 2760
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 0.0003914423076923077,
+      "loss": 0.5069,
+      "step": 2761
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 0.00039139423076923075,
+      "loss": 0.5924,
+      "step": 2762
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 0.00039134615384615385,
+      "loss": 0.597,
+      "step": 2763
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 0.00039129807692307695,
+      "loss": 0.5202,
+      "step": 2764
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 0.00039125,
+      "loss": 0.4787,
+      "step": 2765
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 0.0003912019230769231,
+      "loss": 0.6174,
+      "step": 2766
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 0.00039115384615384614,
+      "loss": 0.4152,
+      "step": 2767
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 0.00039110576923076924,
+      "loss": 0.5107,
+      "step": 2768
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 0.00039105769230769234,
+      "loss": 0.614,
+      "step": 2769
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 0.0003910096153846154,
+      "loss": 0.5374,
+      "step": 2770
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 0.0003909615384615385,
+      "loss": 0.6396,
+      "step": 2771
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 0.00039091346153846153,
+      "loss": 0.464,
+      "step": 2772
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 0.00039086538461538463,
+      "loss": 0.5795,
+      "step": 2773
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 0.00039081730769230773,
+      "loss": 0.534,
+      "step": 2774
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 0.0003907692307692308,
+      "loss": 0.4762,
+      "step": 2775
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 0.0003907211538461539,
+      "loss": 0.6435,
+      "step": 2776
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 0.0003906730769230769,
+      "loss": 0.4886,
+      "step": 2777
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 0.000390625,
+      "loss": 0.5179,
+      "step": 2778
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 0.0003905769230769231,
+      "loss": 0.5136,
+      "step": 2779
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 0.00039052884615384617,
+      "loss": 0.7018,
+      "step": 2780
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 0.00039048076923076927,
+      "loss": 0.5649,
+      "step": 2781
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 0.0003904326923076923,
+      "loss": 0.5979,
+      "step": 2782
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 0.0003903846153846154,
+      "loss": 0.5599,
+      "step": 2783
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 0.0003903365384615385,
+      "loss": 0.594,
+      "step": 2784
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 0.00039028846153846156,
+      "loss": 0.6326,
+      "step": 2785
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 0.00039024038461538466,
+      "loss": 0.5068,
+      "step": 2786
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 0.0003901923076923077,
+      "loss": 0.5188,
+      "step": 2787
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 0.0003901442307692308,
+      "loss": 0.6191,
+      "step": 2788
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 0.0003900961538461539,
+      "loss": 0.5191,
+      "step": 2789
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 0.00039004807692307695,
+      "loss": 0.5632,
+      "step": 2790
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 0.4714,
+      "step": 2791
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 0.00038995192307692304,
+      "loss": 0.5423,
+      "step": 2792
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 0.00038990384615384614,
+      "loss": 0.5255,
+      "step": 2793
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 0.00038985576923076924,
+      "loss": 0.4405,
+      "step": 2794
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 0.0003898076923076923,
+      "loss": 0.67,
+      "step": 2795
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 0.0003897596153846154,
+      "loss": 0.5527,
+      "step": 2796
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 0.00038971153846153843,
+      "loss": 0.6636,
+      "step": 2797
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 0.00038966346153846153,
+      "loss": 0.6702,
+      "step": 2798
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 0.00038961538461538463,
+      "loss": 0.5352,
+      "step": 2799
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 0.0003895673076923077,
+      "loss": 0.6286,
+      "step": 2800
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 0.0003895192307692308,
+      "loss": 0.5956,
+      "step": 2801
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 0.0003894711538461538,
+      "loss": 0.4048,
+      "step": 2802
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 0.0003894230769230769,
+      "loss": 0.577,
+      "step": 2803
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 0.000389375,
+      "loss": 0.624,
+      "step": 2804
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 0.00038932692307692306,
+      "loss": 0.586,
+      "step": 2805
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 0.00038927884615384616,
+      "loss": 0.4524,
+      "step": 2806
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 0.0003892307692307692,
+      "loss": 0.6823,
+      "step": 2807
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 0.0003891826923076923,
+      "loss": 0.6618,
+      "step": 2808
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 0.0003891346153846154,
+      "loss": 0.6205,
+      "step": 2809
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 0.00038908653846153845,
+      "loss": 0.4724,
+      "step": 2810
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 0.00038903846153846155,
+      "loss": 0.5882,
+      "step": 2811
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 0.0003889903846153846,
+      "loss": 0.5359,
+      "step": 2812
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 0.0003889423076923077,
+      "loss": 0.4039,
+      "step": 2813
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 0.00038889423076923074,
+      "loss": 0.5945,
+      "step": 2814
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 0.00038884615384615384,
+      "loss": 0.5235,
+      "step": 2815
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 0.00038879807692307694,
+      "loss": 0.4607,
+      "step": 2816
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 0.00038875,
+      "loss": 0.5486,
+      "step": 2817
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 0.0003887019230769231,
+      "loss": 0.6251,
+      "step": 2818
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 0.00038865384615384613,
+      "loss": 0.5498,
+      "step": 2819
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 0.00038860576923076923,
+      "loss": 0.4949,
+      "step": 2820
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 0.00038855769230769233,
+      "loss": 0.6317,
+      "step": 2821
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 0.0003885096153846154,
+      "loss": 0.5853,
+      "step": 2822
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 0.0003884615384615385,
+      "loss": 0.6299,
+      "step": 2823
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 0.0003884134615384615,
+      "loss": 0.3827,
+      "step": 2824
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 0.0003883653846153846,
+      "loss": 0.479,
+      "step": 2825
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 0.0003883173076923077,
+      "loss": 0.5514,
+      "step": 2826
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 0.00038826923076923077,
+      "loss": 0.5305,
+      "step": 2827
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 0.00038822115384615387,
+      "loss": 0.613,
+      "step": 2828
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 0.0003881730769230769,
+      "loss": 0.5811,
+      "step": 2829
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 0.000388125,
+      "loss": 0.4676,
+      "step": 2830
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 0.0003880769230769231,
+      "loss": 0.5443,
+      "step": 2831
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 0.00038802884615384616,
+      "loss": 0.6361,
+      "step": 2832
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 0.00038798076923076926,
+      "loss": 0.5137,
+      "step": 2833
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 0.0003879326923076923,
+      "loss": 0.5945,
+      "step": 2834
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 0.0003878846153846154,
+      "loss": 0.6448,
+      "step": 2835
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 0.0003878365384615385,
+      "loss": 0.4611,
+      "step": 2836
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 0.00038778846153846155,
+      "loss": 0.6191,
+      "step": 2837
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 0.00038774038461538465,
+      "loss": 0.5608,
+      "step": 2838
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 0.0003876923076923077,
+      "loss": 0.6766,
+      "step": 2839
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 0.0003876442307692308,
+      "loss": 0.6206,
+      "step": 2840
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 0.0003875961538461539,
+      "loss": 0.4514,
+      "step": 2841
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 0.00038754807692307694,
+      "loss": 0.5582,
+      "step": 2842
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 0.00038750000000000004,
+      "loss": 0.4312,
+      "step": 2843
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 0.0003874519230769231,
+      "loss": 0.5389,
+      "step": 2844
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 0.0003874038461538462,
+      "loss": 0.5027,
+      "step": 2845
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 0.0003873557692307693,
+      "loss": 0.5593,
+      "step": 2846
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 0.00038730769230769233,
+      "loss": 0.5774,
+      "step": 2847
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 0.00038725961538461543,
+      "loss": 0.5396,
+      "step": 2848
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 0.0003872115384615384,
+      "loss": 0.6153,
+      "step": 2849
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 0.0003871634615384615,
+      "loss": 0.5015,
+      "step": 2850
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 0.0003871153846153846,
+      "loss": 0.4644,
+      "step": 2851
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 0.00038706730769230767,
+      "loss": 0.5993,
+      "step": 2852
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 0.00038701923076923077,
+      "loss": 0.6176,
+      "step": 2853
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 0.0003869711538461538,
+      "loss": 0.4388,
+      "step": 2854
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 0.0003869230769230769,
+      "loss": 0.5199,
+      "step": 2855
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 0.000386875,
+      "loss": 0.6205,
+      "step": 2856
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 0.00038682692307692306,
+      "loss": 0.4613,
+      "step": 2857
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 0.00038677884615384616,
+      "loss": 0.6005,
+      "step": 2858
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 0.0003867307692307692,
+      "loss": 0.6235,
+      "step": 2859
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 0.0003866826923076923,
+      "loss": 0.5057,
+      "step": 2860
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 0.0003866346153846154,
+      "loss": 0.3892,
+      "step": 2861
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 0.00038658653846153845,
+      "loss": 0.6132,
+      "step": 2862
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 0.00038653846153846155,
+      "loss": 0.6242,
+      "step": 2863
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 0.0003864903846153846,
+      "loss": 0.5775,
+      "step": 2864
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 0.0003864423076923077,
+      "loss": 0.6153,
+      "step": 2865
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 0.0003863942307692308,
+      "loss": 0.4656,
+      "step": 2866
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 0.00038634615384615384,
+      "loss": 0.5363,
+      "step": 2867
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 0.00038629807692307694,
+      "loss": 0.5227,
+      "step": 2868
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 0.00038625,
+      "loss": 0.4503,
+      "step": 2869
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 0.0003862019230769231,
+      "loss": 0.5,
+      "step": 2870
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 0.0003861538461538462,
+      "loss": 0.5462,
+      "step": 2871
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 0.00038610576923076923,
+      "loss": 0.5726,
+      "step": 2872
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 0.00038605769230769233,
+      "loss": 0.4916,
+      "step": 2873
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 0.0003860096153846154,
+      "loss": 0.53,
+      "step": 2874
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 0.0003859615384615385,
+      "loss": 0.5334,
+      "step": 2875
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 0.0003859134615384615,
+      "loss": 0.627,
+      "step": 2876
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 0.0003858653846153846,
+      "loss": 0.6107,
+      "step": 2877
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 0.0003858173076923077,
+      "loss": 0.4612,
+      "step": 2878
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 0.00038576923076923076,
+      "loss": 0.5746,
+      "step": 2879
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 0.00038572115384615386,
+      "loss": 0.43,
+      "step": 2880
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 0.0003856730769230769,
+      "loss": 0.5969,
+      "step": 2881
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 0.000385625,
+      "loss": 0.5614,
+      "step": 2882
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 0.0003855769230769231,
+      "loss": 0.4961,
+      "step": 2883
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 0.00038552884615384615,
+      "loss": 0.5097,
+      "step": 2884
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 0.00038548076923076925,
+      "loss": 0.543,
+      "step": 2885
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 0.0003854326923076923,
+      "loss": 0.6328,
+      "step": 2886
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 0.0003853846153846154,
+      "loss": 0.5582,
+      "step": 2887
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 0.0003853365384615385,
+      "loss": 0.5759,
+      "step": 2888
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 0.00038528846153846154,
+      "loss": 0.813,
+      "step": 2889
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 0.00038524038461538464,
+      "loss": 0.6362,
+      "step": 2890
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 0.0003851923076923077,
+      "loss": 0.6859,
+      "step": 2891
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 0.0003851442307692308,
+      "loss": 0.6175,
+      "step": 2892
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 0.0003850961538461539,
+      "loss": 0.6317,
+      "step": 2893
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 0.00038504807692307693,
+      "loss": 0.5719,
+      "step": 2894
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 0.00038500000000000003,
+      "loss": 0.5326,
+      "step": 2895
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 0.0003849519230769231,
+      "loss": 0.5586,
+      "step": 2896
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 0.0003849038461538462,
+      "loss": 0.516,
+      "step": 2897
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 0.0003848557692307693,
+      "loss": 0.4926,
+      "step": 2898
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 0.0003848076923076923,
+      "loss": 0.5307,
+      "step": 2899
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 0.0003847596153846154,
+      "loss": 0.4995,
+      "step": 2900
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 0.00038471153846153847,
+      "loss": 0.576,
+      "step": 2901
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 0.00038466346153846157,
+      "loss": 0.481,
+      "step": 2902
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 0.00038461538461538467,
+      "loss": 0.6336,
+      "step": 2903
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 0.0003845673076923077,
+      "loss": 0.524,
+      "step": 2904
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 0.0003845192307692308,
+      "loss": 0.596,
+      "step": 2905
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 0.0003844711538461538,
+      "loss": 0.4377,
+      "step": 2906
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 0.0003844230769230769,
+      "loss": 0.6358,
+      "step": 2907
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 0.000384375,
+      "loss": 0.5734,
+      "step": 2908
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 0.00038432692307692305,
+      "loss": 0.6247,
+      "step": 2909
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 0.00038427884615384615,
+      "loss": 0.5728,
+      "step": 2910
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 0.0003842307692307692,
+      "loss": 0.4783,
+      "step": 2911
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 0.0003841826923076923,
+      "loss": 0.6257,
+      "step": 2912
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 0.0003841346153846154,
+      "loss": 0.5488,
+      "step": 2913
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 0.00038408653846153844,
+      "loss": 0.5136,
+      "step": 2914
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 0.00038403846153846154,
+      "loss": 0.4637,
+      "step": 2915
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 0.0003839903846153846,
+      "loss": 0.6497,
+      "step": 2916
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 0.0003839423076923077,
+      "loss": 0.526,
+      "step": 2917
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 0.0003838942307692308,
+      "loss": 0.6754,
+      "step": 2918
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 0.00038384615384615383,
+      "loss": 0.6659,
+      "step": 2919
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 0.00038379807692307693,
+      "loss": 0.4612,
+      "step": 2920
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 0.00038375,
+      "loss": 0.6187,
+      "step": 2921
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 0.0003837019230769231,
+      "loss": 0.5538,
+      "step": 2922
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 0.0003836538461538462,
+      "loss": 0.4943,
+      "step": 2923
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 0.0003836057692307692,
+      "loss": 0.5777,
+      "step": 2924
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 0.0003835576923076923,
+      "loss": 0.4794,
+      "step": 2925
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 0.00038350961538461537,
+      "loss": 0.522,
+      "step": 2926
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 0.00038346153846153847,
+      "loss": 0.6446,
+      "step": 2927
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 0.00038341346153846157,
+      "loss": 0.5921,
+      "step": 2928
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 0.0003833653846153846,
+      "loss": 0.5354,
+      "step": 2929
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 0.0003833173076923077,
+      "loss": 0.6577,
+      "step": 2930
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 0.00038326923076923076,
+      "loss": 0.6223,
+      "step": 2931
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 0.00038322115384615386,
+      "loss": 0.6549,
+      "step": 2932
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 0.00038317307692307696,
+      "loss": 0.367,
+      "step": 2933
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 0.000383125,
+      "loss": 0.5651,
+      "step": 2934
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 0.0003830769230769231,
+      "loss": 0.5562,
+      "step": 2935
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 0.00038302884615384615,
+      "loss": 0.612,
+      "step": 2936
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 0.00038298076923076925,
+      "loss": 0.6198,
+      "step": 2937
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 0.00038293269230769235,
+      "loss": 0.5591,
+      "step": 2938
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 0.0003828846153846154,
+      "loss": 0.5471,
+      "step": 2939
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 0.0003828365384615385,
+      "loss": 0.5013,
+      "step": 2940
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 0.00038278846153846154,
+      "loss": 0.635,
+      "step": 2941
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 0.00038274038461538464,
+      "loss": 0.4455,
+      "step": 2942
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 0.0003826923076923077,
+      "loss": 0.5263,
+      "step": 2943
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 0.0003826442307692308,
+      "loss": 0.5802,
+      "step": 2944
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 0.0003825961538461539,
+      "loss": 0.5957,
+      "step": 2945
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 0.00038254807692307693,
+      "loss": 0.5694,
+      "step": 2946
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 0.00038250000000000003,
+      "loss": 0.5798,
+      "step": 2947
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 0.0003824519230769231,
+      "loss": 0.5521,
+      "step": 2948
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 0.00038240384615384617,
+      "loss": 0.6418,
+      "step": 2949
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 0.00038235576923076927,
+      "loss": 0.4189,
+      "step": 2950
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 0.0003823076923076923,
+      "loss": 0.514,
+      "step": 2951
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 0.0003822596153846154,
+      "loss": 0.5897,
+      "step": 2952
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 0.00038221153846153846,
+      "loss": 0.5225,
+      "step": 2953
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 0.00038216346153846156,
+      "loss": 0.4985,
+      "step": 2954
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 0.00038211538461538466,
+      "loss": 0.4537,
+      "step": 2955
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 0.0003820673076923077,
+      "loss": 0.5544,
+      "step": 2956
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 0.0003820192307692308,
+      "loss": 0.5335,
+      "step": 2957
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 0.00038197115384615385,
+      "loss": 0.4597,
+      "step": 2958
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 0.00038192307692307695,
+      "loss": 0.6317,
+      "step": 2959
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 0.00038187500000000005,
+      "loss": 0.5275,
+      "step": 2960
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 0.0003818269230769231,
+      "loss": 0.4268,
+      "step": 2961
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 0.00038177884615384614,
+      "loss": 0.6314,
+      "step": 2962
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 0.0003817307692307692,
+      "loss": 0.5054,
+      "step": 2963
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 0.0003816826923076923,
+      "loss": 0.4128,
+      "step": 2964
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 0.0003816346153846154,
+      "loss": 0.5114,
+      "step": 2965
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 0.00038158653846153843,
+      "loss": 0.4643,
+      "step": 2966
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 0.00038153846153846153,
+      "loss": 0.4673,
+      "step": 2967
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 0.0003814903846153846,
+      "loss": 0.6345,
+      "step": 2968
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 0.0003814423076923077,
+      "loss": 0.6023,
+      "step": 2969
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 0.0003813942307692308,
+      "loss": 0.5502,
+      "step": 2970
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 0.0003813461538461538,
+      "loss": 0.6927,
+      "step": 2971
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 0.0003812980769230769,
+      "loss": 0.4282,
+      "step": 2972
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 0.00038124999999999997,
+      "loss": 0.5137,
+      "step": 2973
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 0.00038120192307692307,
+      "loss": 0.4987,
+      "step": 2974
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 0.00038115384615384617,
+      "loss": 0.6296,
+      "step": 2975
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 0.0003811057692307692,
+      "loss": 0.4939,
+      "step": 2976
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 0.0003810576923076923,
+      "loss": 0.6331,
+      "step": 2977
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 0.00038100961538461536,
+      "loss": 0.5276,
+      "step": 2978
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 0.00038096153846153846,
+      "loss": 0.5522,
+      "step": 2979
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 0.00038091346153846156,
+      "loss": 0.4772,
+      "step": 2980
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 0.0003808653846153846,
+      "loss": 0.6305,
+      "step": 2981
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 0.0003808173076923077,
+      "loss": 0.5313,
+      "step": 2982
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 0.00038076923076923075,
+      "loss": 0.4743,
+      "step": 2983
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 0.00038072115384615385,
+      "loss": 0.5418,
+      "step": 2984
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 0.00038067307692307695,
+      "loss": 0.5528,
+      "step": 2985
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 0.000380625,
+      "loss": 0.5731,
+      "step": 2986
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 0.0003805769230769231,
+      "loss": 0.4431,
+      "step": 2987
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 0.00038052884615384614,
+      "loss": 0.5834,
+      "step": 2988
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 0.00038048076923076924,
+      "loss": 0.5941,
+      "step": 2989
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 0.00038043269230769234,
+      "loss": 0.5744,
+      "step": 2990
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 0.0003803846153846154,
+      "loss": 0.5861,
+      "step": 2991
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 0.0003803365384615385,
+      "loss": 0.523,
+      "step": 2992
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 0.00038028846153846153,
+      "loss": 0.4214,
+      "step": 2993
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 0.00038024038461538463,
+      "loss": 0.5593,
+      "step": 2994
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 0.00038019230769230773,
+      "loss": 0.625,
+      "step": 2995
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 0.0003801442307692308,
+      "loss": 0.5459,
+      "step": 2996
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 0.0003800961538461539,
+      "loss": 0.3196,
+      "step": 2997
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 0.0003800480769230769,
+      "loss": 0.5955,
+      "step": 2998
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 0.00038,
+      "loss": 0.6885,
+      "step": 2999
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 0.0003799519230769231,
+      "loss": 0.624,
+      "step": 3000
+    },
+    {
+      "epoch": 27.52,
+      "eval_cer": 0.18202671704223453,
+      "eval_loss": 0.6940117478370667,
+      "eval_runtime": 87.4711,
+      "eval_samples_per_second": 18.829,
+      "eval_steps_per_second": 2.355,
+      "eval_wer": 0.6115820651618834,
+      "step": 3000
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 0.00037990384615384617,
+      "loss": 0.6165,
+      "step": 3001
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 0.00037985576923076927,
+      "loss": 0.5871,
+      "step": 3002
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 0.0003798076923076923,
+      "loss": 0.5926,
+      "step": 3003
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 0.0003797596153846154,
+      "loss": 0.5124,
+      "step": 3004
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 0.0003797115384615385,
+      "loss": 0.6283,
+      "step": 3005
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 0.00037966346153846156,
+      "loss": 0.5408,
+      "step": 3006
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 0.00037961538461538466,
+      "loss": 0.5754,
+      "step": 3007
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 0.0003795673076923077,
+      "loss": 0.6286,
+      "step": 3008
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 0.0003795192307692308,
+      "loss": 0.5521,
+      "step": 3009
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 0.00037947115384615385,
+      "loss": 0.5175,
+      "step": 3010
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 0.00037942307692307695,
+      "loss": 0.5282,
+      "step": 3011
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 0.00037937500000000005,
+      "loss": 0.5873,
+      "step": 3012
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 0.0003793269230769231,
+      "loss": 0.4465,
+      "step": 3013
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 0.0003792788461538462,
+      "loss": 0.5161,
+      "step": 3014
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 0.00037923076923076924,
+      "loss": 0.541,
+      "step": 3015
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 0.00037918269230769234,
+      "loss": 0.5143,
+      "step": 3016
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 0.00037913461538461544,
+      "loss": 0.5854,
+      "step": 3017
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 0.00037908653846153843,
+      "loss": 0.5277,
+      "step": 3018
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 0.00037903846153846153,
+      "loss": 0.484,
+      "step": 3019
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 0.0003789903846153846,
+      "loss": 0.5042,
+      "step": 3020
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 0.0003789423076923077,
+      "loss": 0.54,
+      "step": 3021
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 0.00037889423076923077,
+      "loss": 0.5237,
+      "step": 3022
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 0.0003788461538461538,
+      "loss": 0.4752,
+      "step": 3023
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 0.0003787980769230769,
+      "loss": 0.6015,
+      "step": 3024
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 0.00037874999999999996,
+      "loss": 0.6935,
+      "step": 3025
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 0.00037870192307692306,
+      "loss": 0.6946,
+      "step": 3026
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 0.00037865384615384616,
+      "loss": 0.637,
+      "step": 3027
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 0.0003786057692307692,
+      "loss": 0.5038,
+      "step": 3028
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 0.0003785576923076923,
+      "loss": 0.6329,
+      "step": 3029
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 0.00037850961538461535,
+      "loss": 0.6164,
+      "step": 3030
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 0.00037846153846153845,
+      "loss": 0.4595,
+      "step": 3031
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 0.00037841346153846155,
+      "loss": 0.516,
+      "step": 3032
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 0.0003783653846153846,
+      "loss": 0.5519,
+      "step": 3033
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 0.0003783173076923077,
+      "loss": 0.4839,
+      "step": 3034
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 0.00037826923076923074,
+      "loss": 0.5074,
+      "step": 3035
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 0.00037822115384615384,
+      "loss": 0.5267,
+      "step": 3036
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 0.00037817307692307694,
+      "loss": 0.5083,
+      "step": 3037
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 0.000378125,
+      "loss": 0.4624,
+      "step": 3038
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 0.0003780769230769231,
+      "loss": 0.4997,
+      "step": 3039
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 0.00037802884615384613,
+      "loss": 0.4866,
+      "step": 3040
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 0.00037798076923076923,
+      "loss": 0.4322,
+      "step": 3041
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 0.00037793269230769233,
+      "loss": 0.5435,
+      "step": 3042
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 0.0003778846153846154,
+      "loss": 0.5312,
+      "step": 3043
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 0.0003778365384615385,
+      "loss": 0.5971,
+      "step": 3044
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 0.0003777884615384615,
+      "loss": 0.4738,
+      "step": 3045
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 0.0003777403846153846,
+      "loss": 0.5589,
+      "step": 3046
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 0.0003776923076923077,
+      "loss": 0.4791,
+      "step": 3047
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 0.00037764423076923077,
+      "loss": 0.6133,
+      "step": 3048
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 0.00037759615384615387,
+      "loss": 0.6265,
+      "step": 3049
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 0.0003775480769230769,
+      "loss": 0.5571,
+      "step": 3050
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 0.0003775,
+      "loss": 0.5098,
+      "step": 3051
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 0.0003774519230769231,
+      "loss": 0.5865,
+      "step": 3052
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 0.00037740384615384616,
+      "loss": 0.6495,
+      "step": 3053
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 0.00037735576923076926,
+      "loss": 0.6043,
+      "step": 3054
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 0.0003773076923076923,
+      "loss": 0.5159,
+      "step": 3055
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 0.0003772596153846154,
+      "loss": 0.5664,
+      "step": 3056
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 0.0003772115384615385,
+      "loss": 0.4996,
+      "step": 3057
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 0.00037716346153846155,
+      "loss": 0.508,
+      "step": 3058
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 0.00037711538461538465,
+      "loss": 0.4687,
+      "step": 3059
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 0.0003770673076923077,
+      "loss": 0.5343,
+      "step": 3060
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 0.0003770192307692308,
+      "loss": 0.4435,
+      "step": 3061
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 0.0003769711538461539,
+      "loss": 0.5721,
+      "step": 3062
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 0.00037692307692307694,
+      "loss": 0.6033,
+      "step": 3063
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 0.00037687500000000004,
+      "loss": 0.4552,
+      "step": 3064
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 0.0003768269230769231,
+      "loss": 0.4453,
+      "step": 3065
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 0.0003767788461538462,
+      "loss": 0.5549,
+      "step": 3066
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 0.0003767307692307693,
+      "loss": 0.5482,
+      "step": 3067
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 0.00037668269230769233,
+      "loss": 0.4581,
+      "step": 3068
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 0.00037663461538461543,
+      "loss": 0.4517,
+      "step": 3069
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 0.0003765865384615385,
+      "loss": 0.6173,
+      "step": 3070
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 0.0003765384615384616,
+      "loss": 0.4873,
+      "step": 3071
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 0.0003764903846153846,
+      "loss": 0.453,
+      "step": 3072
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 0.0003764423076923077,
+      "loss": 0.4929,
+      "step": 3073
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 0.0003763942307692308,
+      "loss": 0.4773,
+      "step": 3074
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 0.0003763461538461538,
+      "loss": 0.5097,
+      "step": 3075
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 0.0003762980769230769,
+      "loss": 0.5417,
+      "step": 3076
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 0.00037624999999999996,
+      "loss": 0.4977,
+      "step": 3077
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 0.00037620192307692306,
+      "loss": 0.4854,
+      "step": 3078
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 0.00037615384615384616,
+      "loss": 0.5225,
+      "step": 3079
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 0.0003761057692307692,
+      "loss": 0.5839,
+      "step": 3080
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 0.0003760576923076923,
+      "loss": 0.5905,
+      "step": 3081
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 0.00037600961538461535,
+      "loss": 0.6653,
+      "step": 3082
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 0.00037596153846153845,
+      "loss": 0.5423,
+      "step": 3083
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 0.00037591346153846155,
+      "loss": 0.5146,
+      "step": 3084
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 0.0003758653846153846,
+      "loss": 0.4782,
+      "step": 3085
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 0.0003758173076923077,
+      "loss": 0.5599,
+      "step": 3086
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 0.00037576923076923074,
+      "loss": 0.5438,
+      "step": 3087
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 0.00037572115384615384,
+      "loss": 0.4426,
+      "step": 3088
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 0.00037567307692307694,
+      "loss": 0.452,
+      "step": 3089
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 0.000375625,
+      "loss": 0.6385,
+      "step": 3090
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 0.0003755769230769231,
+      "loss": 0.6636,
+      "step": 3091
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 0.00037552884615384613,
+      "loss": 0.5867,
+      "step": 3092
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 0.00037548076923076923,
+      "loss": 0.391,
+      "step": 3093
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 0.0003754326923076923,
+      "loss": 0.5914,
+      "step": 3094
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 0.00037538461538461537,
+      "loss": 0.5173,
+      "step": 3095
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 0.00037533653846153847,
+      "loss": 0.5316,
+      "step": 3096
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 0.0003752884615384615,
+      "loss": 0.5064,
+      "step": 3097
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 0.0003752403846153846,
+      "loss": 0.4441,
+      "step": 3098
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 0.0003751923076923077,
+      "loss": 0.5225,
+      "step": 3099
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 0.00037514423076923076,
+      "loss": 0.444,
+      "step": 3100
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 0.00037509615384615386,
+      "loss": 0.5019,
+      "step": 3101
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 0.0003750480769230769,
+      "loss": 0.4721,
+      "step": 3102
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 0.000375,
+      "loss": 0.442,
+      "step": 3103
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 0.0003749519230769231,
+      "loss": 0.6518,
+      "step": 3104
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 0.00037490384615384615,
+      "loss": 0.4385,
+      "step": 3105
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 0.00037485576923076925,
+      "loss": 0.7003,
+      "step": 3106
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 0.0003748076923076923,
+      "loss": 0.6604,
+      "step": 3107
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 0.0003747596153846154,
+      "loss": 0.7141,
+      "step": 3108
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 0.0003747115384615385,
+      "loss": 0.695,
+      "step": 3109
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 0.00037466346153846154,
+      "loss": 0.627,
+      "step": 3110
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 0.00037461538461538464,
+      "loss": 0.5289,
+      "step": 3111
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 0.0003745673076923077,
+      "loss": 0.6458,
+      "step": 3112
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 0.0003745192307692308,
+      "loss": 0.5418,
+      "step": 3113
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 0.0003744711538461539,
+      "loss": 0.5431,
+      "step": 3114
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 0.00037442307692307693,
+      "loss": 0.598,
+      "step": 3115
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 0.00037437500000000003,
+      "loss": 0.5042,
+      "step": 3116
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 0.0003743269230769231,
+      "loss": 0.5428,
+      "step": 3117
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 0.0003742788461538462,
+      "loss": 0.5173,
+      "step": 3118
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 0.0003742307692307693,
+      "loss": 0.4932,
+      "step": 3119
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 0.0003741826923076923,
+      "loss": 0.6402,
+      "step": 3120
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 0.0003741346153846154,
+      "loss": 0.552,
+      "step": 3121
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 0.00037408653846153847,
+      "loss": 0.4415,
+      "step": 3122
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 0.00037403846153846157,
+      "loss": 0.6018,
+      "step": 3123
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 0.00037399038461538467,
+      "loss": 0.5461,
+      "step": 3124
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 0.0003739423076923077,
+      "loss": 0.4947,
+      "step": 3125
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 0.0003738942307692308,
+      "loss": 0.5448,
+      "step": 3126
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 0.00037384615384615386,
+      "loss": 0.4864,
+      "step": 3127
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 0.00037379807692307696,
+      "loss": 0.5516,
+      "step": 3128
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 0.00037375000000000006,
+      "loss": 0.441,
+      "step": 3129
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 0.0003737019230769231,
+      "loss": 0.5365,
+      "step": 3130
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 0.0003736538461538462,
+      "loss": 0.6,
+      "step": 3131
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 0.0003736057692307692,
+      "loss": 0.511,
+      "step": 3132
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 0.0003735576923076923,
+      "loss": 0.4849,
+      "step": 3133
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 0.0003735096153846154,
+      "loss": 0.6405,
+      "step": 3134
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 0.00037346153846153844,
+      "loss": 0.5334,
+      "step": 3135
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 0.00037341346153846154,
+      "loss": 0.505,
+      "step": 3136
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 0.0003733653846153846,
+      "loss": 0.5442,
+      "step": 3137
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 0.0003733173076923077,
+      "loss": 0.6251,
+      "step": 3138
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 0.00037326923076923073,
+      "loss": 0.5223,
+      "step": 3139
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 0.00037322115384615383,
+      "loss": 0.6276,
+      "step": 3140
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 0.00037317307692307693,
+      "loss": 0.527,
+      "step": 3141
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 0.000373125,
+      "loss": 0.4283,
+      "step": 3142
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 0.0003730769230769231,
+      "loss": 0.4373,
+      "step": 3143
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 0.0003730288461538461,
+      "loss": 0.5681,
+      "step": 3144
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 0.0003729807692307692,
+      "loss": 0.5358,
+      "step": 3145
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 0.0003729326923076923,
+      "loss": 0.5029,
+      "step": 3146
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 0.00037288461538461537,
+      "loss": 0.5374,
+      "step": 3147
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 0.00037283653846153847,
+      "loss": 0.5374,
+      "step": 3148
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 0.0003727884615384615,
+      "loss": 0.4467,
+      "step": 3149
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 0.0003727403846153846,
+      "loss": 0.5343,
+      "step": 3150
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 0.0003726923076923077,
+      "loss": 0.6367,
+      "step": 3151
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 0.00037264423076923076,
+      "loss": 0.5295,
+      "step": 3152
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 0.00037259615384615386,
+      "loss": 0.474,
+      "step": 3153
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 0.0003725480769230769,
+      "loss": 0.6136,
+      "step": 3154
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 0.0003725,
+      "loss": 0.3314,
+      "step": 3155
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 0.0003724519230769231,
+      "loss": 0.4807,
+      "step": 3156
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 0.00037240384615384615,
+      "loss": 0.4997,
+      "step": 3157
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 0.00037235576923076925,
+      "loss": 0.6398,
+      "step": 3158
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 0.0003723076923076923,
+      "loss": 0.5685,
+      "step": 3159
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 0.0003722596153846154,
+      "loss": 0.4676,
+      "step": 3160
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 0.0003722115384615385,
+      "loss": 0.5819,
+      "step": 3161
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 0.00037216346153846154,
+      "loss": 0.5479,
+      "step": 3162
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 0.00037211538461538464,
+      "loss": 0.6533,
+      "step": 3163
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 0.0003720673076923077,
+      "loss": 0.6419,
+      "step": 3164
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 0.0003720192307692308,
+      "loss": 0.5394,
+      "step": 3165
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 0.0003719711538461539,
+      "loss": 0.5797,
+      "step": 3166
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 0.0003719230769230769,
+      "loss": 0.453,
+      "step": 3167
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 0.000371875,
+      "loss": 0.5168,
+      "step": 3168
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 0.00037182692307692307,
+      "loss": 0.4915,
+      "step": 3169
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 0.00037177884615384617,
+      "loss": 0.5299,
+      "step": 3170
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 0.00037173076923076927,
+      "loss": 0.5494,
+      "step": 3171
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 0.0003716826923076923,
+      "loss": 0.4564,
+      "step": 3172
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 0.0003716346153846154,
+      "loss": 0.4914,
+      "step": 3173
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 0.00037158653846153846,
+      "loss": 0.4769,
+      "step": 3174
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 0.00037153846153846156,
+      "loss": 0.5432,
+      "step": 3175
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 0.00037149038461538466,
+      "loss": 0.5491,
+      "step": 3176
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 0.0003714423076923077,
+      "loss": 0.5219,
+      "step": 3177
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 0.0003713942307692308,
+      "loss": 0.5283,
+      "step": 3178
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 0.00037134615384615385,
+      "loss": 0.5258,
+      "step": 3179
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 0.00037129807692307695,
+      "loss": 0.5491,
+      "step": 3180
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 0.00037125000000000005,
+      "loss": 0.4704,
+      "step": 3181
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 0.0003712019230769231,
+      "loss": 0.5237,
+      "step": 3182
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 0.0003711538461538462,
+      "loss": 0.495,
+      "step": 3183
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 0.00037110576923076924,
+      "loss": 0.3715,
+      "step": 3184
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 0.00037105769230769234,
+      "loss": 0.4377,
+      "step": 3185
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 0.00037100961538461544,
+      "loss": 0.6325,
+      "step": 3186
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 0.0003709615384615385,
+      "loss": 0.6082,
+      "step": 3187
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 0.0003709134615384616,
+      "loss": 0.4836,
+      "step": 3188
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 0.0003708653846153846,
+      "loss": 0.6556,
+      "step": 3189
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 0.0003708173076923077,
+      "loss": 0.5943,
+      "step": 3190
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 0.0003707692307692308,
+      "loss": 0.5854,
+      "step": 3191
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 0.0003707211538461538,
+      "loss": 0.5208,
+      "step": 3192
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 0.0003706730769230769,
+      "loss": 0.5207,
+      "step": 3193
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 0.00037062499999999997,
+      "loss": 0.5615,
+      "step": 3194
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 0.00037057692307692307,
+      "loss": 0.6012,
+      "step": 3195
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 0.00037052884615384617,
+      "loss": 0.5806,
+      "step": 3196
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 0.0003704807692307692,
+      "loss": 0.6022,
+      "step": 3197
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 0.0003704326923076923,
+      "loss": 0.4155,
+      "step": 3198
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 0.00037038461538461536,
+      "loss": 0.4232,
+      "step": 3199
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 0.00037033653846153846,
+      "loss": 0.4693,
+      "step": 3200
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 0.0003702884615384615,
+      "loss": 0.4576,
+      "step": 3201
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 0.0003702403846153846,
+      "loss": 0.496,
+      "step": 3202
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 0.0003701923076923077,
+      "loss": 0.5257,
+      "step": 3203
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 0.00037014423076923075,
+      "loss": 0.4457,
+      "step": 3204
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 0.00037009615384615385,
+      "loss": 0.5196,
+      "step": 3205
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 0.0003700480769230769,
+      "loss": 0.3798,
+      "step": 3206
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 0.00037,
+      "loss": 0.5768,
+      "step": 3207
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 0.0003699519230769231,
+      "loss": 0.5259,
+      "step": 3208
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 0.00036990384615384614,
+      "loss": 0.4235,
+      "step": 3209
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 0.00036985576923076924,
+      "loss": 0.428,
+      "step": 3210
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 0.0003698076923076923,
+      "loss": 0.3741,
+      "step": 3211
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 0.0003697596153846154,
+      "loss": 0.3929,
+      "step": 3212
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 0.0003697115384615385,
+      "loss": 0.4132,
+      "step": 3213
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 0.00036966346153846153,
+      "loss": 0.4735,
+      "step": 3214
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 0.00036961538461538463,
+      "loss": 0.424,
+      "step": 3215
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 0.0003695673076923077,
+      "loss": 0.6209,
+      "step": 3216
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 0.0003695192307692308,
+      "loss": 0.5674,
+      "step": 3217
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 0.0003694711538461539,
+      "loss": 0.434,
+      "step": 3218
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 0.0003694230769230769,
+      "loss": 0.5711,
+      "step": 3219
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 0.000369375,
+      "loss": 0.4885,
+      "step": 3220
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 0.00036932692307692307,
+      "loss": 0.4077,
+      "step": 3221
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 0.00036927884615384617,
+      "loss": 0.4465,
+      "step": 3222
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 0.00036923076923076927,
+      "loss": 0.3541,
+      "step": 3223
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 0.0003691826923076923,
+      "loss": 0.4861,
+      "step": 3224
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 0.0003691346153846154,
+      "loss": 0.4296,
+      "step": 3225
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 0.00036908653846153846,
+      "loss": 0.4516,
+      "step": 3226
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 0.00036903846153846156,
+      "loss": 0.4504,
+      "step": 3227
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 0.00036899038461538466,
+      "loss": 0.4473,
+      "step": 3228
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 0.0003689423076923077,
+      "loss": 0.4453,
+      "step": 3229
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 0.0003688942307692308,
+      "loss": 0.4474,
+      "step": 3230
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 0.00036884615384615385,
+      "loss": 0.5471,
+      "step": 3231
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 0.00036879807692307695,
+      "loss": 0.4423,
+      "step": 3232
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 0.00036875000000000005,
+      "loss": 0.5389,
+      "step": 3233
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 0.0003687019230769231,
+      "loss": 0.5132,
+      "step": 3234
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 0.0003686538461538462,
+      "loss": 0.5147,
+      "step": 3235
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 0.00036860576923076924,
+      "loss": 0.6035,
+      "step": 3236
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 0.00036855769230769234,
+      "loss": 0.4822,
+      "step": 3237
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 0.00036850961538461544,
+      "loss": 0.4314,
+      "step": 3238
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 0.0003684615384615385,
+      "loss": 0.4216,
+      "step": 3239
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 0.0003684134615384616,
+      "loss": 0.4742,
+      "step": 3240
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 0.0003683653846153846,
+      "loss": 0.5307,
+      "step": 3241
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 0.0003683173076923077,
+      "loss": 0.3455,
+      "step": 3242
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 0.0003682692307692308,
+      "loss": 0.592,
+      "step": 3243
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 0.00036822115384615387,
+      "loss": 0.5906,
+      "step": 3244
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 0.0003681730769230769,
+      "loss": 0.675,
+      "step": 3245
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 0.00036812499999999996,
+      "loss": 0.4684,
+      "step": 3246
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 0.00036807692307692306,
+      "loss": 0.6224,
+      "step": 3247
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 0.00036802884615384616,
+      "loss": 0.5054,
+      "step": 3248
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 0.0003679807692307692,
+      "loss": 0.5532,
+      "step": 3249
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 0.0003679326923076923,
+      "loss": 0.5248,
+      "step": 3250
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 0.00036788461538461535,
+      "loss": 0.4164,
+      "step": 3251
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 0.00036783653846153845,
+      "loss": 0.4531,
+      "step": 3252
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 0.00036778846153846155,
+      "loss": 0.5125,
+      "step": 3253
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 0.0003677403846153846,
+      "loss": 0.4258,
+      "step": 3254
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 0.0003676923076923077,
+      "loss": 0.6014,
+      "step": 3255
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 0.00036764423076923074,
+      "loss": 0.511,
+      "step": 3256
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 0.00036759615384615384,
+      "loss": 0.493,
+      "step": 3257
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 0.00036754807692307694,
+      "loss": 0.5219,
+      "step": 3258
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 0.0003675,
+      "loss": 0.5679,
+      "step": 3259
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 0.0003674519230769231,
+      "loss": 0.4906,
+      "step": 3260
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 0.00036740384615384613,
+      "loss": 0.5638,
+      "step": 3261
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 0.00036735576923076923,
+      "loss": 0.6356,
+      "step": 3262
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 0.00036730769230769233,
+      "loss": 0.6161,
+      "step": 3263
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 0.0003672596153846154,
+      "loss": 0.5367,
+      "step": 3264
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 0.0003672115384615385,
+      "loss": 0.4574,
+      "step": 3265
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 0.0003671634615384615,
+      "loss": 0.5066,
+      "step": 3266
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 0.0003671153846153846,
+      "loss": 0.4723,
+      "step": 3267
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 0.00036706730769230767,
+      "loss": 0.5064,
+      "step": 3268
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 0.00036701923076923077,
+      "loss": 0.5463,
+      "step": 3269
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 0.00036697115384615387,
+      "loss": 0.5067,
+      "step": 3270
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 0.0003669230769230769,
+      "loss": 0.5806,
+      "step": 3271
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 0.000366875,
+      "loss": 0.7009,
+      "step": 3272
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 0.00036682692307692306,
+      "loss": 0.6856,
+      "step": 3273
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 0.00036677884615384616,
+      "loss": 0.506,
+      "step": 3274
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 0.00036673076923076926,
+      "loss": 0.6059,
+      "step": 3275
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 0.0003666826923076923,
+      "loss": 0.5625,
+      "step": 3276
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 0.0003666346153846154,
+      "loss": 0.5418,
+      "step": 3277
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 0.00036658653846153845,
+      "loss": 0.5467,
+      "step": 3278
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 0.00036653846153846155,
+      "loss": 0.5745,
+      "step": 3279
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 0.00036649038461538465,
+      "loss": 0.4944,
+      "step": 3280
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 0.0003664423076923077,
+      "loss": 0.4517,
+      "step": 3281
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 0.0003663942307692308,
+      "loss": 0.6255,
+      "step": 3282
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 0.00036634615384615384,
+      "loss": 0.4645,
+      "step": 3283
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 0.00036629807692307694,
+      "loss": 0.5899,
+      "step": 3284
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 0.00036625000000000004,
+      "loss": 0.4265,
+      "step": 3285
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 0.0003662019230769231,
+      "loss": 0.4029,
+      "step": 3286
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 0.0003661538461538462,
+      "loss": 0.5081,
+      "step": 3287
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 0.00036610576923076923,
+      "loss": 0.5135,
+      "step": 3288
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 0.00036605769230769233,
+      "loss": 0.5372,
+      "step": 3289
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 0.00036600961538461543,
+      "loss": 0.5346,
+      "step": 3290
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 0.0003659615384615385,
+      "loss": 0.5657,
+      "step": 3291
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 0.0003659134615384616,
+      "loss": 0.4351,
+      "step": 3292
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 0.0003658653846153846,
+      "loss": 0.4897,
+      "step": 3293
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 0.0003658173076923077,
+      "loss": 0.543,
+      "step": 3294
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 0.0003657692307692308,
+      "loss": 0.4672,
+      "step": 3295
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 0.00036572115384615387,
+      "loss": 0.4594,
+      "step": 3296
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 0.00036567307692307696,
+      "loss": 0.5149,
+      "step": 3297
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 0.000365625,
+      "loss": 0.6996,
+      "step": 3298
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 0.0003655769230769231,
+      "loss": 0.4532,
+      "step": 3299
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 0.0003655288461538462,
+      "loss": 0.5036,
+      "step": 3300
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 0.0003654807692307692,
+      "loss": 0.5375,
+      "step": 3301
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 0.0003654326923076923,
+      "loss": 0.4988,
+      "step": 3302
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 0.00036538461538461535,
+      "loss": 0.5011,
+      "step": 3303
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 0.00036533653846153845,
+      "loss": 0.4558,
+      "step": 3304
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 0.00036528846153846155,
+      "loss": 0.4985,
+      "step": 3305
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 0.0003652403846153846,
+      "loss": 0.513,
+      "step": 3306
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 0.0003651923076923077,
+      "loss": 0.6212,
+      "step": 3307
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 0.00036514423076923074,
+      "loss": 0.5318,
+      "step": 3308
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 0.00036509615384615384,
+      "loss": 0.4311,
+      "step": 3309
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 0.00036504807692307694,
+      "loss": 0.4045,
+      "step": 3310
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 0.000365,
+      "loss": 0.5646,
+      "step": 3311
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 0.0003649519230769231,
+      "loss": 0.5052,
+      "step": 3312
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 0.0003649038461538461,
+      "loss": 0.6251,
+      "step": 3313
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 0.0003648557692307692,
+      "loss": 0.5232,
+      "step": 3314
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 0.0003648076923076923,
+      "loss": 0.4686,
+      "step": 3315
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 0.00036475961538461537,
+      "loss": 0.432,
+      "step": 3316
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 0.00036471153846153847,
+      "loss": 0.4236,
+      "step": 3317
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 0.0003646634615384615,
+      "loss": 0.4782,
+      "step": 3318
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 0.0003646153846153846,
+      "loss": 0.5189,
+      "step": 3319
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 0.0003645673076923077,
+      "loss": 0.4331,
+      "step": 3320
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 0.00036451923076923076,
+      "loss": 0.5305,
+      "step": 3321
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 0.00036447115384615386,
+      "loss": 0.4955,
+      "step": 3322
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 0.0003644230769230769,
+      "loss": 0.4303,
+      "step": 3323
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 0.000364375,
+      "loss": 0.5342,
+      "step": 3324
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 0.0003643269230769231,
+      "loss": 0.6115,
+      "step": 3325
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 0.00036427884615384615,
+      "loss": 0.5742,
+      "step": 3326
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 0.00036423076923076925,
+      "loss": 0.6027,
+      "step": 3327
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 0.0003641826923076923,
+      "loss": 0.5557,
+      "step": 3328
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 0.0003641346153846154,
+      "loss": 0.5972,
+      "step": 3329
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 0.0003640865384615385,
+      "loss": 0.5103,
+      "step": 3330
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 0.00036403846153846154,
+      "loss": 0.5578,
+      "step": 3331
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 0.00036399038461538464,
+      "loss": 0.4896,
+      "step": 3332
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 0.0003639423076923077,
+      "loss": 0.5634,
+      "step": 3333
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 0.0003638942307692308,
+      "loss": 0.5666,
+      "step": 3334
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 0.00036384615384615383,
+      "loss": 0.4851,
+      "step": 3335
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 0.00036379807692307693,
+      "loss": 0.48,
+      "step": 3336
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 0.00036375000000000003,
+      "loss": 0.5031,
+      "step": 3337
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 0.0003637019230769231,
+      "loss": 0.4489,
+      "step": 3338
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 0.0003636538461538462,
+      "loss": 0.6061,
+      "step": 3339
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 0.0003636057692307692,
+      "loss": 0.5368,
+      "step": 3340
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 0.0003635576923076923,
+      "loss": 0.5985,
+      "step": 3341
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 0.0003635096153846154,
+      "loss": 0.5099,
+      "step": 3342
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 0.00036346153846153847,
+      "loss": 0.5253,
+      "step": 3343
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 0.00036341346153846157,
+      "loss": 0.5691,
+      "step": 3344
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 0.0003633653846153846,
+      "loss": 0.4706,
+      "step": 3345
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 0.0003633173076923077,
+      "loss": 0.402,
+      "step": 3346
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 0.0003632692307692308,
+      "loss": 0.5065,
+      "step": 3347
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 0.00036322115384615386,
+      "loss": 0.4627,
+      "step": 3348
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 0.00036317307692307696,
+      "loss": 0.6781,
+      "step": 3349
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 0.000363125,
+      "loss": 0.5084,
+      "step": 3350
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 0.0003630769230769231,
+      "loss": 0.4529,
+      "step": 3351
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 0.0003630288461538462,
+      "loss": 0.6949,
+      "step": 3352
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 0.00036298076923076925,
+      "loss": 0.6996,
+      "step": 3353
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 0.00036293269230769235,
+      "loss": 0.5598,
+      "step": 3354
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 0.0003628846153846154,
+      "loss": 0.5699,
+      "step": 3355
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 0.0003628365384615385,
+      "loss": 0.5375,
+      "step": 3356
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 0.0003627884615384616,
+      "loss": 0.4572,
+      "step": 3357
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 0.0003627403846153846,
+      "loss": 0.5688,
+      "step": 3358
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 0.0003626923076923077,
+      "loss": 0.5178,
+      "step": 3359
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 0.00036264423076923073,
+      "loss": 0.4774,
+      "step": 3360
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 0.00036259615384615383,
+      "loss": 0.4991,
+      "step": 3361
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 0.00036254807692307693,
+      "loss": 0.5354,
+      "step": 3362
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 0.0003625,
+      "loss": 0.5378,
+      "step": 3363
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 0.0003624519230769231,
+      "loss": 0.5038,
+      "step": 3364
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 0.0003624038461538461,
+      "loss": 0.6266,
+      "step": 3365
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 0.0003623557692307692,
+      "loss": 0.6383,
+      "step": 3366
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 0.0003623076923076923,
+      "loss": 0.5966,
+      "step": 3367
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 0.00036225961538461537,
+      "loss": 0.4605,
+      "step": 3368
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 0.00036221153846153846,
+      "loss": 0.4943,
+      "step": 3369
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 0.0003621634615384615,
+      "loss": 0.5835,
+      "step": 3370
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 0.0003621153846153846,
+      "loss": 0.4813,
+      "step": 3371
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 0.0003620673076923077,
+      "loss": 0.4851,
+      "step": 3372
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 0.00036201923076923076,
+      "loss": 0.5125,
+      "step": 3373
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 0.00036197115384615386,
+      "loss": 0.5128,
+      "step": 3374
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 0.0003619230769230769,
+      "loss": 0.4165,
+      "step": 3375
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 0.000361875,
+      "loss": 0.5872,
+      "step": 3376
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 0.0003618269230769231,
+      "loss": 0.5881,
+      "step": 3377
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 0.00036177884615384615,
+      "loss": 0.588,
+      "step": 3378
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 0.00036173076923076925,
+      "loss": 0.5766,
+      "step": 3379
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 0.0003616826923076923,
+      "loss": 0.5433,
+      "step": 3380
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 0.0003616346153846154,
+      "loss": 0.5551,
+      "step": 3381
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 0.0003615865384615385,
+      "loss": 0.528,
+      "step": 3382
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 0.00036153846153846154,
+      "loss": 0.5484,
+      "step": 3383
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 0.00036149038461538464,
+      "loss": 0.4291,
+      "step": 3384
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 0.0003614423076923077,
+      "loss": 0.4127,
+      "step": 3385
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 0.0003613942307692308,
+      "loss": 0.4985,
+      "step": 3386
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 0.0003613461538461539,
+      "loss": 0.3594,
+      "step": 3387
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 0.0003612980769230769,
+      "loss": 0.5082,
+      "step": 3388
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 0.00036125,
+      "loss": 0.3819,
+      "step": 3389
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 0.00036120192307692307,
+      "loss": 0.4771,
+      "step": 3390
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 0.00036115384615384617,
+      "loss": 0.5059,
+      "step": 3391
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 0.00036110576923076927,
+      "loss": 0.4171,
+      "step": 3392
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 0.0003610576923076923,
+      "loss": 0.3276,
+      "step": 3393
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 0.0003610096153846154,
+      "loss": 0.3882,
+      "step": 3394
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 0.00036096153846153846,
+      "loss": 0.4821,
+      "step": 3395
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 0.00036091346153846156,
+      "loss": 0.5495,
+      "step": 3396
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 0.0003608653846153846,
+      "loss": 0.4794,
+      "step": 3397
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 0.0003608173076923077,
+      "loss": 0.5025,
+      "step": 3398
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 0.0003607692307692308,
+      "loss": 0.4442,
+      "step": 3399
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 0.00036072115384615385,
+      "loss": 0.4667,
+      "step": 3400
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 0.00036067307692307695,
+      "loss": 0.3575,
+      "step": 3401
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 0.000360625,
+      "loss": 0.516,
+      "step": 3402
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 0.0003605769230769231,
+      "loss": 0.3574,
+      "step": 3403
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 0.0003605288461538462,
+      "loss": 0.4776,
+      "step": 3404
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 0.00036048076923076924,
+      "loss": 0.5368,
+      "step": 3405
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 0.00036043269230769234,
+      "loss": 0.4062,
+      "step": 3406
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 0.0003603846153846154,
+      "loss": 0.6395,
+      "step": 3407
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 0.0003603365384615385,
+      "loss": 0.6123,
+      "step": 3408
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 0.0003602884615384616,
+      "loss": 0.4305,
+      "step": 3409
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 0.00036024038461538463,
+      "loss": 0.5539,
+      "step": 3410
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 0.00036019230769230773,
+      "loss": 0.4291,
+      "step": 3411
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 0.0003601442307692308,
+      "loss": 0.5123,
+      "step": 3412
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 0.0003600961538461539,
+      "loss": 0.5346,
+      "step": 3413
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 0.000360048076923077,
+      "loss": 0.521,
+      "step": 3414
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 0.4315,
+      "step": 3415
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 0.00035995192307692307,
+      "loss": 0.4309,
+      "step": 3416
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 0.0003599038461538461,
+      "loss": 0.5036,
+      "step": 3417
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 0.0003598557692307692,
+      "loss": 0.4542,
+      "step": 3418
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 0.0003598076923076923,
+      "loss": 0.396,
+      "step": 3419
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 0.00035975961538461536,
+      "loss": 0.4887,
+      "step": 3420
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 0.00035971153846153846,
+      "loss": 0.468,
+      "step": 3421
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 0.0003596634615384615,
+      "loss": 0.5697,
+      "step": 3422
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 0.0003596153846153846,
+      "loss": 0.4744,
+      "step": 3423
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 0.0003595673076923077,
+      "loss": 0.3926,
+      "step": 3424
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 0.00035951923076923075,
+      "loss": 0.3844,
+      "step": 3425
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 0.00035947115384615385,
+      "loss": 0.4447,
+      "step": 3426
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 0.0003594230769230769,
+      "loss": 0.5188,
+      "step": 3427
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 0.000359375,
+      "loss": 0.4315,
+      "step": 3428
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 0.0003593269230769231,
+      "loss": 0.5381,
+      "step": 3429
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 0.00035927884615384614,
+      "loss": 0.3637,
+      "step": 3430
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 0.00035923076923076924,
+      "loss": 0.5969,
+      "step": 3431
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 0.0003591826923076923,
+      "loss": 0.4275,
+      "step": 3432
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 0.0003591346153846154,
+      "loss": 0.5024,
+      "step": 3433
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 0.0003590865384615385,
+      "loss": 0.6289,
+      "step": 3434
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 0.00035903846153846153,
+      "loss": 0.4826,
+      "step": 3435
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 0.00035899038461538463,
+      "loss": 0.5422,
+      "step": 3436
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 0.0003589423076923077,
+      "loss": 0.4723,
+      "step": 3437
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 0.0003588942307692308,
+      "loss": 0.5317,
+      "step": 3438
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 0.0003588461538461539,
+      "loss": 0.4575,
+      "step": 3439
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 0.0003587980769230769,
+      "loss": 0.5786,
+      "step": 3440
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 0.00035875,
+      "loss": 0.4073,
+      "step": 3441
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 0.00035870192307692306,
+      "loss": 0.4238,
+      "step": 3442
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 0.00035865384615384616,
+      "loss": 0.4352,
+      "step": 3443
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 0.00035860576923076926,
+      "loss": 0.4577,
+      "step": 3444
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 0.0003585576923076923,
+      "loss": 0.4239,
+      "step": 3445
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 0.0003585096153846154,
+      "loss": 0.6187,
+      "step": 3446
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 0.00035846153846153846,
+      "loss": 0.5706,
+      "step": 3447
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 0.00035841346153846155,
+      "loss": 0.4341,
+      "step": 3448
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 0.00035836538461538465,
+      "loss": 0.5438,
+      "step": 3449
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 0.0003583173076923077,
+      "loss": 0.5353,
+      "step": 3450
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 0.0003582692307692308,
+      "loss": 0.5817,
+      "step": 3451
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 0.00035822115384615385,
+      "loss": 0.4673,
+      "step": 3452
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 0.00035817307692307695,
+      "loss": 0.4492,
+      "step": 3453
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 0.00035812500000000004,
+      "loss": 0.4807,
+      "step": 3454
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 0.0003580769230769231,
+      "loss": 0.3807,
+      "step": 3455
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 0.0003580288461538462,
+      "loss": 0.4563,
+      "step": 3456
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 0.00035798076923076924,
+      "loss": 0.4402,
+      "step": 3457
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 0.00035793269230769234,
+      "loss": 0.4482,
+      "step": 3458
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 0.00035788461538461544,
+      "loss": 0.5212,
+      "step": 3459
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 0.0003578365384615385,
+      "loss": 0.5563,
+      "step": 3460
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 0.0003577884615384616,
+      "loss": 0.547,
+      "step": 3461
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 0.0003577403846153846,
+      "loss": 0.4996,
+      "step": 3462
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 0.0003576923076923077,
+      "loss": 0.528,
+      "step": 3463
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 0.00035764423076923077,
+      "loss": 0.5301,
+      "step": 3464
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 0.00035759615384615387,
+      "loss": 0.4401,
+      "step": 3465
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 0.00035754807692307697,
+      "loss": 0.4168,
+      "step": 3466
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 0.0003575,
+      "loss": 0.4657,
+      "step": 3467
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 0.0003574519230769231,
+      "loss": 0.4615,
+      "step": 3468
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 0.00035740384615384616,
+      "loss": 0.4839,
+      "step": 3469
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 0.00035735576923076926,
+      "loss": 0.4054,
+      "step": 3470
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 0.0003573076923076923,
+      "loss": 0.3703,
+      "step": 3471
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 0.00035725961538461535,
+      "loss": 0.493,
+      "step": 3472
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 0.00035721153846153845,
+      "loss": 0.4781,
+      "step": 3473
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 0.0003571634615384615,
+      "loss": 0.563,
+      "step": 3474
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 0.0003571153846153846,
+      "loss": 0.5248,
+      "step": 3475
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 0.0003570673076923077,
+      "loss": 0.4082,
+      "step": 3476
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 0.00035701923076923074,
+      "loss": 0.5663,
+      "step": 3477
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 0.00035697115384615384,
+      "loss": 0.4722,
+      "step": 3478
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 0.0003569230769230769,
+      "loss": 0.4856,
+      "step": 3479
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 0.000356875,
+      "loss": 0.4612,
+      "step": 3480
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 0.0003568269230769231,
+      "loss": 0.4172,
+      "step": 3481
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 0.00035677884615384613,
+      "loss": 0.571,
+      "step": 3482
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 0.00035673076923076923,
+      "loss": 0.5332,
+      "step": 3483
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 0.0003566826923076923,
+      "loss": 0.3733,
+      "step": 3484
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 0.0003566346153846154,
+      "loss": 0.4672,
+      "step": 3485
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 0.0003565865384615385,
+      "loss": 0.7036,
+      "step": 3486
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 0.0003565384615384615,
+      "loss": 0.4158,
+      "step": 3487
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 0.0003564903846153846,
+      "loss": 0.4292,
+      "step": 3488
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 0.00035644230769230767,
+      "loss": 0.5793,
+      "step": 3489
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 0.00035639423076923077,
+      "loss": 0.5,
+      "step": 3490
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 0.00035634615384615387,
+      "loss": 0.5033,
+      "step": 3491
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 0.0003562980769230769,
+      "loss": 0.4505,
+      "step": 3492
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 0.00035625,
+      "loss": 0.4561,
+      "step": 3493
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 0.00035620192307692306,
+      "loss": 0.4404,
+      "step": 3494
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 0.00035615384615384616,
+      "loss": 0.4124,
+      "step": 3495
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 0.00035610576923076926,
+      "loss": 0.4179,
+      "step": 3496
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 0.0003560576923076923,
+      "loss": 0.4155,
+      "step": 3497
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 0.0003560096153846154,
+      "loss": 0.5199,
+      "step": 3498
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 0.00035596153846153845,
+      "loss": 0.5162,
+      "step": 3499
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 0.00035591346153846155,
+      "loss": 0.4781,
+      "step": 3500
+    },
+    {
+      "epoch": 32.11,
+      "eval_cer": 0.17351084723908905,
+      "eval_loss": 0.6965524554252625,
+      "eval_runtime": 89.4637,
+      "eval_samples_per_second": 18.41,
+      "eval_steps_per_second": 2.303,
+      "eval_wer": 0.5989173730977428,
+      "step": 3500
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 0.00035586538461538465,
+      "loss": 0.5843,
+      "step": 3501
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 0.0003558173076923077,
+      "loss": 0.4667,
+      "step": 3502
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 0.0003557692307692308,
+      "loss": 0.3938,
+      "step": 3503
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 0.00035572115384615384,
+      "loss": 0.5438,
+      "step": 3504
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 0.00035567307692307694,
+      "loss": 0.456,
+      "step": 3505
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 0.00035562500000000004,
+      "loss": 0.373,
+      "step": 3506
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 0.0003555769230769231,
+      "loss": 0.4442,
+      "step": 3507
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 0.0003555288461538462,
+      "loss": 0.4166,
+      "step": 3508
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 0.00035548076923076923,
+      "loss": 0.3844,
+      "step": 3509
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 0.00035543269230769233,
+      "loss": 0.4296,
+      "step": 3510
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 0.00035538461538461543,
+      "loss": 0.5143,
+      "step": 3511
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 0.0003553365384615385,
+      "loss": 0.4286,
+      "step": 3512
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 0.0003552884615384616,
+      "loss": 0.4011,
+      "step": 3513
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 0.0003552403846153846,
+      "loss": 0.3882,
+      "step": 3514
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 0.0003551923076923077,
+      "loss": 0.5255,
+      "step": 3515
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 0.0003551442307692308,
+      "loss": 0.6588,
+      "step": 3516
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 0.00035509615384615386,
+      "loss": 0.5273,
+      "step": 3517
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 0.00035504807692307696,
+      "loss": 0.5824,
+      "step": 3518
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 0.000355,
+      "loss": 0.5197,
+      "step": 3519
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 0.0003549519230769231,
+      "loss": 0.4813,
+      "step": 3520
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 0.0003549038461538462,
+      "loss": 0.5431,
+      "step": 3521
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 0.00035485576923076925,
+      "loss": 0.5577,
+      "step": 3522
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 0.00035480769230769235,
+      "loss": 0.5225,
+      "step": 3523
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 0.0003547596153846154,
+      "loss": 0.5,
+      "step": 3524
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 0.0003547115384615385,
+      "loss": 0.4945,
+      "step": 3525
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 0.00035466346153846155,
+      "loss": 0.514,
+      "step": 3526
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 0.0003546153846153846,
+      "loss": 0.4908,
+      "step": 3527
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 0.0003545673076923077,
+      "loss": 0.3726,
+      "step": 3528
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 0.00035451923076923074,
+      "loss": 0.5621,
+      "step": 3529
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 0.00035447115384615384,
+      "loss": 0.4352,
+      "step": 3530
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 0.0003544230769230769,
+      "loss": 0.5589,
+      "step": 3531
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 0.000354375,
+      "loss": 0.5564,
+      "step": 3532
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 0.0003543269230769231,
+      "loss": 0.4436,
+      "step": 3533
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 0.0003542788461538461,
+      "loss": 0.5038,
+      "step": 3534
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 0.0003542307692307692,
+      "loss": 0.3713,
+      "step": 3535
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 0.00035418269230769227,
+      "loss": 0.4929,
+      "step": 3536
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 0.00035413461538461537,
+      "loss": 0.4994,
+      "step": 3537
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 0.00035408653846153847,
+      "loss": 0.4408,
+      "step": 3538
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 0.0003540384615384615,
+      "loss": 0.5844,
+      "step": 3539
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 0.0003539903846153846,
+      "loss": 0.5674,
+      "step": 3540
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 0.00035394230769230766,
+      "loss": 0.3865,
+      "step": 3541
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 0.00035389423076923076,
+      "loss": 0.4395,
+      "step": 3542
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 0.00035384615384615386,
+      "loss": 0.5361,
+      "step": 3543
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 0.0003537980769230769,
+      "loss": 0.5026,
+      "step": 3544
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 0.00035375,
+      "loss": 0.5485,
+      "step": 3545
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 0.00035370192307692305,
+      "loss": 0.5846,
+      "step": 3546
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 0.00035365384615384615,
+      "loss": 0.52,
+      "step": 3547
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 0.00035360576923076925,
+      "loss": 0.4609,
+      "step": 3548
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 0.0003535576923076923,
+      "loss": 0.5369,
+      "step": 3549
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 0.0003535096153846154,
+      "loss": 0.4905,
+      "step": 3550
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 0.00035346153846153844,
+      "loss": 0.4206,
+      "step": 3551
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 0.00035341346153846154,
+      "loss": 0.4375,
+      "step": 3552
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 0.00035336538461538464,
+      "loss": 0.4772,
+      "step": 3553
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 0.0003533173076923077,
+      "loss": 0.4139,
+      "step": 3554
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 0.0003532692307692308,
+      "loss": 0.4275,
+      "step": 3555
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 0.00035322115384615383,
+      "loss": 0.5281,
+      "step": 3556
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 0.00035317307692307693,
+      "loss": 0.4338,
+      "step": 3557
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 0.00035312500000000003,
+      "loss": 0.4698,
+      "step": 3558
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 0.0003530769230769231,
+      "loss": 0.4755,
+      "step": 3559
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 0.0003530288461538462,
+      "loss": 0.3759,
+      "step": 3560
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 0.0003529807692307692,
+      "loss": 0.4431,
+      "step": 3561
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 0.0003529326923076923,
+      "loss": 0.4072,
+      "step": 3562
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 0.0003528846153846154,
+      "loss": 0.5281,
+      "step": 3563
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 0.00035283653846153847,
+      "loss": 0.5023,
+      "step": 3564
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 0.00035278846153846157,
+      "loss": 0.4122,
+      "step": 3565
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 0.0003527403846153846,
+      "loss": 0.6147,
+      "step": 3566
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 0.0003526923076923077,
+      "loss": 0.4892,
+      "step": 3567
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 0.0003526442307692308,
+      "loss": 0.5033,
+      "step": 3568
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 0.00035259615384615386,
+      "loss": 0.4493,
+      "step": 3569
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 0.00035254807692307696,
+      "loss": 0.6003,
+      "step": 3570
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 0.0003525,
+      "loss": 0.498,
+      "step": 3571
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 0.0003524519230769231,
+      "loss": 0.5923,
+      "step": 3572
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 0.0003524038461538462,
+      "loss": 0.4785,
+      "step": 3573
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 0.00035235576923076925,
+      "loss": 0.482,
+      "step": 3574
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 0.00035230769230769235,
+      "loss": 0.5551,
+      "step": 3575
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 0.0003522596153846154,
+      "loss": 0.4631,
+      "step": 3576
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 0.0003522115384615385,
+      "loss": 0.5654,
+      "step": 3577
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 0.0003521634615384616,
+      "loss": 0.4541,
+      "step": 3578
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 0.00035211538461538464,
+      "loss": 0.5257,
+      "step": 3579
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 0.00035206730769230774,
+      "loss": 0.5766,
+      "step": 3580
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 0.0003520192307692308,
+      "loss": 0.4801,
+      "step": 3581
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 0.0003519711538461539,
+      "loss": 0.4268,
+      "step": 3582
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 0.000351923076923077,
+      "loss": 0.4131,
+      "step": 3583
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 0.000351875,
+      "loss": 0.4625,
+      "step": 3584
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 0.0003518269230769231,
+      "loss": 0.5356,
+      "step": 3585
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 0.0003517788461538461,
+      "loss": 0.5144,
+      "step": 3586
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 0.0003517307692307692,
+      "loss": 0.3912,
+      "step": 3587
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 0.0003516826923076923,
+      "loss": 0.5382,
+      "step": 3588
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 0.00035163461538461536,
+      "loss": 0.4904,
+      "step": 3589
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 0.00035158653846153846,
+      "loss": 0.4847,
+      "step": 3590
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 0.0003515384615384615,
+      "loss": 0.4675,
+      "step": 3591
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 0.0003514903846153846,
+      "loss": 0.494,
+      "step": 3592
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 0.00035144230769230766,
+      "loss": 0.5194,
+      "step": 3593
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 0.00035139423076923075,
+      "loss": 0.4055,
+      "step": 3594
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 0.00035134615384615385,
+      "loss": 0.4924,
+      "step": 3595
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 0.0003512980769230769,
+      "loss": 0.4576,
+      "step": 3596
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 0.00035125,
+      "loss": 0.4928,
+      "step": 3597
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 0.00035120192307692305,
+      "loss": 0.5307,
+      "step": 3598
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 0.00035115384615384614,
+      "loss": 0.4746,
+      "step": 3599
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 0.00035110576923076924,
+      "loss": 0.4848,
+      "step": 3600
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 0.0003510576923076923,
+      "loss": 0.4494,
+      "step": 3601
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 0.0003510096153846154,
+      "loss": 0.5249,
+      "step": 3602
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 0.00035096153846153844,
+      "loss": 0.3997,
+      "step": 3603
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 0.00035091346153846154,
+      "loss": 0.4574,
+      "step": 3604
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 0.00035086538461538463,
+      "loss": 0.3746,
+      "step": 3605
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 0.0003508173076923077,
+      "loss": 0.529,
+      "step": 3606
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 0.0003507692307692308,
+      "loss": 0.446,
+      "step": 3607
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 0.0003507211538461538,
+      "loss": 0.5099,
+      "step": 3608
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 0.0003506730769230769,
+      "loss": 0.4274,
+      "step": 3609
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 0.000350625,
+      "loss": 0.4476,
+      "step": 3610
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 0.00035057692307692307,
+      "loss": 0.5253,
+      "step": 3611
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 0.00035052884615384617,
+      "loss": 0.4198,
+      "step": 3612
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 0.0003504807692307692,
+      "loss": 0.4921,
+      "step": 3613
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 0.0003504326923076923,
+      "loss": 0.4675,
+      "step": 3614
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 0.0003503846153846154,
+      "loss": 0.4553,
+      "step": 3615
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 0.00035033653846153846,
+      "loss": 0.4694,
+      "step": 3616
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 0.00035028846153846156,
+      "loss": 0.3881,
+      "step": 3617
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 0.0003502403846153846,
+      "loss": 0.4127,
+      "step": 3618
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 0.0003501923076923077,
+      "loss": 0.4652,
+      "step": 3619
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 0.0003501442307692308,
+      "loss": 0.3297,
+      "step": 3620
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 0.00035009615384615385,
+      "loss": 0.569,
+      "step": 3621
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 0.00035004807692307695,
+      "loss": 0.4936,
+      "step": 3622
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 0.00035,
+      "loss": 0.4774,
+      "step": 3623
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 0.0003499519230769231,
+      "loss": 0.5052,
+      "step": 3624
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 0.0003499038461538462,
+      "loss": 0.5954,
+      "step": 3625
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 0.00034985576923076924,
+      "loss": 0.5483,
+      "step": 3626
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 0.00034980769230769234,
+      "loss": 0.4913,
+      "step": 3627
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 0.0003497596153846154,
+      "loss": 0.5434,
+      "step": 3628
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 0.0003497115384615385,
+      "loss": 0.4927,
+      "step": 3629
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 0.0003496634615384616,
+      "loss": 0.4582,
+      "step": 3630
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 0.00034961538461538463,
+      "loss": 0.5545,
+      "step": 3631
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 0.00034956730769230773,
+      "loss": 0.4919,
+      "step": 3632
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 0.0003495192307692308,
+      "loss": 0.5225,
+      "step": 3633
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 0.0003494711538461539,
+      "loss": 0.424,
+      "step": 3634
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 0.000349423076923077,
+      "loss": 0.4298,
+      "step": 3635
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 0.000349375,
+      "loss": 0.4973,
+      "step": 3636
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 0.0003493269230769231,
+      "loss": 0.3653,
+      "step": 3637
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 0.00034927884615384617,
+      "loss": 0.5122,
+      "step": 3638
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 0.00034923076923076927,
+      "loss": 0.4616,
+      "step": 3639
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 0.00034918269230769237,
+      "loss": 0.4057,
+      "step": 3640
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 0.00034913461538461536,
+      "loss": 0.3803,
+      "step": 3641
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 0.00034908653846153846,
+      "loss": 0.4458,
+      "step": 3642
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 0.0003490384615384615,
+      "loss": 0.3756,
+      "step": 3643
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 0.0003489903846153846,
+      "loss": 0.4117,
+      "step": 3644
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 0.0003489423076923077,
+      "loss": 0.4607,
+      "step": 3645
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 0.00034889423076923075,
+      "loss": 0.4876,
+      "step": 3646
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 0.00034884615384615385,
+      "loss": 0.3921,
+      "step": 3647
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 0.0003487980769230769,
+      "loss": 0.5889,
+      "step": 3648
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 0.00034875,
+      "loss": 0.3475,
+      "step": 3649
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 0.0003487019230769231,
+      "loss": 0.4399,
+      "step": 3650
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 0.00034865384615384614,
+      "loss": 0.3443,
+      "step": 3651
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 0.00034860576923076924,
+      "loss": 0.6044,
+      "step": 3652
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 0.0003485576923076923,
+      "loss": 0.5192,
+      "step": 3653
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 0.0003485096153846154,
+      "loss": 0.5377,
+      "step": 3654
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 0.0003484615384615385,
+      "loss": 0.5165,
+      "step": 3655
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 0.00034841346153846153,
+      "loss": 0.4434,
+      "step": 3656
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 0.00034836538461538463,
+      "loss": 0.46,
+      "step": 3657
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 0.0003483173076923077,
+      "loss": 0.4831,
+      "step": 3658
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 0.0003482692307692308,
+      "loss": 0.4862,
+      "step": 3659
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 0.0003482211538461538,
+      "loss": 0.42,
+      "step": 3660
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 0.0003481730769230769,
+      "loss": 0.4182,
+      "step": 3661
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 0.000348125,
+      "loss": 0.5244,
+      "step": 3662
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 0.00034807692307692306,
+      "loss": 0.4594,
+      "step": 3663
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 0.00034802884615384616,
+      "loss": 0.4234,
+      "step": 3664
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 0.0003479807692307692,
+      "loss": 0.4239,
+      "step": 3665
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 0.0003479326923076923,
+      "loss": 0.338,
+      "step": 3666
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 0.0003478846153846154,
+      "loss": 0.3338,
+      "step": 3667
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 0.00034783653846153845,
+      "loss": 0.4481,
+      "step": 3668
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 0.00034778846153846155,
+      "loss": 0.4074,
+      "step": 3669
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 0.0003477403846153846,
+      "loss": 0.3318,
+      "step": 3670
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 0.0003476923076923077,
+      "loss": 0.5234,
+      "step": 3671
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 0.0003476442307692308,
+      "loss": 0.4531,
+      "step": 3672
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 0.00034759615384615384,
+      "loss": 0.4215,
+      "step": 3673
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 0.00034754807692307694,
+      "loss": 0.2669,
+      "step": 3674
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 0.0003475,
+      "loss": 0.4251,
+      "step": 3675
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 0.0003474519230769231,
+      "loss": 0.4842,
+      "step": 3676
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 0.0003474038461538462,
+      "loss": 0.3683,
+      "step": 3677
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 0.00034735576923076923,
+      "loss": 0.3202,
+      "step": 3678
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 0.00034730769230769233,
+      "loss": 0.5845,
+      "step": 3679
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 0.0003472596153846154,
+      "loss": 0.6117,
+      "step": 3680
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 0.0003472115384615385,
+      "loss": 0.5708,
+      "step": 3681
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 0.0003471634615384616,
+      "loss": 0.6129,
+      "step": 3682
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 0.0003471153846153846,
+      "loss": 0.4908,
+      "step": 3683
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 0.0003470673076923077,
+      "loss": 0.568,
+      "step": 3684
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 0.00034701923076923077,
+      "loss": 0.472,
+      "step": 3685
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 0.00034697115384615387,
+      "loss": 0.5323,
+      "step": 3686
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 0.00034692307692307697,
+      "loss": 0.4919,
+      "step": 3687
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 0.000346875,
+      "loss": 0.416,
+      "step": 3688
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 0.0003468269230769231,
+      "loss": 0.3948,
+      "step": 3689
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 0.00034677884615384616,
+      "loss": 0.4564,
+      "step": 3690
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 0.00034673076923076926,
+      "loss": 0.4475,
+      "step": 3691
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 0.00034668269230769236,
+      "loss": 0.4672,
+      "step": 3692
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 0.0003466346153846154,
+      "loss": 0.404,
+      "step": 3693
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 0.0003465865384615385,
+      "loss": 0.3915,
+      "step": 3694
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 0.00034653846153846155,
+      "loss": 0.4091,
+      "step": 3695
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 0.00034649038461538465,
+      "loss": 0.389,
+      "step": 3696
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 0.00034644230769230775,
+      "loss": 0.4457,
+      "step": 3697
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 0.00034639423076923074,
+      "loss": 0.3827,
+      "step": 3698
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 0.00034634615384615384,
+      "loss": 0.4269,
+      "step": 3699
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 0.0003462980769230769,
+      "loss": 0.5298,
+      "step": 3700
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 0.00034625,
+      "loss": 0.4524,
+      "step": 3701
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 0.0003462019230769231,
+      "loss": 0.4332,
+      "step": 3702
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 0.00034615384615384613,
+      "loss": 0.5838,
+      "step": 3703
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 0.00034610576923076923,
+      "loss": 0.5007,
+      "step": 3704
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 0.0003460576923076923,
+      "loss": 0.5765,
+      "step": 3705
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 0.0003460096153846154,
+      "loss": 0.5138,
+      "step": 3706
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 0.0003459615384615385,
+      "loss": 0.6479,
+      "step": 3707
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 0.0003459134615384615,
+      "loss": 0.473,
+      "step": 3708
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 0.0003458653846153846,
+      "loss": 0.5221,
+      "step": 3709
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 0.00034581730769230767,
+      "loss": 0.4964,
+      "step": 3710
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 0.00034576923076923077,
+      "loss": 0.6002,
+      "step": 3711
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 0.00034572115384615387,
+      "loss": 0.4785,
+      "step": 3712
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 0.0003456730769230769,
+      "loss": 0.5443,
+      "step": 3713
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 0.000345625,
+      "loss": 0.498,
+      "step": 3714
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 0.00034557692307692306,
+      "loss": 0.435,
+      "step": 3715
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 0.00034552884615384616,
+      "loss": 0.394,
+      "step": 3716
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 0.00034548076923076926,
+      "loss": 0.444,
+      "step": 3717
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 0.0003454326923076923,
+      "loss": 0.4347,
+      "step": 3718
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 0.0003453846153846154,
+      "loss": 0.4626,
+      "step": 3719
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 0.00034533653846153845,
+      "loss": 0.3632,
+      "step": 3720
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 0.00034528846153846155,
+      "loss": 0.5351,
+      "step": 3721
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 0.0003452403846153846,
+      "loss": 0.3152,
+      "step": 3722
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 0.0003451923076923077,
+      "loss": 0.3816,
+      "step": 3723
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 0.0003451442307692308,
+      "loss": 0.4241,
+      "step": 3724
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 0.00034509615384615384,
+      "loss": 0.5663,
+      "step": 3725
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 0.00034504807692307694,
+      "loss": 0.4424,
+      "step": 3726
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 0.000345,
+      "loss": 0.3819,
+      "step": 3727
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 0.0003449519230769231,
+      "loss": 0.3989,
+      "step": 3728
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 0.0003449038461538462,
+      "loss": 0.4891,
+      "step": 3729
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 0.00034485576923076923,
+      "loss": 0.4698,
+      "step": 3730
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 0.00034480769230769233,
+      "loss": 0.4884,
+      "step": 3731
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 0.0003447596153846154,
+      "loss": 0.627,
+      "step": 3732
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 0.0003447115384615385,
+      "loss": 0.5152,
+      "step": 3733
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 0.0003446634615384616,
+      "loss": 0.4629,
+      "step": 3734
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 0.0003446153846153846,
+      "loss": 0.5429,
+      "step": 3735
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 0.0003445673076923077,
+      "loss": 0.4672,
+      "step": 3736
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 0.00034451923076923076,
+      "loss": 0.4682,
+      "step": 3737
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 0.00034447115384615386,
+      "loss": 0.4505,
+      "step": 3738
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 0.00034442307692307696,
+      "loss": 0.5612,
+      "step": 3739
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 0.000344375,
+      "loss": 0.4218,
+      "step": 3740
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 0.0003443269230769231,
+      "loss": 0.3543,
+      "step": 3741
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 0.00034427884615384615,
+      "loss": 0.4158,
+      "step": 3742
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 0.00034423076923076925,
+      "loss": 0.4226,
+      "step": 3743
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 0.00034418269230769235,
+      "loss": 0.4595,
+      "step": 3744
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 0.0003441346153846154,
+      "loss": 0.4329,
+      "step": 3745
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 0.0003440865384615385,
+      "loss": 0.44,
+      "step": 3746
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 0.00034403846153846154,
+      "loss": 0.4284,
+      "step": 3747
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 0.00034399038461538464,
+      "loss": 0.41,
+      "step": 3748
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 0.00034394230769230774,
+      "loss": 0.5913,
+      "step": 3749
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 0.0003438942307692308,
+      "loss": 0.4561,
+      "step": 3750
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 0.0003438461538461539,
+      "loss": 0.3666,
+      "step": 3751
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 0.00034379807692307693,
+      "loss": 0.4521,
+      "step": 3752
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 0.00034375,
+      "loss": 0.5001,
+      "step": 3753
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 0.0003437019230769231,
+      "loss": 0.5463,
+      "step": 3754
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 0.0003436538461538461,
+      "loss": 0.3544,
+      "step": 3755
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 0.0003436057692307692,
+      "loss": 0.298,
+      "step": 3756
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 0.00034355769230769227,
+      "loss": 0.458,
+      "step": 3757
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 0.00034350961538461537,
+      "loss": 0.4374,
+      "step": 3758
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 0.00034346153846153847,
+      "loss": 0.5389,
+      "step": 3759
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 0.0003434134615384615,
+      "loss": 0.363,
+      "step": 3760
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 0.0003433653846153846,
+      "loss": 0.5834,
+      "step": 3761
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 0.00034331730769230766,
+      "loss": 0.4674,
+      "step": 3762
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 0.00034326923076923076,
+      "loss": 0.5521,
+      "step": 3763
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 0.00034322115384615386,
+      "loss": 0.5103,
+      "step": 3764
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 0.0003431730769230769,
+      "loss": 0.381,
+      "step": 3765
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 0.000343125,
+      "loss": 0.4991,
+      "step": 3766
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 0.00034307692307692305,
+      "loss": 0.4708,
+      "step": 3767
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 0.00034302884615384615,
+      "loss": 0.4654,
+      "step": 3768
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 0.00034298076923076925,
+      "loss": 0.3949,
+      "step": 3769
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 0.0003429326923076923,
+      "loss": 0.4461,
+      "step": 3770
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 0.0003428846153846154,
+      "loss": 0.477,
+      "step": 3771
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 0.00034283653846153844,
+      "loss": 0.5352,
+      "step": 3772
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 0.00034278846153846154,
+      "loss": 0.5034,
+      "step": 3773
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 0.00034274038461538464,
+      "loss": 0.5143,
+      "step": 3774
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 0.0003426923076923077,
+      "loss": 0.4236,
+      "step": 3775
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 0.0003426442307692308,
+      "loss": 0.387,
+      "step": 3776
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 0.00034259615384615383,
+      "loss": 0.3683,
+      "step": 3777
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 0.00034254807692307693,
+      "loss": 0.4712,
+      "step": 3778
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 0.00034250000000000003,
+      "loss": 0.4177,
+      "step": 3779
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 0.0003424519230769231,
+      "loss": 0.5619,
+      "step": 3780
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 0.0003424038461538462,
+      "loss": 0.3566,
+      "step": 3781
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 0.0003423557692307692,
+      "loss": 0.3628,
+      "step": 3782
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 0.0003423076923076923,
+      "loss": 0.3863,
+      "step": 3783
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 0.0003422596153846154,
+      "loss": 0.3447,
+      "step": 3784
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 0.00034221153846153847,
+      "loss": 0.3854,
+      "step": 3785
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 0.00034216346153846157,
+      "loss": 0.4653,
+      "step": 3786
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 0.0003421153846153846,
+      "loss": 0.3198,
+      "step": 3787
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 0.0003420673076923077,
+      "loss": 0.5011,
+      "step": 3788
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 0.00034201923076923076,
+      "loss": 0.4532,
+      "step": 3789
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 0.00034197115384615386,
+      "loss": 0.4505,
+      "step": 3790
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 0.00034192307692307696,
+      "loss": 0.4629,
+      "step": 3791
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 0.000341875,
+      "loss": 0.4956,
+      "step": 3792
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 0.0003418269230769231,
+      "loss": 0.4301,
+      "step": 3793
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 0.00034177884615384615,
+      "loss": 0.4351,
+      "step": 3794
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 0.00034173076923076925,
+      "loss": 0.4813,
+      "step": 3795
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 0.00034168269230769235,
+      "loss": 0.4481,
+      "step": 3796
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 0.0003416346153846154,
+      "loss": 0.3514,
+      "step": 3797
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 0.0003415865384615385,
+      "loss": 0.4264,
+      "step": 3798
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 0.00034153846153846154,
+      "loss": 0.4844,
+      "step": 3799
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 0.00034149038461538464,
+      "loss": 0.4321,
+      "step": 3800
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 0.00034144230769230774,
+      "loss": 0.4431,
+      "step": 3801
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 0.0003413942307692308,
+      "loss": 0.377,
+      "step": 3802
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 0.0003413461538461539,
+      "loss": 0.4475,
+      "step": 3803
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 0.00034129807692307693,
+      "loss": 0.5601,
+      "step": 3804
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 0.00034125000000000003,
+      "loss": 0.5937,
+      "step": 3805
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 0.00034120192307692313,
+      "loss": 0.4545,
+      "step": 3806
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 0.0003411538461538462,
+      "loss": 0.4721,
+      "step": 3807
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 0.00034110576923076927,
+      "loss": 0.4801,
+      "step": 3808
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 0.00034105769230769226,
+      "loss": 0.497,
+      "step": 3809
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 0.00034100961538461536,
+      "loss": 0.6259,
+      "step": 3810
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 0.00034096153846153846,
+      "loss": 0.3706,
+      "step": 3811
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 0.0003409134615384615,
+      "loss": 0.3819,
+      "step": 3812
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 0.0003408653846153846,
+      "loss": 0.5379,
+      "step": 3813
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 0.00034081730769230765,
+      "loss": 0.4045,
+      "step": 3814
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 0.00034076923076923075,
+      "loss": 0.4504,
+      "step": 3815
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 0.00034072115384615385,
+      "loss": 0.5654,
+      "step": 3816
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 0.0003406730769230769,
+      "loss": 0.4405,
+      "step": 3817
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 0.000340625,
+      "loss": 0.5105,
+      "step": 3818
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 0.00034057692307692304,
+      "loss": 0.5194,
+      "step": 3819
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 0.00034052884615384614,
+      "loss": 0.4813,
+      "step": 3820
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 0.00034048076923076924,
+      "loss": 0.4544,
+      "step": 3821
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 0.0003404326923076923,
+      "loss": 0.4221,
+      "step": 3822
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 0.0003403846153846154,
+      "loss": 0.4484,
+      "step": 3823
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 0.00034033653846153843,
+      "loss": 0.4302,
+      "step": 3824
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 0.00034028846153846153,
+      "loss": 0.3791,
+      "step": 3825
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 0.00034024038461538463,
+      "loss": 0.4566,
+      "step": 3826
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 0.0003401923076923077,
+      "loss": 0.3472,
+      "step": 3827
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 0.0003401442307692308,
+      "loss": 0.419,
+      "step": 3828
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 0.0003400961538461538,
+      "loss": 0.3239,
+      "step": 3829
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 0.0003400480769230769,
+      "loss": 0.4976,
+      "step": 3830
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 0.00034,
+      "loss": 0.5178,
+      "step": 3831
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 0.00033995192307692307,
+      "loss": 0.4155,
+      "step": 3832
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 0.00033990384615384617,
+      "loss": 0.4403,
+      "step": 3833
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 0.0003398557692307692,
+      "loss": 0.4671,
+      "step": 3834
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 0.0003398076923076923,
+      "loss": 0.4288,
+      "step": 3835
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 0.0003397596153846154,
+      "loss": 0.4384,
+      "step": 3836
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 0.00033971153846153846,
+      "loss": 0.4519,
+      "step": 3837
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 0.00033966346153846156,
+      "loss": 0.5302,
+      "step": 3838
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 0.0003396153846153846,
+      "loss": 0.2873,
+      "step": 3839
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 0.0003395673076923077,
+      "loss": 0.4651,
+      "step": 3840
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 0.0003395192307692308,
+      "loss": 0.4202,
+      "step": 3841
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 0.00033947115384615385,
+      "loss": 0.3673,
+      "step": 3842
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 0.00033942307692307695,
+      "loss": 0.5643,
+      "step": 3843
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 0.000339375,
+      "loss": 0.4633,
+      "step": 3844
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 0.0003393269230769231,
+      "loss": 0.4299,
+      "step": 3845
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 0.0003392788461538462,
+      "loss": 0.4743,
+      "step": 3846
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 0.00033923076923076924,
+      "loss": 0.4516,
+      "step": 3847
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 0.00033918269230769234,
+      "loss": 0.5331,
+      "step": 3848
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 0.0003391346153846154,
+      "loss": 0.4264,
+      "step": 3849
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 0.0003390865384615385,
+      "loss": 0.5305,
+      "step": 3850
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 0.00033903846153846153,
+      "loss": 0.4938,
+      "step": 3851
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 0.00033899038461538463,
+      "loss": 0.3765,
+      "step": 3852
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 0.00033894230769230773,
+      "loss": 0.4539,
+      "step": 3853
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 0.0003388942307692308,
+      "loss": 0.4647,
+      "step": 3854
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 0.0003388461538461539,
+      "loss": 0.415,
+      "step": 3855
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 0.0003387980769230769,
+      "loss": 0.5063,
+      "step": 3856
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 0.00033875,
+      "loss": 0.539,
+      "step": 3857
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 0.0003387019230769231,
+      "loss": 0.4509,
+      "step": 3858
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 0.00033865384615384617,
+      "loss": 0.5027,
+      "step": 3859
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 0.00033860576923076927,
+      "loss": 0.4279,
+      "step": 3860
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 0.0003385576923076923,
+      "loss": 0.4646,
+      "step": 3861
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 0.0003385096153846154,
+      "loss": 0.3806,
+      "step": 3862
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 0.0003384615384615385,
+      "loss": 0.4576,
+      "step": 3863
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 0.00033841346153846156,
+      "loss": 0.4958,
+      "step": 3864
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 0.00033836538461538466,
+      "loss": 0.3828,
+      "step": 3865
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 0.00033831730769230765,
+      "loss": 0.4931,
+      "step": 3866
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 0.00033826923076923075,
+      "loss": 0.5319,
+      "step": 3867
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 0.00033822115384615385,
+      "loss": 0.386,
+      "step": 3868
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 0.0003381730769230769,
+      "loss": 0.3751,
+      "step": 3869
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 0.000338125,
+      "loss": 0.5135,
+      "step": 3870
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 0.00033807692307692304,
+      "loss": 0.5515,
+      "step": 3871
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 0.00033802884615384614,
+      "loss": 0.5331,
+      "step": 3872
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 0.00033798076923076924,
+      "loss": 0.4587,
+      "step": 3873
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 0.0003379326923076923,
+      "loss": 0.4914,
+      "step": 3874
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 0.0003378846153846154,
+      "loss": 0.4806,
+      "step": 3875
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 0.00033783653846153843,
+      "loss": 0.5286,
+      "step": 3876
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 0.00033778846153846153,
+      "loss": 0.4575,
+      "step": 3877
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 0.00033774038461538463,
+      "loss": 0.4135,
+      "step": 3878
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 0.0003376923076923077,
+      "loss": 0.435,
+      "step": 3879
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 0.00033764423076923077,
+      "loss": 0.4461,
+      "step": 3880
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 0.0003375961538461538,
+      "loss": 0.3924,
+      "step": 3881
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 0.0003375480769230769,
+      "loss": 0.4267,
+      "step": 3882
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 0.0003375,
+      "loss": 0.4311,
+      "step": 3883
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 0.00033745192307692306,
+      "loss": 0.4691,
+      "step": 3884
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 0.00033740384615384616,
+      "loss": 0.4626,
+      "step": 3885
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 0.0003373557692307692,
+      "loss": 0.4495,
+      "step": 3886
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 0.0003373076923076923,
+      "loss": 0.4954,
+      "step": 3887
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 0.0003372596153846154,
+      "loss": 0.4791,
+      "step": 3888
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 0.00033721153846153845,
+      "loss": 0.5002,
+      "step": 3889
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 0.00033716346153846155,
+      "loss": 0.3816,
+      "step": 3890
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 0.0003371153846153846,
+      "loss": 0.3545,
+      "step": 3891
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 0.0003370673076923077,
+      "loss": 0.4919,
+      "step": 3892
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 0.0003370192307692308,
+      "loss": 0.4396,
+      "step": 3893
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 0.00033697115384615384,
+      "loss": 0.4699,
+      "step": 3894
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 0.00033692307692307694,
+      "loss": 0.4649,
+      "step": 3895
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 0.000336875,
+      "loss": 0.479,
+      "step": 3896
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 0.0003368269230769231,
+      "loss": 0.6151,
+      "step": 3897
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 0.0003367788461538462,
+      "loss": 0.5528,
+      "step": 3898
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 0.00033673076923076923,
+      "loss": 0.5335,
+      "step": 3899
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 0.00033668269230769233,
+      "loss": 0.6146,
+      "step": 3900
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 0.0003366346153846154,
+      "loss": 0.3904,
+      "step": 3901
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 0.0003365865384615385,
+      "loss": 0.5622,
+      "step": 3902
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 0.0003365384615384616,
+      "loss": 0.5122,
+      "step": 3903
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 0.0003364903846153846,
+      "loss": 0.4814,
+      "step": 3904
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 0.0003364423076923077,
+      "loss": 0.4589,
+      "step": 3905
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 0.00033639423076923077,
+      "loss": 0.3997,
+      "step": 3906
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 0.00033634615384615387,
+      "loss": 0.559,
+      "step": 3907
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 0.00033629807692307697,
+      "loss": 0.4428,
+      "step": 3908
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 0.00033625,
+      "loss": 0.541,
+      "step": 3909
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 0.0003362019230769231,
+      "loss": 0.5332,
+      "step": 3910
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 0.00033615384615384616,
+      "loss": 0.4463,
+      "step": 3911
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 0.00033610576923076926,
+      "loss": 0.4399,
+      "step": 3912
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 0.00033605769230769236,
+      "loss": 0.4751,
+      "step": 3913
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 0.0003360096153846154,
+      "loss": 0.3688,
+      "step": 3914
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 0.0003359615384615385,
+      "loss": 0.4438,
+      "step": 3915
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 0.00033591346153846155,
+      "loss": 0.4255,
+      "step": 3916
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 0.00033586538461538465,
+      "loss": 0.4152,
+      "step": 3917
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 0.0003358173076923077,
+      "loss": 0.516,
+      "step": 3918
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 0.0003357692307692308,
+      "loss": 0.5637,
+      "step": 3919
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 0.0003357211538461539,
+      "loss": 0.5483,
+      "step": 3920
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 0.00033567307692307694,
+      "loss": 0.4895,
+      "step": 3921
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 0.00033562500000000004,
+      "loss": 0.3971,
+      "step": 3922
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 0.00033557692307692303,
+      "loss": 0.3822,
+      "step": 3923
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 0.00033552884615384613,
+      "loss": 0.5082,
+      "step": 3924
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 0.00033548076923076923,
+      "loss": 0.4289,
+      "step": 3925
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 0.0003354326923076923,
+      "loss": 0.5034,
+      "step": 3926
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 0.0003353846153846154,
+      "loss": 0.4636,
+      "step": 3927
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 0.0003353365384615384,
+      "loss": 0.5371,
+      "step": 3928
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 0.0003352884615384615,
+      "loss": 0.4338,
+      "step": 3929
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 0.0003352403846153846,
+      "loss": 0.5593,
+      "step": 3930
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 0.00033519230769230767,
+      "loss": 0.4283,
+      "step": 3931
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 0.00033514423076923077,
+      "loss": 0.4228,
+      "step": 3932
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 0.0003350961538461538,
+      "loss": 0.4508,
+      "step": 3933
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 0.0003350480769230769,
+      "loss": 0.3849,
+      "step": 3934
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 0.000335,
+      "loss": 0.4784,
+      "step": 3935
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 0.00033495192307692306,
+      "loss": 0.4976,
+      "step": 3936
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 0.00033490384615384616,
+      "loss": 0.4551,
+      "step": 3937
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 0.0003348557692307692,
+      "loss": 0.5462,
+      "step": 3938
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 0.0003348076923076923,
+      "loss": 0.4865,
+      "step": 3939
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 0.0003347596153846154,
+      "loss": 0.533,
+      "step": 3940
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 0.00033471153846153845,
+      "loss": 0.3817,
+      "step": 3941
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 0.00033466346153846155,
+      "loss": 0.419,
+      "step": 3942
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 0.0003346153846153846,
+      "loss": 0.4719,
+      "step": 3943
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 0.0003345673076923077,
+      "loss": 0.4853,
+      "step": 3944
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 0.0003345192307692308,
+      "loss": 0.4579,
+      "step": 3945
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 0.00033447115384615384,
+      "loss": 0.4568,
+      "step": 3946
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 0.00033442307692307694,
+      "loss": 0.4256,
+      "step": 3947
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 0.000334375,
+      "loss": 0.4169,
+      "step": 3948
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 0.0003343269230769231,
+      "loss": 0.4941,
+      "step": 3949
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 0.0003342788461538462,
+      "loss": 0.4732,
+      "step": 3950
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 0.00033423076923076923,
+      "loss": 0.4812,
+      "step": 3951
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 0.00033418269230769233,
+      "loss": 0.6014,
+      "step": 3952
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 0.00033413461538461537,
+      "loss": 0.4385,
+      "step": 3953
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 0.00033408653846153847,
+      "loss": 0.5557,
+      "step": 3954
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 0.00033403846153846157,
+      "loss": 0.4732,
+      "step": 3955
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 0.0003339903846153846,
+      "loss": 0.5547,
+      "step": 3956
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 0.0003339423076923077,
+      "loss": 0.4406,
+      "step": 3957
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 0.00033389423076923076,
+      "loss": 0.4499,
+      "step": 3958
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 0.00033384615384615386,
+      "loss": 0.418,
+      "step": 3959
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 0.00033379807692307696,
+      "loss": 0.4572,
+      "step": 3960
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 0.00033375,
+      "loss": 0.5031,
+      "step": 3961
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 0.0003337019230769231,
+      "loss": 0.6023,
+      "step": 3962
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 0.00033365384615384615,
+      "loss": 0.4245,
+      "step": 3963
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 0.00033360576923076925,
+      "loss": 0.4406,
+      "step": 3964
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 0.00033355769230769235,
+      "loss": 0.5245,
+      "step": 3965
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 0.0003335096153846154,
+      "loss": 0.5341,
+      "step": 3966
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 0.0003334615384615385,
+      "loss": 0.5272,
+      "step": 3967
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 0.00033341346153846154,
+      "loss": 0.4463,
+      "step": 3968
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 0.00033336538461538464,
+      "loss": 0.6207,
+      "step": 3969
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 0.00033331730769230774,
+      "loss": 0.4568,
+      "step": 3970
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 0.0003332692307692308,
+      "loss": 0.4685,
+      "step": 3971
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 0.0003332211538461539,
+      "loss": 0.4312,
+      "step": 3972
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 0.00033317307692307693,
+      "loss": 0.3267,
+      "step": 3973
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 0.00033312500000000003,
+      "loss": 0.3176,
+      "step": 3974
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 0.00033307692307692313,
+      "loss": 0.5674,
+      "step": 3975
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 0.0003330288461538462,
+      "loss": 0.5442,
+      "step": 3976
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 0.0003329807692307693,
+      "loss": 0.3545,
+      "step": 3977
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 0.0003329326923076923,
+      "loss": 0.3829,
+      "step": 3978
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 0.0003328846153846154,
+      "loss": 0.4137,
+      "step": 3979
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 0.00033283653846153847,
+      "loss": 0.5232,
+      "step": 3980
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 0.0003327884615384615,
+      "loss": 0.475,
+      "step": 3981
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 0.0003327403846153846,
+      "loss": 0.439,
+      "step": 3982
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 0.00033269230769230766,
+      "loss": 0.4637,
+      "step": 3983
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 0.00033264423076923076,
+      "loss": 0.4438,
+      "step": 3984
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 0.0003325961538461538,
+      "loss": 0.3535,
+      "step": 3985
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 0.0003325480769230769,
+      "loss": 0.372,
+      "step": 3986
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 0.0003325,
+      "loss": 0.3663,
+      "step": 3987
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 0.00033245192307692305,
+      "loss": 0.4423,
+      "step": 3988
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 0.00033240384615384615,
+      "loss": 0.4314,
+      "step": 3989
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 0.0003323557692307692,
+      "loss": 0.3809,
+      "step": 3990
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 0.0003323076923076923,
+      "loss": 0.4723,
+      "step": 3991
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 0.0003322596153846154,
+      "loss": 0.4836,
+      "step": 3992
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 0.00033221153846153844,
+      "loss": 0.4587,
+      "step": 3993
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 0.00033216346153846154,
+      "loss": 0.5143,
+      "step": 3994
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 0.0003321153846153846,
+      "loss": 0.4639,
+      "step": 3995
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 0.0003320673076923077,
+      "loss": 0.478,
+      "step": 3996
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 0.0003320192307692308,
+      "loss": 0.49,
+      "step": 3997
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 0.00033197115384615383,
+      "loss": 0.4669,
+      "step": 3998
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 0.00033192307692307693,
+      "loss": 0.3945,
+      "step": 3999
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 0.000331875,
+      "loss": 0.5685,
+      "step": 4000
+    },
+    {
+      "epoch": 36.7,
+      "eval_cer": 0.1768692184290619,
+      "eval_loss": 0.6741782426834106,
+      "eval_runtime": 87.3927,
+      "eval_samples_per_second": 18.846,
+      "eval_steps_per_second": 2.357,
+      "eval_wer": 0.5970789500561741,
+      "step": 4000
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 0.0003318269230769231,
+      "loss": 0.3373,
+      "step": 4001
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 0.0003317788461538462,
+      "loss": 0.4397,
+      "step": 4002
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 0.0003317307692307692,
+      "loss": 0.3835,
+      "step": 4003
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 0.0003316826923076923,
+      "loss": 0.3998,
+      "step": 4004
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 0.00033163461538461537,
+      "loss": 0.325,
+      "step": 4005
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 0.00033158653846153847,
+      "loss": 0.575,
+      "step": 4006
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 0.00033153846153846157,
+      "loss": 0.5034,
+      "step": 4007
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 0.0003314903846153846,
+      "loss": 0.5631,
+      "step": 4008
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 0.0003314423076923077,
+      "loss": 0.5477,
+      "step": 4009
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 0.00033139423076923076,
+      "loss": 0.4036,
+      "step": 4010
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 0.00033134615384615386,
+      "loss": 0.5328,
+      "step": 4011
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 0.00033129807692307696,
+      "loss": 0.4346,
+      "step": 4012
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 0.00033125,
+      "loss": 0.4133,
+      "step": 4013
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 0.0003312019230769231,
+      "loss": 0.3977,
+      "step": 4014
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 0.00033115384615384615,
+      "loss": 0.4987,
+      "step": 4015
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 0.00033110576923076925,
+      "loss": 0.4252,
+      "step": 4016
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 0.00033105769230769235,
+      "loss": 0.4333,
+      "step": 4017
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 0.0003310096153846154,
+      "loss": 0.4023,
+      "step": 4018
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 0.0003309615384615385,
+      "loss": 0.4043,
+      "step": 4019
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 0.00033091346153846154,
+      "loss": 0.4601,
+      "step": 4020
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 0.00033086538461538464,
+      "loss": 0.4935,
+      "step": 4021
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 0.00033081730769230774,
+      "loss": 0.425,
+      "step": 4022
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 0.0003307692307692308,
+      "loss": 0.393,
+      "step": 4023
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 0.0003307211538461539,
+      "loss": 0.5754,
+      "step": 4024
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 0.00033067307692307693,
+      "loss": 0.4483,
+      "step": 4025
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 0.000330625,
+      "loss": 0.5291,
+      "step": 4026
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 0.0003305769230769231,
+      "loss": 0.4876,
+      "step": 4027
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 0.00033052884615384617,
+      "loss": 0.4385,
+      "step": 4028
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 0.00033048076923076927,
+      "loss": 0.4908,
+      "step": 4029
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 0.0003304326923076923,
+      "loss": 0.4198,
+      "step": 4030
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 0.0003303846153846154,
+      "loss": 0.4517,
+      "step": 4031
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 0.0003303365384615385,
+      "loss": 0.3266,
+      "step": 4032
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 0.00033028846153846156,
+      "loss": 0.5478,
+      "step": 4033
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 0.00033024038461538466,
+      "loss": 0.6142,
+      "step": 4034
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 0.0003301923076923077,
+      "loss": 0.5927,
+      "step": 4035
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 0.00033014423076923075,
+      "loss": 0.5079,
+      "step": 4036
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 0.00033009615384615385,
+      "loss": 0.5589,
+      "step": 4037
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 0.0003300480769230769,
+      "loss": 0.4199,
+      "step": 4038
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 0.00033,
+      "loss": 0.4388,
+      "step": 4039
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 0.00032995192307692304,
+      "loss": 0.5174,
+      "step": 4040
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 0.00032990384615384614,
+      "loss": 0.3278,
+      "step": 4041
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 0.00032985576923076924,
+      "loss": 0.5056,
+      "step": 4042
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 0.0003298076923076923,
+      "loss": 0.4489,
+      "step": 4043
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 0.0003297596153846154,
+      "loss": 0.4979,
+      "step": 4044
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 0.00032971153846153843,
+      "loss": 0.4235,
+      "step": 4045
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 0.00032966346153846153,
+      "loss": 0.3384,
+      "step": 4046
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 0.0003296153846153846,
+      "loss": 0.4822,
+      "step": 4047
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 0.0003295673076923077,
+      "loss": 0.4175,
+      "step": 4048
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 0.0003295192307692308,
+      "loss": 0.4577,
+      "step": 4049
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 0.0003294711538461538,
+      "loss": 0.4204,
+      "step": 4050
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 0.0003294230769230769,
+      "loss": 0.4724,
+      "step": 4051
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 0.00032937499999999997,
+      "loss": 0.4064,
+      "step": 4052
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 0.00032932692307692307,
+      "loss": 0.4905,
+      "step": 4053
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 0.00032927884615384617,
+      "loss": 0.5541,
+      "step": 4054
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 0.0003292307692307692,
+      "loss": 0.3636,
+      "step": 4055
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 0.0003291826923076923,
+      "loss": 0.4665,
+      "step": 4056
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 0.00032913461538461536,
+      "loss": 0.5157,
+      "step": 4057
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 0.00032908653846153846,
+      "loss": 0.6501,
+      "step": 4058
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 0.00032903846153846156,
+      "loss": 0.4602,
+      "step": 4059
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 0.0003289903846153846,
+      "loss": 0.5822,
+      "step": 4060
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 0.0003289423076923077,
+      "loss": 0.6343,
+      "step": 4061
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 0.00032889423076923075,
+      "loss": 0.5653,
+      "step": 4062
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 0.00032884615384615385,
+      "loss": 0.5959,
+      "step": 4063
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 0.00032879807692307695,
+      "loss": 0.4701,
+      "step": 4064
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 0.00032875,
+      "loss": 0.52,
+      "step": 4065
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 0.0003287019230769231,
+      "loss": 0.4271,
+      "step": 4066
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 0.00032865384615384614,
+      "loss": 0.5013,
+      "step": 4067
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 0.00032860576923076924,
+      "loss": 0.3825,
+      "step": 4068
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 0.00032855769230769234,
+      "loss": 0.4238,
+      "step": 4069
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 0.0003285096153846154,
+      "loss": 0.4789,
+      "step": 4070
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 0.0003284615384615385,
+      "loss": 0.4444,
+      "step": 4071
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 0.00032841346153846153,
+      "loss": 0.5018,
+      "step": 4072
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 0.00032836538461538463,
+      "loss": 0.4506,
+      "step": 4073
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 0.00032831730769230773,
+      "loss": 0.4415,
+      "step": 4074
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 0.0003282692307692308,
+      "loss": 0.4242,
+      "step": 4075
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 0.0003282211538461539,
+      "loss": 0.4799,
+      "step": 4076
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 0.0003281730769230769,
+      "loss": 0.3664,
+      "step": 4077
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 0.000328125,
+      "loss": 0.4847,
+      "step": 4078
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 0.0003280769230769231,
+      "loss": 0.4275,
+      "step": 4079
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 0.00032802884615384617,
+      "loss": 0.4617,
+      "step": 4080
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 0.00032798076923076927,
+      "loss": 0.3394,
+      "step": 4081
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 0.0003279326923076923,
+      "loss": 0.4847,
+      "step": 4082
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 0.0003278846153846154,
+      "loss": 0.4744,
+      "step": 4083
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 0.0003278365384615385,
+      "loss": 0.3824,
+      "step": 4084
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 0.00032778846153846156,
+      "loss": 0.4779,
+      "step": 4085
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 0.00032774038461538466,
+      "loss": 0.4043,
+      "step": 4086
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 0.0003276923076923077,
+      "loss": 0.6063,
+      "step": 4087
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 0.0003276442307692308,
+      "loss": 0.5558,
+      "step": 4088
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 0.0003275961538461539,
+      "loss": 0.5115,
+      "step": 4089
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 0.00032754807692307695,
+      "loss": 0.537,
+      "step": 4090
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 0.00032750000000000005,
+      "loss": 0.5671,
+      "step": 4091
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 0.00032745192307692304,
+      "loss": 0.4358,
+      "step": 4092
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 0.00032740384615384614,
+      "loss": 0.4223,
+      "step": 4093
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 0.00032735576923076924,
+      "loss": 0.5496,
+      "step": 4094
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 0.0003273076923076923,
+      "loss": 0.4218,
+      "step": 4095
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 0.0003272596153846154,
+      "loss": 0.455,
+      "step": 4096
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 0.00032721153846153843,
+      "loss": 0.4466,
+      "step": 4097
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 0.00032716346153846153,
+      "loss": 0.4882,
+      "step": 4098
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 0.0003271153846153846,
+      "loss": 0.4043,
+      "step": 4099
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 0.00032706730769230767,
+      "loss": 0.4105,
+      "step": 4100
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 0.00032701923076923077,
+      "loss": 0.4853,
+      "step": 4101
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 0.0003269711538461538,
+      "loss": 0.5638,
+      "step": 4102
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 0.0003269230769230769,
+      "loss": 0.4784,
+      "step": 4103
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 0.000326875,
+      "loss": 0.439,
+      "step": 4104
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 0.00032682692307692306,
+      "loss": 0.4919,
+      "step": 4105
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 0.00032677884615384616,
+      "loss": 0.4584,
+      "step": 4106
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 0.0003267307692307692,
+      "loss": 0.4648,
+      "step": 4107
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 0.0003266826923076923,
+      "loss": 0.4247,
+      "step": 4108
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 0.0003266346153846154,
+      "loss": 0.3726,
+      "step": 4109
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 0.00032658653846153845,
+      "loss": 0.5592,
+      "step": 4110
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 0.00032653846153846155,
+      "loss": 0.4541,
+      "step": 4111
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 0.0003264903846153846,
+      "loss": 0.5114,
+      "step": 4112
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 0.0003264423076923077,
+      "loss": 0.5379,
+      "step": 4113
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 0.00032639423076923074,
+      "loss": 0.5188,
+      "step": 4114
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 0.00032634615384615384,
+      "loss": 0.5279,
+      "step": 4115
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 0.00032629807692307694,
+      "loss": 0.5163,
+      "step": 4116
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 0.00032625,
+      "loss": 0.4904,
+      "step": 4117
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 0.0003262019230769231,
+      "loss": 0.4074,
+      "step": 4118
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 0.00032615384615384613,
+      "loss": 0.5204,
+      "step": 4119
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 0.00032610576923076923,
+      "loss": 0.489,
+      "step": 4120
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 0.00032605769230769233,
+      "loss": 0.4147,
+      "step": 4121
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 0.0003260096153846154,
+      "loss": 0.5157,
+      "step": 4122
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 0.0003259615384615385,
+      "loss": 0.4418,
+      "step": 4123
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 0.0003259134615384615,
+      "loss": 0.4213,
+      "step": 4124
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 0.0003258653846153846,
+      "loss": 0.432,
+      "step": 4125
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 0.0003258173076923077,
+      "loss": 0.5076,
+      "step": 4126
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 0.00032576923076923077,
+      "loss": 0.4437,
+      "step": 4127
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 0.00032572115384615387,
+      "loss": 0.4104,
+      "step": 4128
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 0.0003256730769230769,
+      "loss": 0.599,
+      "step": 4129
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 0.000325625,
+      "loss": 0.5196,
+      "step": 4130
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 0.0003255769230769231,
+      "loss": 0.3777,
+      "step": 4131
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 0.00032552884615384616,
+      "loss": 0.5383,
+      "step": 4132
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 0.00032548076923076926,
+      "loss": 0.4262,
+      "step": 4133
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 0.0003254326923076923,
+      "loss": 0.4072,
+      "step": 4134
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 0.0003253846153846154,
+      "loss": 0.3492,
+      "step": 4135
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 0.0003253365384615385,
+      "loss": 0.4691,
+      "step": 4136
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 0.00032528846153846155,
+      "loss": 0.3812,
+      "step": 4137
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 0.00032524038461538465,
+      "loss": 0.4363,
+      "step": 4138
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 0.0003251923076923077,
+      "loss": 0.415,
+      "step": 4139
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 0.0003251442307692308,
+      "loss": 0.2483,
+      "step": 4140
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 0.0003250961538461539,
+      "loss": 0.5313,
+      "step": 4141
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 0.00032504807692307694,
+      "loss": 0.3452,
+      "step": 4142
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 0.515,
+      "step": 4143
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 0.0003249519230769231,
+      "loss": 0.4899,
+      "step": 4144
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 0.0003249038461538462,
+      "loss": 0.371,
+      "step": 4145
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 0.0003248557692307693,
+      "loss": 0.4421,
+      "step": 4146
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 0.00032480769230769233,
+      "loss": 0.44,
+      "step": 4147
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 0.00032475961538461543,
+      "loss": 0.4309,
+      "step": 4148
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 0.0003247115384615384,
+      "loss": 0.4188,
+      "step": 4149
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 0.0003246634615384615,
+      "loss": 0.3585,
+      "step": 4150
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 0.0003246153846153846,
+      "loss": 0.3722,
+      "step": 4151
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 0.00032456730769230767,
+      "loss": 0.3929,
+      "step": 4152
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 0.00032451923076923077,
+      "loss": 0.3865,
+      "step": 4153
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 0.0003244711538461538,
+      "loss": 0.3746,
+      "step": 4154
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 0.0003244230769230769,
+      "loss": 0.427,
+      "step": 4155
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 0.000324375,
+      "loss": 0.4119,
+      "step": 4156
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 0.00032432692307692306,
+      "loss": 0.5132,
+      "step": 4157
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 0.00032427884615384616,
+      "loss": 0.4025,
+      "step": 4158
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 0.0003242307692307692,
+      "loss": 0.3765,
+      "step": 4159
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 0.0003241826923076923,
+      "loss": 0.3627,
+      "step": 4160
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 0.0003241346153846154,
+      "loss": 0.4301,
+      "step": 4161
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 0.00032408653846153845,
+      "loss": 0.3299,
+      "step": 4162
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 0.00032403846153846155,
+      "loss": 0.4692,
+      "step": 4163
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 0.0003239903846153846,
+      "loss": 0.4407,
+      "step": 4164
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 0.0003239423076923077,
+      "loss": 0.3864,
+      "step": 4165
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 0.0003238942307692308,
+      "loss": 0.5271,
+      "step": 4166
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 0.00032384615384615384,
+      "loss": 0.3769,
+      "step": 4167
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 0.00032379807692307694,
+      "loss": 0.3746,
+      "step": 4168
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 0.00032375,
+      "loss": 0.2441,
+      "step": 4169
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 0.0003237019230769231,
+      "loss": 0.5475,
+      "step": 4170
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 0.0003236538461538462,
+      "loss": 0.4107,
+      "step": 4171
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 0.0003236057692307692,
+      "loss": 0.4117,
+      "step": 4172
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 0.0003235576923076923,
+      "loss": 0.5026,
+      "step": 4173
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 0.00032350961538461537,
+      "loss": 0.3988,
+      "step": 4174
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 0.00032346153846153847,
+      "loss": 0.5221,
+      "step": 4175
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 0.0003234134615384615,
+      "loss": 0.4409,
+      "step": 4176
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 0.0003233653846153846,
+      "loss": 0.385,
+      "step": 4177
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 0.0003233173076923077,
+      "loss": 0.4415,
+      "step": 4178
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 0.00032326923076923076,
+      "loss": 0.4388,
+      "step": 4179
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 0.00032322115384615386,
+      "loss": 0.4435,
+      "step": 4180
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 0.0003231730769230769,
+      "loss": 0.4371,
+      "step": 4181
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 0.000323125,
+      "loss": 0.4661,
+      "step": 4182
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 0.0003230769230769231,
+      "loss": 0.4203,
+      "step": 4183
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 0.00032302884615384615,
+      "loss": 0.5087,
+      "step": 4184
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 0.00032298076923076925,
+      "loss": 0.4992,
+      "step": 4185
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 0.0003229326923076923,
+      "loss": 0.4027,
+      "step": 4186
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 0.0003228846153846154,
+      "loss": 0.4368,
+      "step": 4187
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 0.0003228365384615385,
+      "loss": 0.5291,
+      "step": 4188
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 0.00032278846153846154,
+      "loss": 0.4123,
+      "step": 4189
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 0.00032274038461538464,
+      "loss": 0.5415,
+      "step": 4190
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 0.0003226923076923077,
+      "loss": 0.5147,
+      "step": 4191
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 0.0003226442307692308,
+      "loss": 0.3739,
+      "step": 4192
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 0.0003225961538461539,
+      "loss": 0.358,
+      "step": 4193
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 0.00032254807692307693,
+      "loss": 0.4781,
+      "step": 4194
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 0.00032250000000000003,
+      "loss": 0.5192,
+      "step": 4195
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 0.0003224519230769231,
+      "loss": 0.5201,
+      "step": 4196
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 0.0003224038461538462,
+      "loss": 0.5494,
+      "step": 4197
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 0.0003223557692307693,
+      "loss": 0.566,
+      "step": 4198
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 0.0003223076923076923,
+      "loss": 0.5707,
+      "step": 4199
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 0.0003222596153846154,
+      "loss": 0.4699,
+      "step": 4200
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 0.00032221153846153847,
+      "loss": 0.4697,
+      "step": 4201
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 0.00032216346153846157,
+      "loss": 0.511,
+      "step": 4202
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 0.00032211538461538467,
+      "loss": 0.5202,
+      "step": 4203
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 0.0003220673076923077,
+      "loss": 0.3494,
+      "step": 4204
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 0.0003220192307692308,
+      "loss": 0.4567,
+      "step": 4205
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 0.0003219711538461538,
+      "loss": 0.4458,
+      "step": 4206
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 0.0003219230769230769,
+      "loss": 0.4439,
+      "step": 4207
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 0.000321875,
+      "loss": 0.412,
+      "step": 4208
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 0.00032182692307692305,
+      "loss": 0.3739,
+      "step": 4209
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 0.00032177884615384615,
+      "loss": 0.355,
+      "step": 4210
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 0.0003217307692307692,
+      "loss": 0.4501,
+      "step": 4211
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 0.0003216826923076923,
+      "loss": 0.4357,
+      "step": 4212
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 0.0003216346153846154,
+      "loss": 0.5065,
+      "step": 4213
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 0.00032158653846153844,
+      "loss": 0.4299,
+      "step": 4214
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 0.00032153846153846154,
+      "loss": 0.4043,
+      "step": 4215
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 0.0003214903846153846,
+      "loss": 0.5249,
+      "step": 4216
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 0.0003214423076923077,
+      "loss": 0.3197,
+      "step": 4217
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 0.0003213942307692308,
+      "loss": 0.457,
+      "step": 4218
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 0.00032134615384615383,
+      "loss": 0.472,
+      "step": 4219
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 0.00032129807692307693,
+      "loss": 0.4471,
+      "step": 4220
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 0.00032125,
+      "loss": 0.4611,
+      "step": 4221
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 0.0003212019230769231,
+      "loss": 0.4841,
+      "step": 4222
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 0.0003211538461538462,
+      "loss": 0.3654,
+      "step": 4223
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 0.0003211057692307692,
+      "loss": 0.4979,
+      "step": 4224
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 0.0003210576923076923,
+      "loss": 0.5268,
+      "step": 4225
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 0.00032100961538461537,
+      "loss": 0.5105,
+      "step": 4226
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 0.00032096153846153847,
+      "loss": 0.3835,
+      "step": 4227
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 0.00032091346153846156,
+      "loss": 0.4692,
+      "step": 4228
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 0.0003208653846153846,
+      "loss": 0.3452,
+      "step": 4229
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 0.0003208173076923077,
+      "loss": 0.4312,
+      "step": 4230
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 0.00032076923076923076,
+      "loss": 0.3863,
+      "step": 4231
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 0.00032072115384615386,
+      "loss": 0.4916,
+      "step": 4232
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 0.00032067307692307696,
+      "loss": 0.4042,
+      "step": 4233
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 0.000320625,
+      "loss": 0.445,
+      "step": 4234
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 0.0003205769230769231,
+      "loss": 0.5019,
+      "step": 4235
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 0.00032052884615384615,
+      "loss": 0.3927,
+      "step": 4236
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 0.00032048076923076925,
+      "loss": 0.3464,
+      "step": 4237
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 0.00032043269230769235,
+      "loss": 0.4469,
+      "step": 4238
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 0.0003203846153846154,
+      "loss": 0.4125,
+      "step": 4239
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 0.0003203365384615385,
+      "loss": 0.3808,
+      "step": 4240
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 0.00032028846153846154,
+      "loss": 0.4365,
+      "step": 4241
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 0.00032024038461538464,
+      "loss": 0.5879,
+      "step": 4242
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 0.0003201923076923077,
+      "loss": 0.4797,
+      "step": 4243
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 0.0003201442307692308,
+      "loss": 0.3592,
+      "step": 4244
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 0.0003200961538461539,
+      "loss": 0.3997,
+      "step": 4245
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 0.0003200480769230769,
+      "loss": 0.3986,
+      "step": 4246
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 0.00032,
+      "loss": 0.479,
+      "step": 4247
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 0.00031995192307692307,
+      "loss": 0.4735,
+      "step": 4248
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 0.00031990384615384617,
+      "loss": 0.4257,
+      "step": 4249
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 0.00031985576923076927,
+      "loss": 0.4978,
+      "step": 4250
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 0.0003198076923076923,
+      "loss": 0.3773,
+      "step": 4251
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 0.0003197596153846154,
+      "loss": 0.5535,
+      "step": 4252
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 0.00031971153846153846,
+      "loss": 0.4524,
+      "step": 4253
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 0.00031966346153846156,
+      "loss": 0.4237,
+      "step": 4254
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 0.00031961538461538466,
+      "loss": 0.5525,
+      "step": 4255
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 0.0003195673076923077,
+      "loss": 0.4077,
+      "step": 4256
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 0.0003195192307692308,
+      "loss": 0.4521,
+      "step": 4257
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 0.00031947115384615385,
+      "loss": 0.4353,
+      "step": 4258
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 0.00031942307692307695,
+      "loss": 0.4798,
+      "step": 4259
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 0.00031937500000000005,
+      "loss": 0.4213,
+      "step": 4260
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 0.0003193269230769231,
+      "loss": 0.5224,
+      "step": 4261
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 0.00031927884615384614,
+      "loss": 0.399,
+      "step": 4262
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 0.0003192307692307692,
+      "loss": 0.413,
+      "step": 4263
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 0.0003191826923076923,
+      "loss": 0.388,
+      "step": 4264
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 0.0003191346153846154,
+      "loss": 0.484,
+      "step": 4265
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 0.00031908653846153843,
+      "loss": 0.4863,
+      "step": 4266
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 0.00031903846153846153,
+      "loss": 0.447,
+      "step": 4267
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 0.0003189903846153846,
+      "loss": 0.3888,
+      "step": 4268
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 0.0003189423076923077,
+      "loss": 0.3777,
+      "step": 4269
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 0.0003188942307692308,
+      "loss": 0.466,
+      "step": 4270
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 0.0003188461538461538,
+      "loss": 0.3773,
+      "step": 4271
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 0.0003187980769230769,
+      "loss": 0.4158,
+      "step": 4272
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 0.00031874999999999997,
+      "loss": 0.3587,
+      "step": 4273
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 0.00031870192307692307,
+      "loss": 0.545,
+      "step": 4274
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 0.00031865384615384617,
+      "loss": 0.4805,
+      "step": 4275
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 0.0003186057692307692,
+      "loss": 0.4773,
+      "step": 4276
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 0.0003185576923076923,
+      "loss": 0.4537,
+      "step": 4277
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 0.00031850961538461536,
+      "loss": 0.2857,
+      "step": 4278
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 0.00031846153846153846,
+      "loss": 0.4218,
+      "step": 4279
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 0.00031841346153846156,
+      "loss": 0.451,
+      "step": 4280
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 0.0003183653846153846,
+      "loss": 0.4903,
+      "step": 4281
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 0.0003183173076923077,
+      "loss": 0.4367,
+      "step": 4282
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 0.00031826923076923075,
+      "loss": 0.3547,
+      "step": 4283
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 0.00031822115384615385,
+      "loss": 0.3493,
+      "step": 4284
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 0.00031817307692307695,
+      "loss": 0.393,
+      "step": 4285
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 0.000318125,
+      "loss": 0.421,
+      "step": 4286
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 0.0003180769230769231,
+      "loss": 0.3657,
+      "step": 4287
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 0.00031802884615384614,
+      "loss": 0.4263,
+      "step": 4288
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 0.00031798076923076924,
+      "loss": 0.4219,
+      "step": 4289
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 0.00031793269230769234,
+      "loss": 0.4026,
+      "step": 4290
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 0.0003178846153846154,
+      "loss": 0.3601,
+      "step": 4291
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 0.0003178365384615385,
+      "loss": 0.3295,
+      "step": 4292
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 0.00031778846153846153,
+      "loss": 0.4895,
+      "step": 4293
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 0.00031774038461538463,
+      "loss": 0.4803,
+      "step": 4294
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 0.00031769230769230773,
+      "loss": 0.4475,
+      "step": 4295
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 0.0003176442307692308,
+      "loss": 0.3979,
+      "step": 4296
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 0.0003175961538461539,
+      "loss": 0.4278,
+      "step": 4297
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 0.0003175480769230769,
+      "loss": 0.4429,
+      "step": 4298
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 0.0003175,
+      "loss": 0.3388,
+      "step": 4299
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 0.0003174519230769231,
+      "loss": 0.3673,
+      "step": 4300
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 0.00031740384615384616,
+      "loss": 0.3715,
+      "step": 4301
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 0.00031735576923076926,
+      "loss": 0.28,
+      "step": 4302
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 0.0003173076923076923,
+      "loss": 0.4246,
+      "step": 4303
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 0.0003172596153846154,
+      "loss": 0.5097,
+      "step": 4304
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 0.0003172115384615385,
+      "loss": 0.2622,
+      "step": 4305
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 0.00031716346153846156,
+      "loss": 0.4449,
+      "step": 4306
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 0.00031711538461538465,
+      "loss": 0.487,
+      "step": 4307
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 0.0003170673076923077,
+      "loss": 0.3633,
+      "step": 4308
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 0.0003170192307692308,
+      "loss": 0.5524,
+      "step": 4309
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 0.00031697115384615385,
+      "loss": 0.3653,
+      "step": 4310
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 0.00031692307692307695,
+      "loss": 0.4685,
+      "step": 4311
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 0.00031687500000000005,
+      "loss": 0.4629,
+      "step": 4312
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 0.0003168269230769231,
+      "loss": 0.4153,
+      "step": 4313
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 0.0003167788461538462,
+      "loss": 0.3919,
+      "step": 4314
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 0.00031673076923076924,
+      "loss": 0.3909,
+      "step": 4315
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 0.00031668269230769234,
+      "loss": 0.4702,
+      "step": 4316
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 0.00031663461538461544,
+      "loss": 0.397,
+      "step": 4317
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 0.0003165865384615384,
+      "loss": 0.4336,
+      "step": 4318
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 0.0003165384615384615,
+      "loss": 0.4116,
+      "step": 4319
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 0.00031649038461538457,
+      "loss": 0.3722,
+      "step": 4320
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 0.00031644230769230767,
+      "loss": 0.3431,
+      "step": 4321
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 0.00031639423076923077,
+      "loss": 0.448,
+      "step": 4322
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 0.0003163461538461538,
+      "loss": 0.4778,
+      "step": 4323
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 0.0003162980769230769,
+      "loss": 0.4572,
+      "step": 4324
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 0.00031624999999999996,
+      "loss": 0.4196,
+      "step": 4325
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 0.00031620192307692306,
+      "loss": 0.4426,
+      "step": 4326
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 0.00031615384615384616,
+      "loss": 0.4003,
+      "step": 4327
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 0.0003161057692307692,
+      "loss": 0.4542,
+      "step": 4328
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 0.0003160576923076923,
+      "loss": 0.3485,
+      "step": 4329
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 0.00031600961538461535,
+      "loss": 0.4251,
+      "step": 4330
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 0.00031596153846153845,
+      "loss": 0.3972,
+      "step": 4331
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 0.00031591346153846155,
+      "loss": 0.3514,
+      "step": 4332
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 0.0003158653846153846,
+      "loss": 0.551,
+      "step": 4333
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 0.0003158173076923077,
+      "loss": 0.4916,
+      "step": 4334
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 0.00031576923076923074,
+      "loss": 0.4341,
+      "step": 4335
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 0.00031572115384615384,
+      "loss": 0.3724,
+      "step": 4336
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 0.00031567307692307694,
+      "loss": 0.4145,
+      "step": 4337
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 0.000315625,
+      "loss": 0.3854,
+      "step": 4338
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 0.0003155769230769231,
+      "loss": 0.4884,
+      "step": 4339
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 0.00031552884615384613,
+      "loss": 0.4601,
+      "step": 4340
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 0.00031548076923076923,
+      "loss": 0.424,
+      "step": 4341
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 0.00031543269230769233,
+      "loss": 0.3778,
+      "step": 4342
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 0.0003153846153846154,
+      "loss": 0.6015,
+      "step": 4343
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 0.0003153365384615385,
+      "loss": 0.4079,
+      "step": 4344
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 0.0003152884615384615,
+      "loss": 0.4408,
+      "step": 4345
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 0.0003152403846153846,
+      "loss": 0.4789,
+      "step": 4346
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 0.0003151923076923077,
+      "loss": 0.4696,
+      "step": 4347
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 0.00031514423076923077,
+      "loss": 0.4821,
+      "step": 4348
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 0.00031509615384615387,
+      "loss": 0.5498,
+      "step": 4349
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 0.0003150480769230769,
+      "loss": 0.3969,
+      "step": 4350
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 0.000315,
+      "loss": 0.4934,
+      "step": 4351
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 0.0003149519230769231,
+      "loss": 0.397,
+      "step": 4352
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 0.00031490384615384616,
+      "loss": 0.4277,
+      "step": 4353
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 0.00031485576923076926,
+      "loss": 0.4473,
+      "step": 4354
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 0.0003148076923076923,
+      "loss": 0.5129,
+      "step": 4355
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 0.0003147596153846154,
+      "loss": 0.4206,
+      "step": 4356
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 0.0003147115384615385,
+      "loss": 0.5043,
+      "step": 4357
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 0.00031466346153846155,
+      "loss": 0.3755,
+      "step": 4358
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 0.00031461538461538465,
+      "loss": 0.3581,
+      "step": 4359
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 0.0003145673076923077,
+      "loss": 0.4577,
+      "step": 4360
+    },
+    {
+      "epoch": 40.01,
+      "learning_rate": 0.0003145192307692308,
+      "loss": 0.4997,
+      "step": 4361
+    },
+    {
+      "epoch": 40.02,
+      "learning_rate": 0.0003144711538461539,
+      "loss": 0.4747,
+      "step": 4362
+    },
+    {
+      "epoch": 40.03,
+      "learning_rate": 0.00031442307692307694,
+      "loss": 0.4682,
+      "step": 4363
+    },
+    {
+      "epoch": 40.04,
+      "learning_rate": 0.00031437500000000004,
+      "loss": 0.4627,
+      "step": 4364
+    },
+    {
+      "epoch": 40.05,
+      "learning_rate": 0.0003143269230769231,
+      "loss": 0.4104,
+      "step": 4365
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 0.0003142788461538462,
+      "loss": 0.3666,
+      "step": 4366
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 0.0003142307692307693,
+      "loss": 0.3648,
+      "step": 4367
+    },
+    {
+      "epoch": 40.07,
+      "learning_rate": 0.00031418269230769233,
+      "loss": 0.4361,
+      "step": 4368
+    },
+    {
+      "epoch": 40.08,
+      "learning_rate": 0.00031413461538461543,
+      "loss": 0.3847,
+      "step": 4369
+    },
+    {
+      "epoch": 40.09,
+      "learning_rate": 0.0003140865384615385,
+      "loss": 0.3743,
+      "step": 4370
+    },
+    {
+      "epoch": 40.1,
+      "learning_rate": 0.0003140384615384616,
+      "loss": 0.3277,
+      "step": 4371
+    },
+    {
+      "epoch": 40.11,
+      "learning_rate": 0.0003139903846153846,
+      "loss": 0.4503,
+      "step": 4372
+    },
+    {
+      "epoch": 40.12,
+      "learning_rate": 0.0003139423076923077,
+      "loss": 0.3855,
+      "step": 4373
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 0.0003138942307692308,
+      "loss": 0.376,
+      "step": 4374
+    },
+    {
+      "epoch": 40.14,
+      "learning_rate": 0.0003138461538461538,
+      "loss": 0.3426,
+      "step": 4375
+    },
+    {
+      "epoch": 40.15,
+      "learning_rate": 0.0003137980769230769,
+      "loss": 0.3894,
+      "step": 4376
+    },
+    {
+      "epoch": 40.16,
+      "learning_rate": 0.00031374999999999996,
+      "loss": 0.3294,
+      "step": 4377
+    },
+    {
+      "epoch": 40.17,
+      "learning_rate": 0.00031370192307692306,
+      "loss": 0.4383,
+      "step": 4378
+    },
+    {
+      "epoch": 40.17,
+      "learning_rate": 0.00031365384615384616,
+      "loss": 0.4334,
+      "step": 4379
+    },
+    {
+      "epoch": 40.18,
+      "learning_rate": 0.0003136057692307692,
+      "loss": 0.3734,
+      "step": 4380
+    },
+    {
+      "epoch": 40.19,
+      "learning_rate": 0.0003135576923076923,
+      "loss": 0.3694,
+      "step": 4381
+    },
+    {
+      "epoch": 40.2,
+      "learning_rate": 0.00031350961538461535,
+      "loss": 0.4256,
+      "step": 4382
+    },
+    {
+      "epoch": 40.21,
+      "learning_rate": 0.00031346153846153845,
+      "loss": 0.4919,
+      "step": 4383
+    },
+    {
+      "epoch": 40.22,
+      "learning_rate": 0.00031341346153846155,
+      "loss": 0.3376,
+      "step": 4384
+    },
+    {
+      "epoch": 40.23,
+      "learning_rate": 0.0003133653846153846,
+      "loss": 0.3711,
+      "step": 4385
+    },
+    {
+      "epoch": 40.24,
+      "learning_rate": 0.0003133173076923077,
+      "loss": 0.3946,
+      "step": 4386
+    },
+    {
+      "epoch": 40.25,
+      "learning_rate": 0.00031326923076923074,
+      "loss": 0.3025,
+      "step": 4387
+    },
+    {
+      "epoch": 40.26,
+      "learning_rate": 0.00031322115384615384,
+      "loss": 0.5628,
+      "step": 4388
+    },
+    {
+      "epoch": 40.27,
+      "learning_rate": 0.00031317307692307694,
+      "loss": 0.5034,
+      "step": 4389
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 0.000313125,
+      "loss": 0.3869,
+      "step": 4390
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 0.0003130769230769231,
+      "loss": 0.4817,
+      "step": 4391
+    },
+    {
+      "epoch": 40.29,
+      "learning_rate": 0.0003130288461538461,
+      "loss": 0.3837,
+      "step": 4392
+    },
+    {
+      "epoch": 40.3,
+      "learning_rate": 0.0003129807692307692,
+      "loss": 0.3566,
+      "step": 4393
+    },
+    {
+      "epoch": 40.31,
+      "learning_rate": 0.0003129326923076923,
+      "loss": 0.4257,
+      "step": 4394
+    },
+    {
+      "epoch": 40.32,
+      "learning_rate": 0.00031288461538461537,
+      "loss": 0.3119,
+      "step": 4395
+    },
+    {
+      "epoch": 40.33,
+      "learning_rate": 0.00031283653846153847,
+      "loss": 0.4176,
+      "step": 4396
+    },
+    {
+      "epoch": 40.34,
+      "learning_rate": 0.0003127884615384615,
+      "loss": 0.3804,
+      "step": 4397
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 0.0003127403846153846,
+      "loss": 0.3288,
+      "step": 4398
+    },
+    {
+      "epoch": 40.36,
+      "learning_rate": 0.0003126923076923077,
+      "loss": 0.3306,
+      "step": 4399
+    },
+    {
+      "epoch": 40.37,
+      "learning_rate": 0.00031264423076923076,
+      "loss": 0.5201,
+      "step": 4400
+    },
+    {
+      "epoch": 40.38,
+      "learning_rate": 0.00031259615384615386,
+      "loss": 0.4682,
+      "step": 4401
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 0.0003125480769230769,
+      "loss": 0.4569,
+      "step": 4402
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 0.0003125,
+      "loss": 0.3256,
+      "step": 4403
+    },
+    {
+      "epoch": 40.4,
+      "learning_rate": 0.0003124519230769231,
+      "loss": 0.3961,
+      "step": 4404
+    },
+    {
+      "epoch": 40.41,
+      "learning_rate": 0.00031240384615384615,
+      "loss": 0.394,
+      "step": 4405
+    },
+    {
+      "epoch": 40.42,
+      "learning_rate": 0.00031235576923076925,
+      "loss": 0.4692,
+      "step": 4406
+    },
+    {
+      "epoch": 40.43,
+      "learning_rate": 0.0003123076923076923,
+      "loss": 0.3276,
+      "step": 4407
+    },
+    {
+      "epoch": 40.44,
+      "learning_rate": 0.0003122596153846154,
+      "loss": 0.417,
+      "step": 4408
+    },
+    {
+      "epoch": 40.45,
+      "learning_rate": 0.0003122115384615385,
+      "loss": 0.4485,
+      "step": 4409
+    },
+    {
+      "epoch": 40.46,
+      "learning_rate": 0.00031216346153846154,
+      "loss": 0.4353,
+      "step": 4410
+    },
+    {
+      "epoch": 40.47,
+      "learning_rate": 0.00031211538461538464,
+      "loss": 0.4768,
+      "step": 4411
+    },
+    {
+      "epoch": 40.48,
+      "learning_rate": 0.0003120673076923077,
+      "loss": 0.3028,
+      "step": 4412
+    },
+    {
+      "epoch": 40.49,
+      "learning_rate": 0.0003120192307692308,
+      "loss": 0.3148,
+      "step": 4413
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 0.0003119711538461539,
+      "loss": 0.2807,
+      "step": 4414
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 0.00031192307692307693,
+      "loss": 0.5019,
+      "step": 4415
+    },
+    {
+      "epoch": 40.51,
+      "learning_rate": 0.00031187500000000003,
+      "loss": 0.4856,
+      "step": 4416
+    },
+    {
+      "epoch": 40.52,
+      "learning_rate": 0.0003118269230769231,
+      "loss": 0.3969,
+      "step": 4417
+    },
+    {
+      "epoch": 40.53,
+      "learning_rate": 0.0003117788461538462,
+      "loss": 0.5088,
+      "step": 4418
+    },
+    {
+      "epoch": 40.54,
+      "learning_rate": 0.0003117307692307693,
+      "loss": 0.4405,
+      "step": 4419
+    },
+    {
+      "epoch": 40.55,
+      "learning_rate": 0.0003116826923076923,
+      "loss": 0.4411,
+      "step": 4420
+    },
+    {
+      "epoch": 40.56,
+      "learning_rate": 0.0003116346153846154,
+      "loss": 0.4547,
+      "step": 4421
+    },
+    {
+      "epoch": 40.57,
+      "learning_rate": 0.00031158653846153847,
+      "loss": 0.3661,
+      "step": 4422
+    },
+    {
+      "epoch": 40.58,
+      "learning_rate": 0.00031153846153846157,
+      "loss": 0.4113,
+      "step": 4423
+    },
+    {
+      "epoch": 40.59,
+      "learning_rate": 0.00031149038461538467,
+      "loss": 0.4656,
+      "step": 4424
+    },
+    {
+      "epoch": 40.6,
+      "learning_rate": 0.0003114423076923077,
+      "loss": 0.3655,
+      "step": 4425
+    },
+    {
+      "epoch": 40.61,
+      "learning_rate": 0.0003113942307692308,
+      "loss": 0.3682,
+      "step": 4426
+    },
+    {
+      "epoch": 40.61,
+      "learning_rate": 0.00031134615384615386,
+      "loss": 0.3552,
+      "step": 4427
+    },
+    {
+      "epoch": 40.62,
+      "learning_rate": 0.00031129807692307696,
+      "loss": 0.4675,
+      "step": 4428
+    },
+    {
+      "epoch": 40.63,
+      "learning_rate": 0.00031125000000000006,
+      "loss": 0.415,
+      "step": 4429
+    },
+    {
+      "epoch": 40.64,
+      "learning_rate": 0.0003112019230769231,
+      "loss": 0.5083,
+      "step": 4430
+    },
+    {
+      "epoch": 40.65,
+      "learning_rate": 0.0003111538461538462,
+      "loss": 0.3984,
+      "step": 4431
+    },
+    {
+      "epoch": 40.66,
+      "learning_rate": 0.0003111057692307692,
+      "loss": 0.4464,
+      "step": 4432
+    },
+    {
+      "epoch": 40.67,
+      "learning_rate": 0.0003110576923076923,
+      "loss": 0.3839,
+      "step": 4433
+    },
+    {
+      "epoch": 40.68,
+      "learning_rate": 0.0003110096153846154,
+      "loss": 0.407,
+      "step": 4434
+    },
+    {
+      "epoch": 40.69,
+      "learning_rate": 0.00031096153846153844,
+      "loss": 0.3533,
+      "step": 4435
+    },
+    {
+      "epoch": 40.7,
+      "learning_rate": 0.00031091346153846154,
+      "loss": 0.4677,
+      "step": 4436
+    },
+    {
+      "epoch": 40.71,
+      "learning_rate": 0.0003108653846153846,
+      "loss": 0.4613,
+      "step": 4437
+    },
+    {
+      "epoch": 40.72,
+      "learning_rate": 0.0003108173076923077,
+      "loss": 0.4984,
+      "step": 4438
+    },
+    {
+      "epoch": 40.72,
+      "learning_rate": 0.00031076923076923073,
+      "loss": 0.364,
+      "step": 4439
+    },
+    {
+      "epoch": 40.73,
+      "learning_rate": 0.00031072115384615383,
+      "loss": 0.3861,
+      "step": 4440
+    },
+    {
+      "epoch": 40.74,
+      "learning_rate": 0.00031067307692307693,
+      "loss": 0.3664,
+      "step": 4441
+    },
+    {
+      "epoch": 40.75,
+      "learning_rate": 0.000310625,
+      "loss": 0.5031,
+      "step": 4442
+    },
+    {
+      "epoch": 40.76,
+      "learning_rate": 0.0003105769230769231,
+      "loss": 0.4005,
+      "step": 4443
+    },
+    {
+      "epoch": 40.77,
+      "learning_rate": 0.0003105288461538461,
+      "loss": 0.5247,
+      "step": 4444
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 0.0003104807692307692,
+      "loss": 0.3551,
+      "step": 4445
+    },
+    {
+      "epoch": 40.79,
+      "learning_rate": 0.0003104326923076923,
+      "loss": 0.3738,
+      "step": 4446
+    },
+    {
+      "epoch": 40.8,
+      "learning_rate": 0.00031038461538461536,
+      "loss": 0.4765,
+      "step": 4447
+    },
+    {
+      "epoch": 40.81,
+      "learning_rate": 0.00031033653846153846,
+      "loss": 0.4764,
+      "step": 4448
+    },
+    {
+      "epoch": 40.82,
+      "learning_rate": 0.0003102884615384615,
+      "loss": 0.4227,
+      "step": 4449
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 0.0003102403846153846,
+      "loss": 0.4365,
+      "step": 4450
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 0.0003101923076923077,
+      "loss": 0.3617,
+      "step": 4451
+    },
+    {
+      "epoch": 40.84,
+      "learning_rate": 0.00031014423076923075,
+      "loss": 0.386,
+      "step": 4452
+    },
+    {
+      "epoch": 40.85,
+      "learning_rate": 0.00031009615384615385,
+      "loss": 0.3314,
+      "step": 4453
+    },
+    {
+      "epoch": 40.86,
+      "learning_rate": 0.0003100480769230769,
+      "loss": 0.4884,
+      "step": 4454
+    },
+    {
+      "epoch": 40.87,
+      "learning_rate": 0.00031,
+      "loss": 0.3785,
+      "step": 4455
+    },
+    {
+      "epoch": 40.88,
+      "learning_rate": 0.0003099519230769231,
+      "loss": 0.4857,
+      "step": 4456
+    },
+    {
+      "epoch": 40.89,
+      "learning_rate": 0.00030990384615384615,
+      "loss": 0.4789,
+      "step": 4457
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 0.00030985576923076924,
+      "loss": 0.5542,
+      "step": 4458
+    },
+    {
+      "epoch": 40.91,
+      "learning_rate": 0.0003098076923076923,
+      "loss": 0.3775,
+      "step": 4459
+    },
+    {
+      "epoch": 40.92,
+      "learning_rate": 0.0003097596153846154,
+      "loss": 0.4659,
+      "step": 4460
+    },
+    {
+      "epoch": 40.93,
+      "learning_rate": 0.0003097115384615385,
+      "loss": 0.4626,
+      "step": 4461
+    },
+    {
+      "epoch": 40.94,
+      "learning_rate": 0.00030966346153846154,
+      "loss": 0.4412,
+      "step": 4462
+    },
+    {
+      "epoch": 40.94,
+      "learning_rate": 0.00030961538461538464,
+      "loss": 0.4236,
+      "step": 4463
+    },
+    {
+      "epoch": 40.95,
+      "learning_rate": 0.0003095673076923077,
+      "loss": 0.3785,
+      "step": 4464
+    },
+    {
+      "epoch": 40.96,
+      "learning_rate": 0.0003095192307692308,
+      "loss": 0.4498,
+      "step": 4465
+    },
+    {
+      "epoch": 40.97,
+      "learning_rate": 0.0003094711538461539,
+      "loss": 0.3332,
+      "step": 4466
+    },
+    {
+      "epoch": 40.98,
+      "learning_rate": 0.0003094230769230769,
+      "loss": 0.3747,
+      "step": 4467
+    },
+    {
+      "epoch": 40.99,
+      "learning_rate": 0.000309375,
+      "loss": 0.4188,
+      "step": 4468
+    },
+    {
+      "epoch": 41.0,
+      "learning_rate": 0.00030932692307692307,
+      "loss": 0.3813,
+      "step": 4469
+    },
+    {
+      "epoch": 41.01,
+      "learning_rate": 0.00030927884615384617,
+      "loss": 0.4473,
+      "step": 4470
+    },
+    {
+      "epoch": 41.02,
+      "learning_rate": 0.00030923076923076927,
+      "loss": 0.4083,
+      "step": 4471
+    },
+    {
+      "epoch": 41.03,
+      "learning_rate": 0.0003091826923076923,
+      "loss": 0.4266,
+      "step": 4472
+    },
+    {
+      "epoch": 41.04,
+      "learning_rate": 0.0003091346153846154,
+      "loss": 0.3817,
+      "step": 4473
+    },
+    {
+      "epoch": 41.05,
+      "learning_rate": 0.00030908653846153846,
+      "loss": 0.4096,
+      "step": 4474
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 0.00030903846153846156,
+      "loss": 0.3883,
+      "step": 4475
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 0.00030899038461538466,
+      "loss": 0.4345,
+      "step": 4476
+    },
+    {
+      "epoch": 41.07,
+      "learning_rate": 0.0003089423076923077,
+      "loss": 0.3815,
+      "step": 4477
+    },
+    {
+      "epoch": 41.08,
+      "learning_rate": 0.0003088942307692308,
+      "loss": 0.3771,
+      "step": 4478
+    },
+    {
+      "epoch": 41.09,
+      "learning_rate": 0.00030884615384615385,
+      "loss": 0.4643,
+      "step": 4479
+    },
+    {
+      "epoch": 41.1,
+      "learning_rate": 0.00030879807692307695,
+      "loss": 0.4114,
+      "step": 4480
+    },
+    {
+      "epoch": 41.11,
+      "learning_rate": 0.00030875000000000005,
+      "loss": 0.3651,
+      "step": 4481
+    },
+    {
+      "epoch": 41.12,
+      "learning_rate": 0.0003087019230769231,
+      "loss": 0.4136,
+      "step": 4482
+    },
+    {
+      "epoch": 41.13,
+      "learning_rate": 0.0003086538461538462,
+      "loss": 0.3855,
+      "step": 4483
+    },
+    {
+      "epoch": 41.14,
+      "learning_rate": 0.00030860576923076924,
+      "loss": 0.3446,
+      "step": 4484
+    },
+    {
+      "epoch": 41.15,
+      "learning_rate": 0.00030855769230769234,
+      "loss": 0.3344,
+      "step": 4485
+    },
+    {
+      "epoch": 41.16,
+      "learning_rate": 0.00030850961538461544,
+      "loss": 0.4033,
+      "step": 4486
+    },
+    {
+      "epoch": 41.17,
+      "learning_rate": 0.0003084615384615385,
+      "loss": 0.3558,
+      "step": 4487
+    },
+    {
+      "epoch": 41.17,
+      "learning_rate": 0.0003084134615384616,
+      "loss": 0.4044,
+      "step": 4488
+    },
+    {
+      "epoch": 41.18,
+      "learning_rate": 0.0003083653846153846,
+      "loss": 0.3794,
+      "step": 4489
+    },
+    {
+      "epoch": 41.19,
+      "learning_rate": 0.0003083173076923077,
+      "loss": 0.3458,
+      "step": 4490
+    },
+    {
+      "epoch": 41.2,
+      "learning_rate": 0.0003082692307692308,
+      "loss": 0.442,
+      "step": 4491
+    },
+    {
+      "epoch": 41.21,
+      "learning_rate": 0.0003082211538461538,
+      "loss": 0.4034,
+      "step": 4492
+    },
+    {
+      "epoch": 41.22,
+      "learning_rate": 0.0003081730769230769,
+      "loss": 0.5666,
+      "step": 4493
+    },
+    {
+      "epoch": 41.23,
+      "learning_rate": 0.00030812499999999997,
+      "loss": 0.4708,
+      "step": 4494
+    },
+    {
+      "epoch": 41.24,
+      "learning_rate": 0.00030807692307692307,
+      "loss": 0.3333,
+      "step": 4495
+    },
+    {
+      "epoch": 41.25,
+      "learning_rate": 0.00030802884615384617,
+      "loss": 0.301,
+      "step": 4496
+    },
+    {
+      "epoch": 41.26,
+      "learning_rate": 0.0003079807692307692,
+      "loss": 0.4879,
+      "step": 4497
+    },
+    {
+      "epoch": 41.27,
+      "learning_rate": 0.0003079326923076923,
+      "loss": 0.5082,
+      "step": 4498
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 0.00030788461538461536,
+      "loss": 0.4475,
+      "step": 4499
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 0.00030783653846153846,
+      "loss": 0.4384,
+      "step": 4500
+    },
+    {
+      "epoch": 41.28,
+      "eval_cer": 0.17673428387232193,
+      "eval_loss": 0.6904355883598328,
+      "eval_runtime": 87.306,
+      "eval_samples_per_second": 18.865,
+      "eval_steps_per_second": 2.36,
+      "eval_wer": 0.5999387192319477,
+      "step": 4500
+    },
+    {
+      "epoch": 41.29,
+      "learning_rate": 0.0003077884615384615,
+      "loss": 0.5041,
+      "step": 4501
+    },
+    {
+      "epoch": 41.3,
+      "learning_rate": 0.0003077403846153846,
+      "loss": 0.4558,
+      "step": 4502
+    },
+    {
+      "epoch": 41.31,
+      "learning_rate": 0.0003076923076923077,
+      "loss": 0.3563,
+      "step": 4503
+    },
+    {
+      "epoch": 41.32,
+      "learning_rate": 0.00030764423076923075,
+      "loss": 0.3274,
+      "step": 4504
+    },
+    {
+      "epoch": 41.33,
+      "learning_rate": 0.00030759615384615385,
+      "loss": 0.3774,
+      "step": 4505
+    },
+    {
+      "epoch": 41.34,
+      "learning_rate": 0.0003075480769230769,
+      "loss": 0.3825,
+      "step": 4506
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 0.0003075,
+      "loss": 0.3727,
+      "step": 4507
+    },
+    {
+      "epoch": 41.36,
+      "learning_rate": 0.0003074519230769231,
+      "loss": 0.4114,
+      "step": 4508
+    },
+    {
+      "epoch": 41.37,
+      "learning_rate": 0.00030740384615384614,
+      "loss": 0.3709,
+      "step": 4509
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 0.00030735576923076924,
+      "loss": 0.5144,
+      "step": 4510
+    },
+    {
+      "epoch": 41.39,
+      "learning_rate": 0.0003073076923076923,
+      "loss": 0.3538,
+      "step": 4511
+    },
+    {
+      "epoch": 41.39,
+      "learning_rate": 0.0003072596153846154,
+      "loss": 0.3495,
+      "step": 4512
+    },
+    {
+      "epoch": 41.4,
+      "learning_rate": 0.0003072115384615385,
+      "loss": 0.3273,
+      "step": 4513
+    },
+    {
+      "epoch": 41.41,
+      "learning_rate": 0.00030716346153846153,
+      "loss": 0.3399,
+      "step": 4514
+    },
+    {
+      "epoch": 41.42,
+      "learning_rate": 0.00030711538461538463,
+      "loss": 0.3498,
+      "step": 4515
+    },
+    {
+      "epoch": 41.43,
+      "learning_rate": 0.0003070673076923077,
+      "loss": 0.5117,
+      "step": 4516
+    },
+    {
+      "epoch": 41.44,
+      "learning_rate": 0.0003070192307692308,
+      "loss": 0.3249,
+      "step": 4517
+    },
+    {
+      "epoch": 41.45,
+      "learning_rate": 0.0003069711538461539,
+      "loss": 0.4333,
+      "step": 4518
+    },
+    {
+      "epoch": 41.46,
+      "learning_rate": 0.0003069230769230769,
+      "loss": 0.4453,
+      "step": 4519
+    },
+    {
+      "epoch": 41.47,
+      "learning_rate": 0.000306875,
+      "loss": 0.3533,
+      "step": 4520
+    },
+    {
+      "epoch": 41.48,
+      "learning_rate": 0.00030682692307692306,
+      "loss": 0.3893,
+      "step": 4521
+    },
+    {
+      "epoch": 41.49,
+      "learning_rate": 0.00030677884615384616,
+      "loss": 0.4303,
+      "step": 4522
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 0.00030673076923076926,
+      "loss": 0.3946,
+      "step": 4523
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 0.0003066826923076923,
+      "loss": 0.4685,
+      "step": 4524
+    },
+    {
+      "epoch": 41.51,
+      "learning_rate": 0.0003066346153846154,
+      "loss": 0.431,
+      "step": 4525
+    },
+    {
+      "epoch": 41.52,
+      "learning_rate": 0.00030658653846153845,
+      "loss": 0.4692,
+      "step": 4526
+    },
+    {
+      "epoch": 41.53,
+      "learning_rate": 0.00030653846153846155,
+      "loss": 0.4961,
+      "step": 4527
+    },
+    {
+      "epoch": 41.54,
+      "learning_rate": 0.00030649038461538465,
+      "loss": 0.3211,
+      "step": 4528
+    },
+    {
+      "epoch": 41.55,
+      "learning_rate": 0.0003064423076923077,
+      "loss": 0.4078,
+      "step": 4529
+    },
+    {
+      "epoch": 41.56,
+      "learning_rate": 0.0003063942307692308,
+      "loss": 0.4799,
+      "step": 4530
+    },
+    {
+      "epoch": 41.57,
+      "learning_rate": 0.00030634615384615384,
+      "loss": 0.3733,
+      "step": 4531
+    },
+    {
+      "epoch": 41.58,
+      "learning_rate": 0.00030629807692307694,
+      "loss": 0.3451,
+      "step": 4532
+    },
+    {
+      "epoch": 41.59,
+      "learning_rate": 0.00030625000000000004,
+      "loss": 0.4838,
+      "step": 4533
+    },
+    {
+      "epoch": 41.6,
+      "learning_rate": 0.0003062019230769231,
+      "loss": 0.4211,
+      "step": 4534
+    },
+    {
+      "epoch": 41.61,
+      "learning_rate": 0.0003061538461538462,
+      "loss": 0.3849,
+      "step": 4535
+    },
+    {
+      "epoch": 41.61,
+      "learning_rate": 0.00030610576923076924,
+      "loss": 0.3692,
+      "step": 4536
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 0.00030605769230769233,
+      "loss": 0.4715,
+      "step": 4537
+    },
+    {
+      "epoch": 41.63,
+      "learning_rate": 0.00030600961538461543,
+      "loss": 0.5336,
+      "step": 4538
+    },
+    {
+      "epoch": 41.64,
+      "learning_rate": 0.0003059615384615385,
+      "loss": 0.35,
+      "step": 4539
+    },
+    {
+      "epoch": 41.65,
+      "learning_rate": 0.0003059134615384616,
+      "loss": 0.4237,
+      "step": 4540
+    },
+    {
+      "epoch": 41.66,
+      "learning_rate": 0.0003058653846153846,
+      "loss": 0.5314,
+      "step": 4541
+    },
+    {
+      "epoch": 41.67,
+      "learning_rate": 0.0003058173076923077,
+      "loss": 0.3945,
+      "step": 4542
+    },
+    {
+      "epoch": 41.68,
+      "learning_rate": 0.0003057692307692308,
+      "loss": 0.3901,
+      "step": 4543
+    },
+    {
+      "epoch": 41.69,
+      "learning_rate": 0.00030572115384615387,
+      "loss": 0.3809,
+      "step": 4544
+    },
+    {
+      "epoch": 41.7,
+      "learning_rate": 0.0003056730769230769,
+      "loss": 0.3378,
+      "step": 4545
+    },
+    {
+      "epoch": 41.71,
+      "learning_rate": 0.00030562499999999996,
+      "loss": 0.4328,
+      "step": 4546
+    },
+    {
+      "epoch": 41.72,
+      "learning_rate": 0.00030557692307692306,
+      "loss": 0.4739,
+      "step": 4547
+    },
+    {
+      "epoch": 41.72,
+      "learning_rate": 0.00030552884615384616,
+      "loss": 0.6172,
+      "step": 4548
+    },
+    {
+      "epoch": 41.73,
+      "learning_rate": 0.0003054807692307692,
+      "loss": 0.412,
+      "step": 4549
+    },
+    {
+      "epoch": 41.74,
+      "learning_rate": 0.0003054326923076923,
+      "loss": 0.357,
+      "step": 4550
+    },
+    {
+      "epoch": 41.75,
+      "learning_rate": 0.00030538461538461535,
+      "loss": 0.4653,
+      "step": 4551
+    },
+    {
+      "epoch": 41.76,
+      "learning_rate": 0.00030533653846153845,
+      "loss": 0.5773,
+      "step": 4552
+    },
+    {
+      "epoch": 41.77,
+      "learning_rate": 0.00030528846153846155,
+      "loss": 0.4481,
+      "step": 4553
+    },
+    {
+      "epoch": 41.78,
+      "learning_rate": 0.0003052403846153846,
+      "loss": 0.3482,
+      "step": 4554
+    },
+    {
+      "epoch": 41.79,
+      "learning_rate": 0.0003051923076923077,
+      "loss": 0.4537,
+      "step": 4555
+    },
+    {
+      "epoch": 41.8,
+      "learning_rate": 0.00030514423076923074,
+      "loss": 0.4384,
+      "step": 4556
+    },
+    {
+      "epoch": 41.81,
+      "learning_rate": 0.00030509615384615384,
+      "loss": 0.3845,
+      "step": 4557
+    },
+    {
+      "epoch": 41.82,
+      "learning_rate": 0.00030504807692307694,
+      "loss": 0.3708,
+      "step": 4558
+    },
+    {
+      "epoch": 41.83,
+      "learning_rate": 0.000305,
+      "loss": 0.4063,
+      "step": 4559
+    },
+    {
+      "epoch": 41.83,
+      "learning_rate": 0.0003049519230769231,
+      "loss": 0.4348,
+      "step": 4560
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 0.00030490384615384613,
+      "loss": 0.4954,
+      "step": 4561
+    },
+    {
+      "epoch": 41.85,
+      "learning_rate": 0.00030485576923076923,
+      "loss": 0.469,
+      "step": 4562
+    },
+    {
+      "epoch": 41.86,
+      "learning_rate": 0.00030480769230769233,
+      "loss": 0.382,
+      "step": 4563
+    },
+    {
+      "epoch": 41.87,
+      "learning_rate": 0.0003047596153846154,
+      "loss": 0.3873,
+      "step": 4564
+    },
+    {
+      "epoch": 41.88,
+      "learning_rate": 0.0003047115384615385,
+      "loss": 0.4004,
+      "step": 4565
+    },
+    {
+      "epoch": 41.89,
+      "learning_rate": 0.0003046634615384615,
+      "loss": 0.3285,
+      "step": 4566
+    },
+    {
+      "epoch": 41.9,
+      "learning_rate": 0.0003046153846153846,
+      "loss": 0.3766,
+      "step": 4567
+    },
+    {
+      "epoch": 41.91,
+      "learning_rate": 0.00030456730769230767,
+      "loss": 0.4152,
+      "step": 4568
+    },
+    {
+      "epoch": 41.92,
+      "learning_rate": 0.00030451923076923077,
+      "loss": 0.4848,
+      "step": 4569
+    },
+    {
+      "epoch": 41.93,
+      "learning_rate": 0.00030447115384615387,
+      "loss": 0.4022,
+      "step": 4570
+    },
+    {
+      "epoch": 41.94,
+      "learning_rate": 0.0003044230769230769,
+      "loss": 0.3398,
+      "step": 4571
+    },
+    {
+      "epoch": 41.94,
+      "learning_rate": 0.000304375,
+      "loss": 0.4275,
+      "step": 4572
+    },
+    {
+      "epoch": 41.95,
+      "learning_rate": 0.00030432692307692306,
+      "loss": 0.2719,
+      "step": 4573
+    },
+    {
+      "epoch": 41.96,
+      "learning_rate": 0.00030427884615384616,
+      "loss": 0.4292,
+      "step": 4574
+    },
+    {
+      "epoch": 41.97,
+      "learning_rate": 0.00030423076923076926,
+      "loss": 0.4552,
+      "step": 4575
+    },
+    {
+      "epoch": 41.98,
+      "learning_rate": 0.0003041826923076923,
+      "loss": 0.4791,
+      "step": 4576
+    },
+    {
+      "epoch": 41.99,
+      "learning_rate": 0.0003041346153846154,
+      "loss": 0.3916,
+      "step": 4577
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 0.00030408653846153845,
+      "loss": 0.4903,
+      "step": 4578
+    },
+    {
+      "epoch": 42.01,
+      "learning_rate": 0.00030403846153846155,
+      "loss": 0.5543,
+      "step": 4579
+    },
+    {
+      "epoch": 42.02,
+      "learning_rate": 0.00030399038461538465,
+      "loss": 0.4049,
+      "step": 4580
+    },
+    {
+      "epoch": 42.03,
+      "learning_rate": 0.0003039423076923077,
+      "loss": 0.3688,
+      "step": 4581
+    },
+    {
+      "epoch": 42.04,
+      "learning_rate": 0.0003038942307692308,
+      "loss": 0.4252,
+      "step": 4582
+    },
+    {
+      "epoch": 42.05,
+      "learning_rate": 0.00030384615384615384,
+      "loss": 0.3208,
+      "step": 4583
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 0.00030379807692307694,
+      "loss": 0.3702,
+      "step": 4584
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 0.00030375000000000004,
+      "loss": 0.4022,
+      "step": 4585
+    },
+    {
+      "epoch": 42.07,
+      "learning_rate": 0.0003037019230769231,
+      "loss": 0.3886,
+      "step": 4586
+    },
+    {
+      "epoch": 42.08,
+      "learning_rate": 0.0003036538461538462,
+      "loss": 0.4528,
+      "step": 4587
+    },
+    {
+      "epoch": 42.09,
+      "learning_rate": 0.00030360576923076923,
+      "loss": 0.408,
+      "step": 4588
+    },
+    {
+      "epoch": 42.1,
+      "learning_rate": 0.00030355769230769233,
+      "loss": 0.3948,
+      "step": 4589
+    },
+    {
+      "epoch": 42.11,
+      "learning_rate": 0.00030350961538461543,
+      "loss": 0.3431,
+      "step": 4590
+    },
+    {
+      "epoch": 42.12,
+      "learning_rate": 0.0003034615384615385,
+      "loss": 0.3976,
+      "step": 4591
+    },
+    {
+      "epoch": 42.13,
+      "learning_rate": 0.0003034134615384616,
+      "loss": 0.3502,
+      "step": 4592
+    },
+    {
+      "epoch": 42.14,
+      "learning_rate": 0.0003033653846153846,
+      "loss": 0.5463,
+      "step": 4593
+    },
+    {
+      "epoch": 42.15,
+      "learning_rate": 0.0003033173076923077,
+      "loss": 0.4091,
+      "step": 4594
+    },
+    {
+      "epoch": 42.16,
+      "learning_rate": 0.0003032692307692308,
+      "loss": 0.4057,
+      "step": 4595
+    },
+    {
+      "epoch": 42.17,
+      "learning_rate": 0.00030322115384615386,
+      "loss": 0.45,
+      "step": 4596
+    },
+    {
+      "epoch": 42.17,
+      "learning_rate": 0.00030317307692307696,
+      "loss": 0.3458,
+      "step": 4597
+    },
+    {
+      "epoch": 42.18,
+      "learning_rate": 0.000303125,
+      "loss": 0.3856,
+      "step": 4598
+    },
+    {
+      "epoch": 42.19,
+      "learning_rate": 0.0003030769230769231,
+      "loss": 0.3553,
+      "step": 4599
+    },
+    {
+      "epoch": 42.2,
+      "learning_rate": 0.0003030288461538462,
+      "loss": 0.363,
+      "step": 4600
+    },
+    {
+      "epoch": 42.21,
+      "learning_rate": 0.0003029807692307692,
+      "loss": 0.4244,
+      "step": 4601
+    },
+    {
+      "epoch": 42.22,
+      "learning_rate": 0.0003029326923076923,
+      "loss": 0.4575,
+      "step": 4602
+    },
+    {
+      "epoch": 42.23,
+      "learning_rate": 0.00030288461538461535,
+      "loss": 0.5037,
+      "step": 4603
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 0.00030283653846153844,
+      "loss": 0.3479,
+      "step": 4604
+    },
+    {
+      "epoch": 42.25,
+      "learning_rate": 0.00030278846153846154,
+      "loss": 0.3239,
+      "step": 4605
+    },
+    {
+      "epoch": 42.26,
+      "learning_rate": 0.0003027403846153846,
+      "loss": 0.4626,
+      "step": 4606
+    },
+    {
+      "epoch": 42.27,
+      "learning_rate": 0.0003026923076923077,
+      "loss": 0.4233,
+      "step": 4607
+    },
+    {
+      "epoch": 42.28,
+      "learning_rate": 0.00030264423076923074,
+      "loss": 0.3963,
+      "step": 4608
+    },
+    {
+      "epoch": 42.28,
+      "learning_rate": 0.00030259615384615384,
+      "loss": 0.484,
+      "step": 4609
+    },
+    {
+      "epoch": 42.29,
+      "learning_rate": 0.00030254807692307693,
+      "loss": 0.3727,
+      "step": 4610
+    },
+    {
+      "epoch": 42.3,
+      "learning_rate": 0.0003025,
+      "loss": 0.4301,
+      "step": 4611
+    },
+    {
+      "epoch": 42.31,
+      "learning_rate": 0.0003024519230769231,
+      "loss": 0.3944,
+      "step": 4612
+    },
+    {
+      "epoch": 42.32,
+      "learning_rate": 0.0003024038461538461,
+      "loss": 0.3655,
+      "step": 4613
+    },
+    {
+      "epoch": 42.33,
+      "learning_rate": 0.0003023557692307692,
+      "loss": 0.4271,
+      "step": 4614
+    },
+    {
+      "epoch": 42.34,
+      "learning_rate": 0.0003023076923076923,
+      "loss": 0.3555,
+      "step": 4615
+    },
+    {
+      "epoch": 42.35,
+      "learning_rate": 0.00030225961538461537,
+      "loss": 0.4868,
+      "step": 4616
+    },
+    {
+      "epoch": 42.36,
+      "learning_rate": 0.00030221153846153847,
+      "loss": 0.3668,
+      "step": 4617
+    },
+    {
+      "epoch": 42.37,
+      "learning_rate": 0.0003021634615384615,
+      "loss": 0.3451,
+      "step": 4618
+    },
+    {
+      "epoch": 42.38,
+      "learning_rate": 0.0003021153846153846,
+      "loss": 0.4495,
+      "step": 4619
+    },
+    {
+      "epoch": 42.39,
+      "learning_rate": 0.0003020673076923077,
+      "loss": 0.4372,
+      "step": 4620
+    },
+    {
+      "epoch": 42.39,
+      "learning_rate": 0.00030201923076923076,
+      "loss": 0.4392,
+      "step": 4621
+    },
+    {
+      "epoch": 42.4,
+      "learning_rate": 0.00030197115384615386,
+      "loss": 0.2759,
+      "step": 4622
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 0.0003019230769230769,
+      "loss": 0.4718,
+      "step": 4623
+    },
+    {
+      "epoch": 42.42,
+      "learning_rate": 0.000301875,
+      "loss": 0.4059,
+      "step": 4624
+    },
+    {
+      "epoch": 42.43,
+      "learning_rate": 0.0003018269230769231,
+      "loss": 0.4461,
+      "step": 4625
+    },
+    {
+      "epoch": 42.44,
+      "learning_rate": 0.00030177884615384615,
+      "loss": 0.473,
+      "step": 4626
+    },
+    {
+      "epoch": 42.45,
+      "learning_rate": 0.00030173076923076925,
+      "loss": 0.3917,
+      "step": 4627
+    },
+    {
+      "epoch": 42.46,
+      "learning_rate": 0.0003016826923076923,
+      "loss": 0.4488,
+      "step": 4628
+    },
+    {
+      "epoch": 42.47,
+      "learning_rate": 0.0003016346153846154,
+      "loss": 0.4627,
+      "step": 4629
+    },
+    {
+      "epoch": 42.48,
+      "learning_rate": 0.0003015865384615385,
+      "loss": 0.3459,
+      "step": 4630
+    },
+    {
+      "epoch": 42.49,
+      "learning_rate": 0.00030153846153846154,
+      "loss": 0.2809,
+      "step": 4631
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 0.00030149038461538464,
+      "loss": 0.279,
+      "step": 4632
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 0.0003014423076923077,
+      "loss": 0.4768,
+      "step": 4633
+    },
+    {
+      "epoch": 42.51,
+      "learning_rate": 0.0003013942307692308,
+      "loss": 0.5228,
+      "step": 4634
+    },
+    {
+      "epoch": 42.52,
+      "learning_rate": 0.00030134615384615383,
+      "loss": 0.4342,
+      "step": 4635
+    },
+    {
+      "epoch": 42.53,
+      "learning_rate": 0.00030129807692307693,
+      "loss": 0.4369,
+      "step": 4636
+    },
+    {
+      "epoch": 42.54,
+      "learning_rate": 0.00030125000000000003,
+      "loss": 0.4061,
+      "step": 4637
+    },
+    {
+      "epoch": 42.55,
+      "learning_rate": 0.0003012019230769231,
+      "loss": 0.4262,
+      "step": 4638
+    },
+    {
+      "epoch": 42.56,
+      "learning_rate": 0.0003011538461538462,
+      "loss": 0.4435,
+      "step": 4639
+    },
+    {
+      "epoch": 42.57,
+      "learning_rate": 0.0003011057692307692,
+      "loss": 0.4313,
+      "step": 4640
+    },
+    {
+      "epoch": 42.58,
+      "learning_rate": 0.0003010576923076923,
+      "loss": 0.4829,
+      "step": 4641
+    },
+    {
+      "epoch": 42.59,
+      "learning_rate": 0.0003010096153846154,
+      "loss": 0.3323,
+      "step": 4642
+    },
+    {
+      "epoch": 42.6,
+      "learning_rate": 0.00030096153846153847,
+      "loss": 0.4186,
+      "step": 4643
+    },
+    {
+      "epoch": 42.61,
+      "learning_rate": 0.00030091346153846157,
+      "loss": 0.5175,
+      "step": 4644
+    },
+    {
+      "epoch": 42.61,
+      "learning_rate": 0.0003008653846153846,
+      "loss": 0.4051,
+      "step": 4645
+    },
+    {
+      "epoch": 42.62,
+      "learning_rate": 0.0003008173076923077,
+      "loss": 0.3437,
+      "step": 4646
+    },
+    {
+      "epoch": 42.63,
+      "learning_rate": 0.0003007692307692308,
+      "loss": 0.4161,
+      "step": 4647
+    },
+    {
+      "epoch": 42.64,
+      "learning_rate": 0.00030072115384615386,
+      "loss": 0.4631,
+      "step": 4648
+    },
+    {
+      "epoch": 42.65,
+      "learning_rate": 0.00030067307692307696,
+      "loss": 0.3807,
+      "step": 4649
+    },
+    {
+      "epoch": 42.66,
+      "learning_rate": 0.000300625,
+      "loss": 0.3438,
+      "step": 4650
+    },
+    {
+      "epoch": 42.67,
+      "learning_rate": 0.0003005769230769231,
+      "loss": 0.3213,
+      "step": 4651
+    },
+    {
+      "epoch": 42.68,
+      "learning_rate": 0.0003005288461538462,
+      "loss": 0.4226,
+      "step": 4652
+    },
+    {
+      "epoch": 42.69,
+      "learning_rate": 0.00030048076923076925,
+      "loss": 0.4426,
+      "step": 4653
+    },
+    {
+      "epoch": 42.7,
+      "learning_rate": 0.00030043269230769235,
+      "loss": 0.283,
+      "step": 4654
+    },
+    {
+      "epoch": 42.71,
+      "learning_rate": 0.0003003846153846154,
+      "loss": 0.4801,
+      "step": 4655
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 0.0003003365384615385,
+      "loss": 0.3542,
+      "step": 4656
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 0.0003002884615384616,
+      "loss": 0.4129,
+      "step": 4657
+    },
+    {
+      "epoch": 42.73,
+      "learning_rate": 0.0003002403846153846,
+      "loss": 0.3648,
+      "step": 4658
+    },
+    {
+      "epoch": 42.74,
+      "learning_rate": 0.0003001923076923077,
+      "loss": 0.2937,
+      "step": 4659
+    },
+    {
+      "epoch": 42.75,
+      "learning_rate": 0.00030014423076923073,
+      "loss": 0.5325,
+      "step": 4660
+    },
+    {
+      "epoch": 42.76,
+      "learning_rate": 0.00030009615384615383,
+      "loss": 0.4217,
+      "step": 4661
+    },
+    {
+      "epoch": 42.77,
+      "learning_rate": 0.00030004807692307693,
+      "loss": 0.5057,
+      "step": 4662
+    },
+    {
+      "epoch": 42.78,
+      "learning_rate": 0.0003,
+      "loss": 0.4492,
+      "step": 4663
+    },
+    {
+      "epoch": 42.79,
+      "learning_rate": 0.0002999519230769231,
+      "loss": 0.481,
+      "step": 4664
+    },
+    {
+      "epoch": 42.8,
+      "learning_rate": 0.0002999038461538461,
+      "loss": 0.4531,
+      "step": 4665
+    },
+    {
+      "epoch": 42.81,
+      "learning_rate": 0.0002998557692307692,
+      "loss": 0.4407,
+      "step": 4666
+    },
+    {
+      "epoch": 42.82,
+      "learning_rate": 0.0002998076923076923,
+      "loss": 0.3765,
+      "step": 4667
+    },
+    {
+      "epoch": 42.83,
+      "learning_rate": 0.00029975961538461536,
+      "loss": 0.4006,
+      "step": 4668
+    },
+    {
+      "epoch": 42.83,
+      "learning_rate": 0.00029971153846153846,
+      "loss": 0.4374,
+      "step": 4669
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 0.0002996634615384615,
+      "loss": 0.4007,
+      "step": 4670
+    },
+    {
+      "epoch": 42.85,
+      "learning_rate": 0.0002996153846153846,
+      "loss": 0.4824,
+      "step": 4671
+    },
+    {
+      "epoch": 42.86,
+      "learning_rate": 0.0002995673076923077,
+      "loss": 0.399,
+      "step": 4672
+    },
+    {
+      "epoch": 42.87,
+      "learning_rate": 0.00029951923076923075,
+      "loss": 0.3527,
+      "step": 4673
+    },
+    {
+      "epoch": 42.88,
+      "learning_rate": 0.00029947115384615385,
+      "loss": 0.3592,
+      "step": 4674
+    },
+    {
+      "epoch": 42.89,
+      "learning_rate": 0.0002994230769230769,
+      "loss": 0.3724,
+      "step": 4675
+    },
+    {
+      "epoch": 42.9,
+      "learning_rate": 0.000299375,
+      "loss": 0.438,
+      "step": 4676
+    },
+    {
+      "epoch": 42.91,
+      "learning_rate": 0.0002993269230769231,
+      "loss": 0.4665,
+      "step": 4677
+    },
+    {
+      "epoch": 42.92,
+      "learning_rate": 0.00029927884615384614,
+      "loss": 0.3903,
+      "step": 4678
+    },
+    {
+      "epoch": 42.93,
+      "learning_rate": 0.00029923076923076924,
+      "loss": 0.4456,
+      "step": 4679
+    },
+    {
+      "epoch": 42.94,
+      "learning_rate": 0.0002991826923076923,
+      "loss": 0.4188,
+      "step": 4680
+    },
+    {
+      "epoch": 42.94,
+      "learning_rate": 0.0002991346153846154,
+      "loss": 0.4023,
+      "step": 4681
+    },
+    {
+      "epoch": 42.95,
+      "learning_rate": 0.0002990865384615385,
+      "loss": 0.3502,
+      "step": 4682
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 0.00029903846153846153,
+      "loss": 0.3724,
+      "step": 4683
+    },
+    {
+      "epoch": 42.97,
+      "learning_rate": 0.00029899038461538463,
+      "loss": 0.4131,
+      "step": 4684
+    },
+    {
+      "epoch": 42.98,
+      "learning_rate": 0.0002989423076923077,
+      "loss": 0.4398,
+      "step": 4685
+    },
+    {
+      "epoch": 42.99,
+      "learning_rate": 0.0002988942307692308,
+      "loss": 0.2961,
+      "step": 4686
+    },
+    {
+      "epoch": 43.0,
+      "learning_rate": 0.0002988461538461539,
+      "loss": 0.4211,
+      "step": 4687
+    },
+    {
+      "epoch": 43.01,
+      "learning_rate": 0.0002987980769230769,
+      "loss": 0.4965,
+      "step": 4688
+    },
+    {
+      "epoch": 43.02,
+      "learning_rate": 0.00029875,
+      "loss": 0.4679,
+      "step": 4689
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 0.00029870192307692307,
+      "loss": 0.4512,
+      "step": 4690
+    },
+    {
+      "epoch": 43.04,
+      "learning_rate": 0.00029865384615384617,
+      "loss": 0.4615,
+      "step": 4691
+    },
+    {
+      "epoch": 43.05,
+      "learning_rate": 0.00029860576923076927,
+      "loss": 0.3968,
+      "step": 4692
+    },
+    {
+      "epoch": 43.06,
+      "learning_rate": 0.0002985576923076923,
+      "loss": 0.3858,
+      "step": 4693
+    },
+    {
+      "epoch": 43.06,
+      "learning_rate": 0.0002985096153846154,
+      "loss": 0.4178,
+      "step": 4694
+    },
+    {
+      "epoch": 43.07,
+      "learning_rate": 0.00029846153846153846,
+      "loss": 0.3919,
+      "step": 4695
+    },
+    {
+      "epoch": 43.08,
+      "learning_rate": 0.00029841346153846156,
+      "loss": 0.3542,
+      "step": 4696
+    },
+    {
+      "epoch": 43.09,
+      "learning_rate": 0.0002983653846153846,
+      "loss": 0.3524,
+      "step": 4697
+    },
+    {
+      "epoch": 43.1,
+      "learning_rate": 0.0002983173076923077,
+      "loss": 0.4063,
+      "step": 4698
+    },
+    {
+      "epoch": 43.11,
+      "learning_rate": 0.0002982692307692308,
+      "loss": 0.2836,
+      "step": 4699
+    },
+    {
+      "epoch": 43.12,
+      "learning_rate": 0.00029822115384615385,
+      "loss": 0.3402,
+      "step": 4700
+    },
+    {
+      "epoch": 43.13,
+      "learning_rate": 0.00029817307692307695,
+      "loss": 0.4593,
+      "step": 4701
+    },
+    {
+      "epoch": 43.14,
+      "learning_rate": 0.000298125,
+      "loss": 0.4184,
+      "step": 4702
+    },
+    {
+      "epoch": 43.15,
+      "learning_rate": 0.0002980769230769231,
+      "loss": 0.367,
+      "step": 4703
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 0.0002980288461538462,
+      "loss": 0.3546,
+      "step": 4704
+    },
+    {
+      "epoch": 43.17,
+      "learning_rate": 0.00029798076923076924,
+      "loss": 0.4187,
+      "step": 4705
+    },
+    {
+      "epoch": 43.17,
+      "learning_rate": 0.00029793269230769234,
+      "loss": 0.3913,
+      "step": 4706
+    },
+    {
+      "epoch": 43.18,
+      "learning_rate": 0.0002978846153846154,
+      "loss": 0.4295,
+      "step": 4707
+    },
+    {
+      "epoch": 43.19,
+      "learning_rate": 0.0002978365384615385,
+      "loss": 0.3409,
+      "step": 4708
+    },
+    {
+      "epoch": 43.2,
+      "learning_rate": 0.0002977884615384616,
+      "loss": 0.4502,
+      "step": 4709
+    },
+    {
+      "epoch": 43.21,
+      "learning_rate": 0.00029774038461538463,
+      "loss": 0.3238,
+      "step": 4710
+    },
+    {
+      "epoch": 43.22,
+      "learning_rate": 0.00029769230769230773,
+      "loss": 0.4744,
+      "step": 4711
+    },
+    {
+      "epoch": 43.23,
+      "learning_rate": 0.0002976442307692308,
+      "loss": 0.4041,
+      "step": 4712
+    },
+    {
+      "epoch": 43.24,
+      "learning_rate": 0.0002975961538461539,
+      "loss": 0.3207,
+      "step": 4713
+    },
+    {
+      "epoch": 43.25,
+      "learning_rate": 0.000297548076923077,
+      "loss": 0.3075,
+      "step": 4714
+    },
+    {
+      "epoch": 43.26,
+      "learning_rate": 0.00029749999999999997,
+      "loss": 0.5424,
+      "step": 4715
+    },
+    {
+      "epoch": 43.27,
+      "learning_rate": 0.00029745192307692307,
+      "loss": 0.4311,
+      "step": 4716
+    },
+    {
+      "epoch": 43.28,
+      "learning_rate": 0.0002974038461538461,
+      "loss": 0.4176,
+      "step": 4717
+    },
+    {
+      "epoch": 43.28,
+      "learning_rate": 0.0002973557692307692,
+      "loss": 0.4707,
+      "step": 4718
+    },
+    {
+      "epoch": 43.29,
+      "learning_rate": 0.0002973076923076923,
+      "loss": 0.441,
+      "step": 4719
+    },
+    {
+      "epoch": 43.3,
+      "learning_rate": 0.00029725961538461536,
+      "loss": 0.4436,
+      "step": 4720
+    },
+    {
+      "epoch": 43.31,
+      "learning_rate": 0.00029721153846153846,
+      "loss": 0.4109,
+      "step": 4721
+    },
+    {
+      "epoch": 43.32,
+      "learning_rate": 0.0002971634615384615,
+      "loss": 0.3951,
+      "step": 4722
+    },
+    {
+      "epoch": 43.33,
+      "learning_rate": 0.0002971153846153846,
+      "loss": 0.382,
+      "step": 4723
+    },
+    {
+      "epoch": 43.34,
+      "learning_rate": 0.0002970673076923077,
+      "loss": 0.3355,
+      "step": 4724
+    },
+    {
+      "epoch": 43.35,
+      "learning_rate": 0.00029701923076923075,
+      "loss": 0.4519,
+      "step": 4725
+    },
+    {
+      "epoch": 43.36,
+      "learning_rate": 0.00029697115384615385,
+      "loss": 0.3822,
+      "step": 4726
+    },
+    {
+      "epoch": 43.37,
+      "learning_rate": 0.0002969230769230769,
+      "loss": 0.3581,
+      "step": 4727
+    },
+    {
+      "epoch": 43.38,
+      "learning_rate": 0.000296875,
+      "loss": 0.3922,
+      "step": 4728
+    },
+    {
+      "epoch": 43.39,
+      "learning_rate": 0.0002968269230769231,
+      "loss": 0.4049,
+      "step": 4729
+    },
+    {
+      "epoch": 43.39,
+      "learning_rate": 0.00029677884615384614,
+      "loss": 0.5047,
+      "step": 4730
+    },
+    {
+      "epoch": 43.4,
+      "learning_rate": 0.00029673076923076924,
+      "loss": 0.3541,
+      "step": 4731
+    },
+    {
+      "epoch": 43.41,
+      "learning_rate": 0.0002966826923076923,
+      "loss": 0.414,
+      "step": 4732
+    },
+    {
+      "epoch": 43.42,
+      "learning_rate": 0.0002966346153846154,
+      "loss": 0.3838,
+      "step": 4733
+    },
+    {
+      "epoch": 43.43,
+      "learning_rate": 0.0002965865384615385,
+      "loss": 0.408,
+      "step": 4734
+    },
+    {
+      "epoch": 43.44,
+      "learning_rate": 0.00029653846153846153,
+      "loss": 0.3998,
+      "step": 4735
+    },
+    {
+      "epoch": 43.45,
+      "learning_rate": 0.00029649038461538463,
+      "loss": 0.4319,
+      "step": 4736
+    },
+    {
+      "epoch": 43.46,
+      "learning_rate": 0.0002964423076923077,
+      "loss": 0.3856,
+      "step": 4737
+    },
+    {
+      "epoch": 43.47,
+      "learning_rate": 0.0002963942307692308,
+      "loss": 0.5399,
+      "step": 4738
+    },
+    {
+      "epoch": 43.48,
+      "learning_rate": 0.00029634615384615387,
+      "loss": 0.3951,
+      "step": 4739
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 0.0002962980769230769,
+      "loss": 0.3196,
+      "step": 4740
+    },
+    {
+      "epoch": 43.5,
+      "learning_rate": 0.00029625,
+      "loss": 0.459,
+      "step": 4741
+    },
+    {
+      "epoch": 43.5,
+      "learning_rate": 0.00029620192307692306,
+      "loss": 0.4962,
+      "step": 4742
+    },
+    {
+      "epoch": 43.51,
+      "learning_rate": 0.00029615384615384616,
+      "loss": 0.4101,
+      "step": 4743
+    },
+    {
+      "epoch": 43.52,
+      "learning_rate": 0.00029610576923076926,
+      "loss": 0.4059,
+      "step": 4744
+    },
+    {
+      "epoch": 43.53,
+      "learning_rate": 0.0002960576923076923,
+      "loss": 0.4077,
+      "step": 4745
+    },
+    {
+      "epoch": 43.54,
+      "learning_rate": 0.0002960096153846154,
+      "loss": 0.4469,
+      "step": 4746
+    },
+    {
+      "epoch": 43.55,
+      "learning_rate": 0.00029596153846153845,
+      "loss": 0.4052,
+      "step": 4747
+    },
+    {
+      "epoch": 43.56,
+      "learning_rate": 0.00029591346153846155,
+      "loss": 0.3863,
+      "step": 4748
+    },
+    {
+      "epoch": 43.57,
+      "learning_rate": 0.00029586538461538465,
+      "loss": 0.3681,
+      "step": 4749
+    },
+    {
+      "epoch": 43.58,
+      "learning_rate": 0.0002958173076923077,
+      "loss": 0.4843,
+      "step": 4750
+    },
+    {
+      "epoch": 43.59,
+      "learning_rate": 0.0002957692307692308,
+      "loss": 0.5858,
+      "step": 4751
+    },
+    {
+      "epoch": 43.6,
+      "learning_rate": 0.00029572115384615384,
+      "loss": 0.3993,
+      "step": 4752
+    },
+    {
+      "epoch": 43.61,
+      "learning_rate": 0.00029567307692307694,
+      "loss": 0.4349,
+      "step": 4753
+    },
+    {
+      "epoch": 43.61,
+      "learning_rate": 0.00029562500000000004,
+      "loss": 0.425,
+      "step": 4754
+    },
+    {
+      "epoch": 43.62,
+      "learning_rate": 0.0002955769230769231,
+      "loss": 0.4612,
+      "step": 4755
+    },
+    {
+      "epoch": 43.63,
+      "learning_rate": 0.0002955288461538462,
+      "loss": 0.326,
+      "step": 4756
+    },
+    {
+      "epoch": 43.64,
+      "learning_rate": 0.00029548076923076923,
+      "loss": 0.3778,
+      "step": 4757
+    },
+    {
+      "epoch": 43.65,
+      "learning_rate": 0.00029543269230769233,
+      "loss": 0.4333,
+      "step": 4758
+    },
+    {
+      "epoch": 43.66,
+      "learning_rate": 0.00029538461538461543,
+      "loss": 0.3431,
+      "step": 4759
+    },
+    {
+      "epoch": 43.67,
+      "learning_rate": 0.0002953365384615385,
+      "loss": 0.4325,
+      "step": 4760
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 0.0002952884615384616,
+      "loss": 0.3895,
+      "step": 4761
+    },
+    {
+      "epoch": 43.69,
+      "learning_rate": 0.0002952403846153846,
+      "loss": 0.4506,
+      "step": 4762
+    },
+    {
+      "epoch": 43.7,
+      "learning_rate": 0.0002951923076923077,
+      "loss": 0.3618,
+      "step": 4763
+    },
+    {
+      "epoch": 43.71,
+      "learning_rate": 0.00029514423076923077,
+      "loss": 0.3606,
+      "step": 4764
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 0.00029509615384615387,
+      "loss": 0.3925,
+      "step": 4765
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 0.00029504807692307697,
+      "loss": 0.4541,
+      "step": 4766
+    },
+    {
+      "epoch": 43.73,
+      "learning_rate": 0.000295,
+      "loss": 0.4172,
+      "step": 4767
+    },
+    {
+      "epoch": 43.74,
+      "learning_rate": 0.0002949519230769231,
+      "loss": 0.4041,
+      "step": 4768
+    },
+    {
+      "epoch": 43.75,
+      "learning_rate": 0.00029490384615384616,
+      "loss": 0.4324,
+      "step": 4769
+    },
+    {
+      "epoch": 43.76,
+      "learning_rate": 0.00029485576923076926,
+      "loss": 0.4038,
+      "step": 4770
+    },
+    {
+      "epoch": 43.77,
+      "learning_rate": 0.0002948076923076923,
+      "loss": 0.4635,
+      "step": 4771
+    },
+    {
+      "epoch": 43.78,
+      "learning_rate": 0.00029475961538461535,
+      "loss": 0.4106,
+      "step": 4772
+    },
+    {
+      "epoch": 43.79,
+      "learning_rate": 0.00029471153846153845,
+      "loss": 0.4236,
+      "step": 4773
+    },
+    {
+      "epoch": 43.8,
+      "learning_rate": 0.0002946634615384615,
+      "loss": 0.38,
+      "step": 4774
+    },
+    {
+      "epoch": 43.81,
+      "learning_rate": 0.0002946153846153846,
+      "loss": 0.4626,
+      "step": 4775
+    },
+    {
+      "epoch": 43.82,
+      "learning_rate": 0.0002945673076923077,
+      "loss": 0.3512,
+      "step": 4776
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 0.00029451923076923074,
+      "loss": 0.3238,
+      "step": 4777
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 0.00029447115384615384,
+      "loss": 0.3551,
+      "step": 4778
+    },
+    {
+      "epoch": 43.84,
+      "learning_rate": 0.0002944230769230769,
+      "loss": 0.4466,
+      "step": 4779
+    },
+    {
+      "epoch": 43.85,
+      "learning_rate": 0.000294375,
+      "loss": 0.384,
+      "step": 4780
+    },
+    {
+      "epoch": 43.86,
+      "learning_rate": 0.0002943269230769231,
+      "loss": 0.3464,
+      "step": 4781
+    },
+    {
+      "epoch": 43.87,
+      "learning_rate": 0.00029427884615384613,
+      "loss": 0.3632,
+      "step": 4782
+    },
+    {
+      "epoch": 43.88,
+      "learning_rate": 0.00029423076923076923,
+      "loss": 0.3855,
+      "step": 4783
+    },
+    {
+      "epoch": 43.89,
+      "learning_rate": 0.0002941826923076923,
+      "loss": 0.3996,
+      "step": 4784
+    },
+    {
+      "epoch": 43.9,
+      "learning_rate": 0.0002941346153846154,
+      "loss": 0.4564,
+      "step": 4785
+    },
+    {
+      "epoch": 43.91,
+      "learning_rate": 0.0002940865384615385,
+      "loss": 0.4017,
+      "step": 4786
+    },
+    {
+      "epoch": 43.92,
+      "learning_rate": 0.0002940384615384615,
+      "loss": 0.4178,
+      "step": 4787
+    },
+    {
+      "epoch": 43.93,
+      "learning_rate": 0.0002939903846153846,
+      "loss": 0.412,
+      "step": 4788
+    },
+    {
+      "epoch": 43.94,
+      "learning_rate": 0.00029394230769230767,
+      "loss": 0.4201,
+      "step": 4789
+    },
+    {
+      "epoch": 43.94,
+      "learning_rate": 0.00029389423076923077,
+      "loss": 0.4283,
+      "step": 4790
+    },
+    {
+      "epoch": 43.95,
+      "learning_rate": 0.00029384615384615387,
+      "loss": 0.4689,
+      "step": 4791
+    },
+    {
+      "epoch": 43.96,
+      "learning_rate": 0.0002937980769230769,
+      "loss": 0.3441,
+      "step": 4792
+    },
+    {
+      "epoch": 43.97,
+      "learning_rate": 0.00029375,
+      "loss": 0.4626,
+      "step": 4793
+    },
+    {
+      "epoch": 43.98,
+      "learning_rate": 0.00029370192307692306,
+      "loss": 0.336,
+      "step": 4794
+    },
+    {
+      "epoch": 43.99,
+      "learning_rate": 0.00029365384615384616,
+      "loss": 0.4527,
+      "step": 4795
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 0.00029360576923076926,
+      "loss": 0.4489,
+      "step": 4796
+    },
+    {
+      "epoch": 44.01,
+      "learning_rate": 0.0002935576923076923,
+      "loss": 0.5151,
+      "step": 4797
+    },
+    {
+      "epoch": 44.02,
+      "learning_rate": 0.0002935096153846154,
+      "loss": 0.4743,
+      "step": 4798
+    },
+    {
+      "epoch": 44.03,
+      "learning_rate": 0.00029346153846153845,
+      "loss": 0.3772,
+      "step": 4799
+    },
+    {
+      "epoch": 44.04,
+      "learning_rate": 0.00029341346153846155,
+      "loss": 0.3807,
+      "step": 4800
+    },
+    {
+      "epoch": 44.05,
+      "learning_rate": 0.00029336538461538465,
+      "loss": 0.5146,
+      "step": 4801
+    },
+    {
+      "epoch": 44.06,
+      "learning_rate": 0.0002933173076923077,
+      "loss": 0.3508,
+      "step": 4802
+    },
+    {
+      "epoch": 44.06,
+      "learning_rate": 0.0002932692307692308,
+      "loss": 0.3871,
+      "step": 4803
+    },
+    {
+      "epoch": 44.07,
+      "learning_rate": 0.00029322115384615384,
+      "loss": 0.3593,
+      "step": 4804
+    },
+    {
+      "epoch": 44.08,
+      "learning_rate": 0.00029317307692307694,
+      "loss": 0.4199,
+      "step": 4805
+    },
+    {
+      "epoch": 44.09,
+      "learning_rate": 0.00029312500000000004,
+      "loss": 0.3829,
+      "step": 4806
+    },
+    {
+      "epoch": 44.1,
+      "learning_rate": 0.0002930769230769231,
+      "loss": 0.4483,
+      "step": 4807
+    },
+    {
+      "epoch": 44.11,
+      "learning_rate": 0.0002930288461538462,
+      "loss": 0.3102,
+      "step": 4808
+    },
+    {
+      "epoch": 44.12,
+      "learning_rate": 0.00029298076923076923,
+      "loss": 0.4003,
+      "step": 4809
+    },
+    {
+      "epoch": 44.13,
+      "learning_rate": 0.00029293269230769233,
+      "loss": 0.4558,
+      "step": 4810
+    },
+    {
+      "epoch": 44.14,
+      "learning_rate": 0.00029288461538461543,
+      "loss": 0.4248,
+      "step": 4811
+    },
+    {
+      "epoch": 44.15,
+      "learning_rate": 0.00029283653846153847,
+      "loss": 0.4124,
+      "step": 4812
+    },
+    {
+      "epoch": 44.16,
+      "learning_rate": 0.00029278846153846157,
+      "loss": 0.3884,
+      "step": 4813
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 0.0002927403846153846,
+      "loss": 0.4242,
+      "step": 4814
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 0.0002926923076923077,
+      "loss": 0.3916,
+      "step": 4815
+    },
+    {
+      "epoch": 44.18,
+      "learning_rate": 0.0002926442307692308,
+      "loss": 0.3007,
+      "step": 4816
+    },
+    {
+      "epoch": 44.19,
+      "learning_rate": 0.00029259615384615386,
+      "loss": 0.3408,
+      "step": 4817
+    },
+    {
+      "epoch": 44.2,
+      "learning_rate": 0.00029254807692307696,
+      "loss": 0.2841,
+      "step": 4818
+    },
+    {
+      "epoch": 44.21,
+      "learning_rate": 0.0002925,
+      "loss": 0.3224,
+      "step": 4819
+    },
+    {
+      "epoch": 44.22,
+      "learning_rate": 0.0002924519230769231,
+      "loss": 0.4012,
+      "step": 4820
+    },
+    {
+      "epoch": 44.23,
+      "learning_rate": 0.0002924038461538462,
+      "loss": 0.3442,
+      "step": 4821
+    },
+    {
+      "epoch": 44.24,
+      "learning_rate": 0.00029235576923076925,
+      "loss": 0.3505,
+      "step": 4822
+    },
+    {
+      "epoch": 44.25,
+      "learning_rate": 0.00029230769230769235,
+      "loss": 0.2907,
+      "step": 4823
+    },
+    {
+      "epoch": 44.26,
+      "learning_rate": 0.0002922596153846154,
+      "loss": 0.4925,
+      "step": 4824
+    },
+    {
+      "epoch": 44.27,
+      "learning_rate": 0.0002922115384615385,
+      "loss": 0.4305,
+      "step": 4825
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 0.00029216346153846154,
+      "loss": 0.5735,
+      "step": 4826
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 0.0002921153846153846,
+      "loss": 0.3365,
+      "step": 4827
+    },
+    {
+      "epoch": 44.29,
+      "learning_rate": 0.0002920673076923077,
+      "loss": 0.3954,
+      "step": 4828
+    },
+    {
+      "epoch": 44.3,
+      "learning_rate": 0.00029201923076923073,
+      "loss": 0.4324,
+      "step": 4829
+    },
+    {
+      "epoch": 44.31,
+      "learning_rate": 0.00029197115384615383,
+      "loss": 0.476,
+      "step": 4830
+    },
+    {
+      "epoch": 44.32,
+      "learning_rate": 0.0002919230769230769,
+      "loss": 0.3881,
+      "step": 4831
+    },
+    {
+      "epoch": 44.33,
+      "learning_rate": 0.000291875,
+      "loss": 0.4882,
+      "step": 4832
+    },
+    {
+      "epoch": 44.34,
+      "learning_rate": 0.0002918269230769231,
+      "loss": 0.355,
+      "step": 4833
+    },
+    {
+      "epoch": 44.35,
+      "learning_rate": 0.0002917788461538461,
+      "loss": 0.2907,
+      "step": 4834
+    },
+    {
+      "epoch": 44.36,
+      "learning_rate": 0.0002917307692307692,
+      "loss": 0.4669,
+      "step": 4835
+    },
+    {
+      "epoch": 44.37,
+      "learning_rate": 0.00029168269230769227,
+      "loss": 0.2681,
+      "step": 4836
+    },
+    {
+      "epoch": 44.38,
+      "learning_rate": 0.00029163461538461537,
+      "loss": 0.3825,
+      "step": 4837
+    },
+    {
+      "epoch": 44.39,
+      "learning_rate": 0.00029158653846153847,
+      "loss": 0.4293,
+      "step": 4838
+    },
+    {
+      "epoch": 44.39,
+      "learning_rate": 0.0002915384615384615,
+      "loss": 0.5477,
+      "step": 4839
+    },
+    {
+      "epoch": 44.4,
+      "learning_rate": 0.0002914903846153846,
+      "loss": 0.3248,
+      "step": 4840
+    },
+    {
+      "epoch": 44.41,
+      "learning_rate": 0.00029144230769230766,
+      "loss": 0.3673,
+      "step": 4841
+    },
+    {
+      "epoch": 44.42,
+      "learning_rate": 0.00029139423076923076,
+      "loss": 0.4863,
+      "step": 4842
+    },
+    {
+      "epoch": 44.43,
+      "learning_rate": 0.00029134615384615386,
+      "loss": 0.3451,
+      "step": 4843
+    },
+    {
+      "epoch": 44.44,
+      "learning_rate": 0.0002912980769230769,
+      "loss": 0.3765,
+      "step": 4844
+    },
+    {
+      "epoch": 44.45,
+      "learning_rate": 0.00029125,
+      "loss": 0.3586,
+      "step": 4845
+    },
+    {
+      "epoch": 44.46,
+      "learning_rate": 0.00029120192307692305,
+      "loss": 0.4345,
+      "step": 4846
+    },
+    {
+      "epoch": 44.47,
+      "learning_rate": 0.00029115384615384615,
+      "loss": 0.3002,
+      "step": 4847
+    },
+    {
+      "epoch": 44.48,
+      "learning_rate": 0.00029110576923076925,
+      "loss": 0.2847,
+      "step": 4848
+    },
+    {
+      "epoch": 44.49,
+      "learning_rate": 0.0002910576923076923,
+      "loss": 0.3247,
+      "step": 4849
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 0.0002910096153846154,
+      "loss": 0.3352,
+      "step": 4850
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 0.00029096153846153844,
+      "loss": 0.5209,
+      "step": 4851
+    },
+    {
+      "epoch": 44.51,
+      "learning_rate": 0.00029091346153846154,
+      "loss": 0.4532,
+      "step": 4852
+    },
+    {
+      "epoch": 44.52,
+      "learning_rate": 0.00029086538461538464,
+      "loss": 0.4724,
+      "step": 4853
+    },
+    {
+      "epoch": 44.53,
+      "learning_rate": 0.0002908173076923077,
+      "loss": 0.4273,
+      "step": 4854
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 0.0002907692307692308,
+      "loss": 0.3498,
+      "step": 4855
+    },
+    {
+      "epoch": 44.55,
+      "learning_rate": 0.00029072115384615383,
+      "loss": 0.3921,
+      "step": 4856
+    },
+    {
+      "epoch": 44.56,
+      "learning_rate": 0.00029067307692307693,
+      "loss": 0.3098,
+      "step": 4857
+    },
+    {
+      "epoch": 44.57,
+      "learning_rate": 0.00029062500000000003,
+      "loss": 0.2983,
+      "step": 4858
+    },
+    {
+      "epoch": 44.58,
+      "learning_rate": 0.0002905769230769231,
+      "loss": 0.4122,
+      "step": 4859
+    },
+    {
+      "epoch": 44.59,
+      "learning_rate": 0.0002905288461538462,
+      "loss": 0.4399,
+      "step": 4860
+    },
+    {
+      "epoch": 44.6,
+      "learning_rate": 0.0002904807692307692,
+      "loss": 0.3498,
+      "step": 4861
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 0.0002904326923076923,
+      "loss": 0.3206,
+      "step": 4862
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 0.0002903846153846154,
+      "loss": 0.345,
+      "step": 4863
+    },
+    {
+      "epoch": 44.62,
+      "learning_rate": 0.00029033653846153847,
+      "loss": 0.4405,
+      "step": 4864
+    },
+    {
+      "epoch": 44.63,
+      "learning_rate": 0.00029028846153846157,
+      "loss": 0.4106,
+      "step": 4865
+    },
+    {
+      "epoch": 44.64,
+      "learning_rate": 0.0002902403846153846,
+      "loss": 0.4363,
+      "step": 4866
+    },
+    {
+      "epoch": 44.65,
+      "learning_rate": 0.0002901923076923077,
+      "loss": 0.4018,
+      "step": 4867
+    },
+    {
+      "epoch": 44.66,
+      "learning_rate": 0.0002901442307692308,
+      "loss": 0.3335,
+      "step": 4868
+    },
+    {
+      "epoch": 44.67,
+      "learning_rate": 0.00029009615384615386,
+      "loss": 0.3586,
+      "step": 4869
+    },
+    {
+      "epoch": 44.68,
+      "learning_rate": 0.00029004807692307696,
+      "loss": 0.4551,
+      "step": 4870
+    },
+    {
+      "epoch": 44.69,
+      "learning_rate": 0.00029,
+      "loss": 0.4816,
+      "step": 4871
+    },
+    {
+      "epoch": 44.7,
+      "learning_rate": 0.0002899519230769231,
+      "loss": 0.3513,
+      "step": 4872
+    },
+    {
+      "epoch": 44.71,
+      "learning_rate": 0.0002899038461538462,
+      "loss": 0.5109,
+      "step": 4873
+    },
+    {
+      "epoch": 44.72,
+      "learning_rate": 0.00028985576923076925,
+      "loss": 0.3618,
+      "step": 4874
+    },
+    {
+      "epoch": 44.72,
+      "learning_rate": 0.00028980769230769235,
+      "loss": 0.3959,
+      "step": 4875
+    },
+    {
+      "epoch": 44.73,
+      "learning_rate": 0.0002897596153846154,
+      "loss": 0.4095,
+      "step": 4876
+    },
+    {
+      "epoch": 44.74,
+      "learning_rate": 0.0002897115384615385,
+      "loss": 0.2766,
+      "step": 4877
+    },
+    {
+      "epoch": 44.75,
+      "learning_rate": 0.0002896634615384616,
+      "loss": 0.5161,
+      "step": 4878
+    },
+    {
+      "epoch": 44.76,
+      "learning_rate": 0.00028961538461538464,
+      "loss": 0.452,
+      "step": 4879
+    },
+    {
+      "epoch": 44.77,
+      "learning_rate": 0.00028956730769230774,
+      "loss": 0.4632,
+      "step": 4880
+    },
+    {
+      "epoch": 44.78,
+      "learning_rate": 0.0002895192307692308,
+      "loss": 0.4796,
+      "step": 4881
+    },
+    {
+      "epoch": 44.79,
+      "learning_rate": 0.0002894711538461539,
+      "loss": 0.4375,
+      "step": 4882
+    },
+    {
+      "epoch": 44.8,
+      "learning_rate": 0.000289423076923077,
+      "loss": 0.3928,
+      "step": 4883
+    },
+    {
+      "epoch": 44.81,
+      "learning_rate": 0.000289375,
+      "loss": 0.4577,
+      "step": 4884
+    },
+    {
+      "epoch": 44.82,
+      "learning_rate": 0.00028932692307692307,
+      "loss": 0.4451,
+      "step": 4885
+    },
+    {
+      "epoch": 44.83,
+      "learning_rate": 0.0002892788461538461,
+      "loss": 0.3727,
+      "step": 4886
+    },
+    {
+      "epoch": 44.83,
+      "learning_rate": 0.0002892307692307692,
+      "loss": 0.3781,
+      "step": 4887
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 0.0002891826923076923,
+      "loss": 0.3284,
+      "step": 4888
+    },
+    {
+      "epoch": 44.85,
+      "learning_rate": 0.00028913461538461536,
+      "loss": 0.4312,
+      "step": 4889
+    },
+    {
+      "epoch": 44.86,
+      "learning_rate": 0.00028908653846153846,
+      "loss": 0.3903,
+      "step": 4890
+    },
+    {
+      "epoch": 44.87,
+      "learning_rate": 0.0002890384615384615,
+      "loss": 0.3739,
+      "step": 4891
+    },
+    {
+      "epoch": 44.88,
+      "learning_rate": 0.0002889903846153846,
+      "loss": 0.4502,
+      "step": 4892
+    },
+    {
+      "epoch": 44.89,
+      "learning_rate": 0.00028894230769230765,
+      "loss": 0.4964,
+      "step": 4893
+    },
+    {
+      "epoch": 44.9,
+      "learning_rate": 0.00028889423076923075,
+      "loss": 0.4577,
+      "step": 4894
+    },
+    {
+      "epoch": 44.91,
+      "learning_rate": 0.00028884615384615385,
+      "loss": 0.3749,
+      "step": 4895
+    },
+    {
+      "epoch": 44.92,
+      "learning_rate": 0.0002887980769230769,
+      "loss": 0.4637,
+      "step": 4896
+    },
+    {
+      "epoch": 44.93,
+      "learning_rate": 0.00028875,
+      "loss": 0.3777,
+      "step": 4897
+    },
+    {
+      "epoch": 44.94,
+      "learning_rate": 0.00028870192307692304,
+      "loss": 0.3564,
+      "step": 4898
+    },
+    {
+      "epoch": 44.94,
+      "learning_rate": 0.00028865384615384614,
+      "loss": 0.4703,
+      "step": 4899
+    },
+    {
+      "epoch": 44.95,
+      "learning_rate": 0.00028860576923076924,
+      "loss": 0.3445,
+      "step": 4900
+    },
+    {
+      "epoch": 44.96,
+      "learning_rate": 0.0002885576923076923,
+      "loss": 0.4377,
+      "step": 4901
+    },
+    {
+      "epoch": 44.97,
+      "learning_rate": 0.0002885096153846154,
+      "loss": 0.3272,
+      "step": 4902
+    },
+    {
+      "epoch": 44.98,
+      "learning_rate": 0.00028846153846153843,
+      "loss": 0.4628,
+      "step": 4903
+    },
+    {
+      "epoch": 44.99,
+      "learning_rate": 0.00028841346153846153,
+      "loss": 0.2632,
+      "step": 4904
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 0.00028836538461538463,
+      "loss": 0.2909,
+      "step": 4905
+    },
+    {
+      "epoch": 45.01,
+      "learning_rate": 0.0002883173076923077,
+      "loss": 0.5025,
+      "step": 4906
+    },
+    {
+      "epoch": 45.02,
+      "learning_rate": 0.0002882692307692308,
+      "loss": 0.4006,
+      "step": 4907
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 0.0002882211538461538,
+      "loss": 0.3379,
+      "step": 4908
+    },
+    {
+      "epoch": 45.04,
+      "learning_rate": 0.0002881730769230769,
+      "loss": 0.4272,
+      "step": 4909
+    },
+    {
+      "epoch": 45.05,
+      "learning_rate": 0.000288125,
+      "loss": 0.4113,
+      "step": 4910
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 0.00028807692307692307,
+      "loss": 0.2895,
+      "step": 4911
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 0.00028802884615384617,
+      "loss": 0.4988,
+      "step": 4912
+    },
+    {
+      "epoch": 45.07,
+      "learning_rate": 0.0002879807692307692,
+      "loss": 0.3796,
+      "step": 4913
+    },
+    {
+      "epoch": 45.08,
+      "learning_rate": 0.0002879326923076923,
+      "loss": 0.3721,
+      "step": 4914
+    },
+    {
+      "epoch": 45.09,
+      "learning_rate": 0.0002878846153846154,
+      "loss": 0.3176,
+      "step": 4915
+    },
+    {
+      "epoch": 45.1,
+      "learning_rate": 0.00028783653846153846,
+      "loss": 0.3521,
+      "step": 4916
+    },
+    {
+      "epoch": 45.11,
+      "learning_rate": 0.00028778846153846156,
+      "loss": 0.3411,
+      "step": 4917
+    },
+    {
+      "epoch": 45.12,
+      "learning_rate": 0.0002877403846153846,
+      "loss": 0.3027,
+      "step": 4918
+    },
+    {
+      "epoch": 45.13,
+      "learning_rate": 0.0002876923076923077,
+      "loss": 0.4423,
+      "step": 4919
+    },
+    {
+      "epoch": 45.14,
+      "learning_rate": 0.0002876442307692308,
+      "loss": 0.3952,
+      "step": 4920
+    },
+    {
+      "epoch": 45.15,
+      "learning_rate": 0.00028759615384615385,
+      "loss": 0.4164,
+      "step": 4921
+    },
+    {
+      "epoch": 45.16,
+      "learning_rate": 0.00028754807692307695,
+      "loss": 0.4604,
+      "step": 4922
+    },
+    {
+      "epoch": 45.17,
+      "learning_rate": 0.0002875,
+      "loss": 0.4006,
+      "step": 4923
+    },
+    {
+      "epoch": 45.17,
+      "learning_rate": 0.0002874519230769231,
+      "loss": 0.5096,
+      "step": 4924
+    },
+    {
+      "epoch": 45.18,
+      "learning_rate": 0.0002874038461538462,
+      "loss": 0.4172,
+      "step": 4925
+    },
+    {
+      "epoch": 45.19,
+      "learning_rate": 0.00028735576923076924,
+      "loss": 0.3414,
+      "step": 4926
+    },
+    {
+      "epoch": 45.2,
+      "learning_rate": 0.00028730769230769234,
+      "loss": 0.3605,
+      "step": 4927
+    },
+    {
+      "epoch": 45.21,
+      "learning_rate": 0.0002872596153846154,
+      "loss": 0.337,
+      "step": 4928
+    },
+    {
+      "epoch": 45.22,
+      "learning_rate": 0.0002872115384615385,
+      "loss": 0.2853,
+      "step": 4929
+    },
+    {
+      "epoch": 45.23,
+      "learning_rate": 0.0002871634615384616,
+      "loss": 0.3704,
+      "step": 4930
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 0.00028711538461538463,
+      "loss": 0.348,
+      "step": 4931
+    },
+    {
+      "epoch": 45.25,
+      "learning_rate": 0.00028706730769230773,
+      "loss": 0.2878,
+      "step": 4932
+    },
+    {
+      "epoch": 45.26,
+      "learning_rate": 0.0002870192307692308,
+      "loss": 0.4541,
+      "step": 4933
+    },
+    {
+      "epoch": 45.27,
+      "learning_rate": 0.0002869711538461539,
+      "loss": 0.4518,
+      "step": 4934
+    },
+    {
+      "epoch": 45.28,
+      "learning_rate": 0.000286923076923077,
+      "loss": 0.4971,
+      "step": 4935
+    },
+    {
+      "epoch": 45.28,
+      "learning_rate": 0.000286875,
+      "loss": 0.4151,
+      "step": 4936
+    },
+    {
+      "epoch": 45.29,
+      "learning_rate": 0.0002868269230769231,
+      "loss": 0.3641,
+      "step": 4937
+    },
+    {
+      "epoch": 45.3,
+      "learning_rate": 0.00028677884615384617,
+      "loss": 0.3517,
+      "step": 4938
+    },
+    {
+      "epoch": 45.31,
+      "learning_rate": 0.00028673076923076927,
+      "loss": 0.4039,
+      "step": 4939
+    },
+    {
+      "epoch": 45.32,
+      "learning_rate": 0.00028668269230769237,
+      "loss": 0.3649,
+      "step": 4940
+    },
+    {
+      "epoch": 45.33,
+      "learning_rate": 0.00028663461538461536,
+      "loss": 0.416,
+      "step": 4941
+    },
+    {
+      "epoch": 45.34,
+      "learning_rate": 0.00028658653846153846,
+      "loss": 0.4283,
+      "step": 4942
+    },
+    {
+      "epoch": 45.35,
+      "learning_rate": 0.0002865384615384615,
+      "loss": 0.3772,
+      "step": 4943
+    },
+    {
+      "epoch": 45.36,
+      "learning_rate": 0.0002864903846153846,
+      "loss": 0.3353,
+      "step": 4944
+    },
+    {
+      "epoch": 45.37,
+      "learning_rate": 0.0002864423076923077,
+      "loss": 0.3681,
+      "step": 4945
+    },
+    {
+      "epoch": 45.38,
+      "learning_rate": 0.00028639423076923075,
+      "loss": 0.3684,
+      "step": 4946
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 0.00028634615384615385,
+      "loss": 0.3779,
+      "step": 4947
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 0.0002862980769230769,
+      "loss": 0.3359,
+      "step": 4948
+    },
+    {
+      "epoch": 45.4,
+      "learning_rate": 0.00028625,
+      "loss": 0.3716,
+      "step": 4949
+    },
+    {
+      "epoch": 45.41,
+      "learning_rate": 0.0002862019230769231,
+      "loss": 0.4083,
+      "step": 4950
+    },
+    {
+      "epoch": 45.42,
+      "learning_rate": 0.00028615384615384614,
+      "loss": 0.3625,
+      "step": 4951
+    },
+    {
+      "epoch": 45.43,
+      "learning_rate": 0.00028610576923076924,
+      "loss": 0.3891,
+      "step": 4952
+    },
+    {
+      "epoch": 45.44,
+      "learning_rate": 0.0002860576923076923,
+      "loss": 0.4001,
+      "step": 4953
+    },
+    {
+      "epoch": 45.45,
+      "learning_rate": 0.0002860096153846154,
+      "loss": 0.3526,
+      "step": 4954
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 0.0002859615384615385,
+      "loss": 0.3802,
+      "step": 4955
+    },
+    {
+      "epoch": 45.47,
+      "learning_rate": 0.00028591346153846153,
+      "loss": 0.4418,
+      "step": 4956
+    },
+    {
+      "epoch": 45.48,
+      "learning_rate": 0.0002858653846153846,
+      "loss": 0.3228,
+      "step": 4957
+    },
+    {
+      "epoch": 45.49,
+      "learning_rate": 0.00028581730769230767,
+      "loss": 0.3891,
+      "step": 4958
+    },
+    {
+      "epoch": 45.5,
+      "learning_rate": 0.00028576923076923077,
+      "loss": 0.2784,
+      "step": 4959
+    },
+    {
+      "epoch": 45.5,
+      "learning_rate": 0.0002857211538461538,
+      "loss": 0.4801,
+      "step": 4960
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 0.0002856730769230769,
+      "loss": 0.4048,
+      "step": 4961
+    },
+    {
+      "epoch": 45.52,
+      "learning_rate": 0.000285625,
+      "loss": 0.3847,
+      "step": 4962
+    },
+    {
+      "epoch": 45.53,
+      "learning_rate": 0.00028557692307692306,
+      "loss": 0.4246,
+      "step": 4963
+    },
+    {
+      "epoch": 45.54,
+      "learning_rate": 0.00028552884615384616,
+      "loss": 0.3464,
+      "step": 4964
+    },
+    {
+      "epoch": 45.55,
+      "learning_rate": 0.0002854807692307692,
+      "loss": 0.3603,
+      "step": 4965
+    },
+    {
+      "epoch": 45.56,
+      "learning_rate": 0.0002854326923076923,
+      "loss": 0.4004,
+      "step": 4966
+    },
+    {
+      "epoch": 45.57,
+      "learning_rate": 0.0002853846153846154,
+      "loss": 0.3491,
+      "step": 4967
+    },
+    {
+      "epoch": 45.58,
+      "learning_rate": 0.00028533653846153845,
+      "loss": 0.3366,
+      "step": 4968
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 0.00028528846153846155,
+      "loss": 0.4811,
+      "step": 4969
+    },
+    {
+      "epoch": 45.6,
+      "learning_rate": 0.0002852403846153846,
+      "loss": 0.4334,
+      "step": 4970
+    },
+    {
+      "epoch": 45.61,
+      "learning_rate": 0.0002851923076923077,
+      "loss": 0.3367,
+      "step": 4971
+    },
+    {
+      "epoch": 45.61,
+      "learning_rate": 0.0002851442307692308,
+      "loss": 0.3333,
+      "step": 4972
+    },
+    {
+      "epoch": 45.62,
+      "learning_rate": 0.00028509615384615384,
+      "loss": 0.3618,
+      "step": 4973
+    },
+    {
+      "epoch": 45.63,
+      "learning_rate": 0.00028504807692307694,
+      "loss": 0.3005,
+      "step": 4974
+    },
+    {
+      "epoch": 45.64,
+      "learning_rate": 0.000285,
+      "loss": 0.3911,
+      "step": 4975
+    },
+    {
+      "epoch": 45.65,
+      "learning_rate": 0.0002849519230769231,
+      "loss": 0.2803,
+      "step": 4976
+    },
+    {
+      "epoch": 45.66,
+      "learning_rate": 0.0002849038461538462,
+      "loss": 0.3576,
+      "step": 4977
+    },
+    {
+      "epoch": 45.67,
+      "learning_rate": 0.00028485576923076923,
+      "loss": 0.3549,
+      "step": 4978
+    },
+    {
+      "epoch": 45.68,
+      "learning_rate": 0.00028480769230769233,
+      "loss": 0.3199,
+      "step": 4979
+    },
+    {
+      "epoch": 45.69,
+      "learning_rate": 0.0002847596153846154,
+      "loss": 0.2898,
+      "step": 4980
+    },
+    {
+      "epoch": 45.7,
+      "learning_rate": 0.0002847115384615385,
+      "loss": 0.4599,
+      "step": 4981
+    },
+    {
+      "epoch": 45.71,
+      "learning_rate": 0.0002846634615384616,
+      "loss": 0.4376,
+      "step": 4982
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 0.0002846153846153846,
+      "loss": 0.5105,
+      "step": 4983
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 0.0002845673076923077,
+      "loss": 0.7059,
+      "step": 4984
+    },
+    {
+      "epoch": 45.73,
+      "learning_rate": 0.00028451923076923077,
+      "loss": 0.4228,
+      "step": 4985
+    },
+    {
+      "epoch": 45.74,
+      "learning_rate": 0.00028447115384615387,
+      "loss": 0.2437,
+      "step": 4986
+    },
+    {
+      "epoch": 45.75,
+      "learning_rate": 0.00028442307692307697,
+      "loss": 0.4681,
+      "step": 4987
+    },
+    {
+      "epoch": 45.76,
+      "learning_rate": 0.000284375,
+      "loss": 0.4153,
+      "step": 4988
+    },
+    {
+      "epoch": 45.77,
+      "learning_rate": 0.0002843269230769231,
+      "loss": 0.4229,
+      "step": 4989
+    },
+    {
+      "epoch": 45.78,
+      "learning_rate": 0.00028427884615384616,
+      "loss": 0.4097,
+      "step": 4990
+    },
+    {
+      "epoch": 45.79,
+      "learning_rate": 0.00028423076923076926,
+      "loss": 0.4276,
+      "step": 4991
+    },
+    {
+      "epoch": 45.8,
+      "learning_rate": 0.00028418269230769236,
+      "loss": 0.4198,
+      "step": 4992
+    },
+    {
+      "epoch": 45.81,
+      "learning_rate": 0.0002841346153846154,
+      "loss": 0.3687,
+      "step": 4993
+    },
+    {
+      "epoch": 45.82,
+      "learning_rate": 0.0002840865384615385,
+      "loss": 0.3977,
+      "step": 4994
+    },
+    {
+      "epoch": 45.83,
+      "learning_rate": 0.00028403846153846155,
+      "loss": 0.4223,
+      "step": 4995
+    },
+    {
+      "epoch": 45.83,
+      "learning_rate": 0.00028399038461538465,
+      "loss": 0.3047,
+      "step": 4996
+    },
+    {
+      "epoch": 45.84,
+      "learning_rate": 0.00028394230769230775,
+      "loss": 0.3559,
+      "step": 4997
+    },
+    {
+      "epoch": 45.85,
+      "learning_rate": 0.00028389423076923074,
+      "loss": 0.4489,
+      "step": 4998
+    },
+    {
+      "epoch": 45.86,
+      "learning_rate": 0.00028384615384615384,
+      "loss": 0.4004,
+      "step": 4999
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 0.0002837980769230769,
+      "loss": 0.5509,
+      "step": 5000
+    },
+    {
+      "epoch": 45.87,
+      "eval_cer": 0.16917794869488298,
+      "eval_loss": 0.6734281182289124,
+      "eval_runtime": 86.9574,
+      "eval_samples_per_second": 18.94,
+      "eval_steps_per_second": 2.369,
+      "eval_wer": 0.5640894699213563,
+      "step": 5000
+    },
+    {
+      "epoch": 45.88,
+      "learning_rate": 0.00028375,
+      "loss": 0.3888,
+      "step": 5001
+    },
+    {
+      "epoch": 45.89,
+      "learning_rate": 0.0002837019230769231,
+      "loss": 0.3831,
+      "step": 5002
+    },
+    {
+      "epoch": 45.9,
+      "learning_rate": 0.00028365384615384613,
+      "loss": 0.4324,
+      "step": 5003
+    },
+    {
+      "epoch": 45.91,
+      "learning_rate": 0.00028360576923076923,
+      "loss": 0.3735,
+      "step": 5004
+    },
+    {
+      "epoch": 45.92,
+      "learning_rate": 0.0002835576923076923,
+      "loss": 0.355,
+      "step": 5005
+    },
+    {
+      "epoch": 45.93,
+      "learning_rate": 0.0002835096153846154,
+      "loss": 0.3093,
+      "step": 5006
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 0.0002834615384615385,
+      "loss": 0.4335,
+      "step": 5007
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 0.0002834134615384615,
+      "loss": 0.4209,
+      "step": 5008
+    },
+    {
+      "epoch": 45.95,
+      "learning_rate": 0.0002833653846153846,
+      "loss": 0.3284,
+      "step": 5009
+    },
+    {
+      "epoch": 45.96,
+      "learning_rate": 0.00028331730769230767,
+      "loss": 0.4518,
+      "step": 5010
+    },
+    {
+      "epoch": 45.97,
+      "learning_rate": 0.00028326923076923077,
+      "loss": 0.3612,
+      "step": 5011
+    },
+    {
+      "epoch": 45.98,
+      "learning_rate": 0.00028322115384615387,
+      "loss": 0.3828,
+      "step": 5012
+    },
+    {
+      "epoch": 45.99,
+      "learning_rate": 0.0002831730769230769,
+      "loss": 0.4417,
+      "step": 5013
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 0.000283125,
+      "loss": 0.3895,
+      "step": 5014
+    },
+    {
+      "epoch": 46.01,
+      "learning_rate": 0.00028307692307692306,
+      "loss": 0.4446,
+      "step": 5015
+    },
+    {
+      "epoch": 46.02,
+      "learning_rate": 0.00028302884615384616,
+      "loss": 0.4012,
+      "step": 5016
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 0.00028298076923076926,
+      "loss": 0.4297,
+      "step": 5017
+    },
+    {
+      "epoch": 46.04,
+      "learning_rate": 0.0002829326923076923,
+      "loss": 0.397,
+      "step": 5018
+    },
+    {
+      "epoch": 46.05,
+      "learning_rate": 0.0002828846153846154,
+      "loss": 0.3459,
+      "step": 5019
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 0.00028283653846153845,
+      "loss": 0.3938,
+      "step": 5020
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 0.00028278846153846155,
+      "loss": 0.3921,
+      "step": 5021
+    },
+    {
+      "epoch": 46.07,
+      "learning_rate": 0.0002827403846153846,
+      "loss": 0.3746,
+      "step": 5022
+    },
+    {
+      "epoch": 46.08,
+      "learning_rate": 0.0002826923076923077,
+      "loss": 0.3089,
+      "step": 5023
+    },
+    {
+      "epoch": 46.09,
+      "learning_rate": 0.0002826442307692308,
+      "loss": 0.3131,
+      "step": 5024
+    },
+    {
+      "epoch": 46.1,
+      "learning_rate": 0.00028259615384615384,
+      "loss": 0.4791,
+      "step": 5025
+    },
+    {
+      "epoch": 46.11,
+      "learning_rate": 0.00028254807692307694,
+      "loss": 0.3578,
+      "step": 5026
+    },
+    {
+      "epoch": 46.12,
+      "learning_rate": 0.0002825,
+      "loss": 0.2623,
+      "step": 5027
+    },
+    {
+      "epoch": 46.13,
+      "learning_rate": 0.0002824519230769231,
+      "loss": 0.3517,
+      "step": 5028
+    },
+    {
+      "epoch": 46.14,
+      "learning_rate": 0.0002824038461538462,
+      "loss": 0.4154,
+      "step": 5029
+    },
+    {
+      "epoch": 46.15,
+      "learning_rate": 0.0002823557692307692,
+      "loss": 0.4284,
+      "step": 5030
+    },
+    {
+      "epoch": 46.16,
+      "learning_rate": 0.0002823076923076923,
+      "loss": 0.3527,
+      "step": 5031
+    },
+    {
+      "epoch": 46.17,
+      "learning_rate": 0.00028225961538461537,
+      "loss": 0.4124,
+      "step": 5032
+    },
+    {
+      "epoch": 46.17,
+      "learning_rate": 0.00028221153846153847,
+      "loss": 0.368,
+      "step": 5033
+    },
+    {
+      "epoch": 46.18,
+      "learning_rate": 0.00028216346153846157,
+      "loss": 0.4023,
+      "step": 5034
+    },
+    {
+      "epoch": 46.19,
+      "learning_rate": 0.0002821153846153846,
+      "loss": 0.3172,
+      "step": 5035
+    },
+    {
+      "epoch": 46.2,
+      "learning_rate": 0.0002820673076923077,
+      "loss": 0.4246,
+      "step": 5036
+    },
+    {
+      "epoch": 46.21,
+      "learning_rate": 0.00028201923076923076,
+      "loss": 0.4521,
+      "step": 5037
+    },
+    {
+      "epoch": 46.22,
+      "learning_rate": 0.00028197115384615386,
+      "loss": 0.3715,
+      "step": 5038
+    },
+    {
+      "epoch": 46.23,
+      "learning_rate": 0.00028192307692307696,
+      "loss": 0.3096,
+      "step": 5039
+    },
+    {
+      "epoch": 46.24,
+      "learning_rate": 0.000281875,
+      "loss": 0.3731,
+      "step": 5040
+    },
+    {
+      "epoch": 46.25,
+      "learning_rate": 0.0002818269230769231,
+      "loss": 0.3649,
+      "step": 5041
+    },
+    {
+      "epoch": 46.26,
+      "learning_rate": 0.00028177884615384615,
+      "loss": 0.4737,
+      "step": 5042
+    },
+    {
+      "epoch": 46.27,
+      "learning_rate": 0.00028173076923076925,
+      "loss": 0.4215,
+      "step": 5043
+    },
+    {
+      "epoch": 46.28,
+      "learning_rate": 0.00028168269230769235,
+      "loss": 0.39,
+      "step": 5044
+    },
+    {
+      "epoch": 46.28,
+      "learning_rate": 0.0002816346153846154,
+      "loss": 0.3045,
+      "step": 5045
+    },
+    {
+      "epoch": 46.29,
+      "learning_rate": 0.0002815865384615385,
+      "loss": 0.3363,
+      "step": 5046
+    },
+    {
+      "epoch": 46.3,
+      "learning_rate": 0.00028153846153846154,
+      "loss": 0.3292,
+      "step": 5047
+    },
+    {
+      "epoch": 46.31,
+      "learning_rate": 0.00028149038461538464,
+      "loss": 0.4194,
+      "step": 5048
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 0.00028144230769230774,
+      "loss": 0.3799,
+      "step": 5049
+    },
+    {
+      "epoch": 46.33,
+      "learning_rate": 0.0002813942307692308,
+      "loss": 0.3763,
+      "step": 5050
+    },
+    {
+      "epoch": 46.34,
+      "learning_rate": 0.0002813461538461539,
+      "loss": 0.4039,
+      "step": 5051
+    },
+    {
+      "epoch": 46.35,
+      "learning_rate": 0.00028129807692307693,
+      "loss": 0.4714,
+      "step": 5052
+    },
+    {
+      "epoch": 46.36,
+      "learning_rate": 0.00028125000000000003,
+      "loss": 0.3689,
+      "step": 5053
+    },
+    {
+      "epoch": 46.37,
+      "learning_rate": 0.0002812019230769231,
+      "loss": 0.3937,
+      "step": 5054
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 0.0002811538461538461,
+      "loss": 0.3976,
+      "step": 5055
+    },
+    {
+      "epoch": 46.39,
+      "learning_rate": 0.0002811057692307692,
+      "loss": 0.4033,
+      "step": 5056
+    },
+    {
+      "epoch": 46.39,
+      "learning_rate": 0.00028105769230769227,
+      "loss": 0.3303,
+      "step": 5057
+    },
+    {
+      "epoch": 46.4,
+      "learning_rate": 0.00028100961538461537,
+      "loss": 0.4399,
+      "step": 5058
+    },
+    {
+      "epoch": 46.41,
+      "learning_rate": 0.00028096153846153847,
+      "loss": 0.3043,
+      "step": 5059
+    },
+    {
+      "epoch": 46.42,
+      "learning_rate": 0.0002809134615384615,
+      "loss": 0.4573,
+      "step": 5060
+    },
+    {
+      "epoch": 46.43,
+      "learning_rate": 0.0002808653846153846,
+      "loss": 0.3213,
+      "step": 5061
+    },
+    {
+      "epoch": 46.44,
+      "learning_rate": 0.00028081730769230766,
+      "loss": 0.4057,
+      "step": 5062
+    },
+    {
+      "epoch": 46.45,
+      "learning_rate": 0.00028076923076923076,
+      "loss": 0.3242,
+      "step": 5063
+    },
+    {
+      "epoch": 46.46,
+      "learning_rate": 0.00028072115384615386,
+      "loss": 0.2332,
+      "step": 5064
+    },
+    {
+      "epoch": 46.47,
+      "learning_rate": 0.0002806730769230769,
+      "loss": 0.3087,
+      "step": 5065
+    },
+    {
+      "epoch": 46.48,
+      "learning_rate": 0.000280625,
+      "loss": 0.4075,
+      "step": 5066
+    },
+    {
+      "epoch": 46.49,
+      "learning_rate": 0.00028057692307692305,
+      "loss": 0.391,
+      "step": 5067
+    },
+    {
+      "epoch": 46.5,
+      "learning_rate": 0.00028052884615384615,
+      "loss": 0.2793,
+      "step": 5068
+    },
+    {
+      "epoch": 46.5,
+      "learning_rate": 0.00028048076923076925,
+      "loss": 0.5856,
+      "step": 5069
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 0.0002804326923076923,
+      "loss": 0.472,
+      "step": 5070
+    },
+    {
+      "epoch": 46.52,
+      "learning_rate": 0.0002803846153846154,
+      "loss": 0.4496,
+      "step": 5071
+    },
+    {
+      "epoch": 46.53,
+      "learning_rate": 0.00028033653846153844,
+      "loss": 0.3737,
+      "step": 5072
+    },
+    {
+      "epoch": 46.54,
+      "learning_rate": 0.00028028846153846154,
+      "loss": 0.3942,
+      "step": 5073
+    },
+    {
+      "epoch": 46.55,
+      "learning_rate": 0.00028024038461538464,
+      "loss": 0.3437,
+      "step": 5074
+    },
+    {
+      "epoch": 46.56,
+      "learning_rate": 0.0002801923076923077,
+      "loss": 0.4744,
+      "step": 5075
+    },
+    {
+      "epoch": 46.57,
+      "learning_rate": 0.0002801442307692308,
+      "loss": 0.4248,
+      "step": 5076
+    },
+    {
+      "epoch": 46.58,
+      "learning_rate": 0.00028009615384615383,
+      "loss": 0.3472,
+      "step": 5077
+    },
+    {
+      "epoch": 46.59,
+      "learning_rate": 0.00028004807692307693,
+      "loss": 0.3462,
+      "step": 5078
+    },
+    {
+      "epoch": 46.6,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 0.4038,
+      "step": 5079
+    },
+    {
+      "epoch": 46.61,
+      "learning_rate": 0.0002799519230769231,
+      "loss": 0.3904,
+      "step": 5080
+    },
+    {
+      "epoch": 46.61,
+      "learning_rate": 0.0002799038461538462,
+      "loss": 0.4471,
+      "step": 5081
+    },
+    {
+      "epoch": 46.62,
+      "learning_rate": 0.0002798557692307692,
+      "loss": 0.4298,
+      "step": 5082
+    },
+    {
+      "epoch": 46.63,
+      "learning_rate": 0.0002798076923076923,
+      "loss": 0.4039,
+      "step": 5083
+    },
+    {
+      "epoch": 46.64,
+      "learning_rate": 0.0002797596153846154,
+      "loss": 0.4064,
+      "step": 5084
+    },
+    {
+      "epoch": 46.65,
+      "learning_rate": 0.00027971153846153847,
+      "loss": 0.4291,
+      "step": 5085
+    },
+    {
+      "epoch": 46.66,
+      "learning_rate": 0.00027966346153846157,
+      "loss": 0.3814,
+      "step": 5086
+    },
+    {
+      "epoch": 46.67,
+      "learning_rate": 0.0002796153846153846,
+      "loss": 0.4094,
+      "step": 5087
+    },
+    {
+      "epoch": 46.68,
+      "learning_rate": 0.0002795673076923077,
+      "loss": 0.3744,
+      "step": 5088
+    },
+    {
+      "epoch": 46.69,
+      "learning_rate": 0.00027951923076923076,
+      "loss": 0.3543,
+      "step": 5089
+    },
+    {
+      "epoch": 46.7,
+      "learning_rate": 0.00027947115384615386,
+      "loss": 0.4287,
+      "step": 5090
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 0.00027942307692307696,
+      "loss": 0.3178,
+      "step": 5091
+    },
+    {
+      "epoch": 46.72,
+      "learning_rate": 0.000279375,
+      "loss": 0.3496,
+      "step": 5092
+    },
+    {
+      "epoch": 46.72,
+      "learning_rate": 0.0002793269230769231,
+      "loss": 0.3134,
+      "step": 5093
+    },
+    {
+      "epoch": 46.73,
+      "learning_rate": 0.00027927884615384615,
+      "loss": 0.473,
+      "step": 5094
+    },
+    {
+      "epoch": 46.74,
+      "learning_rate": 0.00027923076923076925,
+      "loss": 0.3685,
+      "step": 5095
+    },
+    {
+      "epoch": 46.75,
+      "learning_rate": 0.00027918269230769235,
+      "loss": 0.5517,
+      "step": 5096
+    },
+    {
+      "epoch": 46.76,
+      "learning_rate": 0.0002791346153846154,
+      "loss": 0.4848,
+      "step": 5097
+    },
+    {
+      "epoch": 46.77,
+      "learning_rate": 0.0002790865384615385,
+      "loss": 0.4225,
+      "step": 5098
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 0.00027903846153846154,
+      "loss": 0.4517,
+      "step": 5099
+    },
+    {
+      "epoch": 46.79,
+      "learning_rate": 0.00027899038461538464,
+      "loss": 0.4592,
+      "step": 5100
+    },
+    {
+      "epoch": 46.8,
+      "learning_rate": 0.00027894230769230774,
+      "loss": 0.4617,
+      "step": 5101
+    },
+    {
+      "epoch": 46.81,
+      "learning_rate": 0.0002788942307692308,
+      "loss": 0.3873,
+      "step": 5102
+    },
+    {
+      "epoch": 46.82,
+      "learning_rate": 0.0002788461538461539,
+      "loss": 0.3705,
+      "step": 5103
+    },
+    {
+      "epoch": 46.83,
+      "learning_rate": 0.0002787980769230769,
+      "loss": 0.3708,
+      "step": 5104
+    },
+    {
+      "epoch": 46.83,
+      "learning_rate": 0.00027875,
+      "loss": 0.4093,
+      "step": 5105
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 0.0002787019230769231,
+      "loss": 0.4068,
+      "step": 5106
+    },
+    {
+      "epoch": 46.85,
+      "learning_rate": 0.00027865384615384617,
+      "loss": 0.3471,
+      "step": 5107
+    },
+    {
+      "epoch": 46.86,
+      "learning_rate": 0.00027860576923076927,
+      "loss": 0.3554,
+      "step": 5108
+    },
+    {
+      "epoch": 46.87,
+      "learning_rate": 0.00027855769230769226,
+      "loss": 0.4473,
+      "step": 5109
+    },
+    {
+      "epoch": 46.88,
+      "learning_rate": 0.00027850961538461536,
+      "loss": 0.2886,
+      "step": 5110
+    },
+    {
+      "epoch": 46.89,
+      "learning_rate": 0.00027846153846153846,
+      "loss": 0.4372,
+      "step": 5111
+    },
+    {
+      "epoch": 46.9,
+      "learning_rate": 0.0002784134615384615,
+      "loss": 0.4627,
+      "step": 5112
+    },
+    {
+      "epoch": 46.91,
+      "learning_rate": 0.0002783653846153846,
+      "loss": 0.4479,
+      "step": 5113
+    },
+    {
+      "epoch": 46.92,
+      "learning_rate": 0.00027831730769230765,
+      "loss": 0.3453,
+      "step": 5114
+    },
+    {
+      "epoch": 46.93,
+      "learning_rate": 0.00027826923076923075,
+      "loss": 0.3528,
+      "step": 5115
+    },
+    {
+      "epoch": 46.94,
+      "learning_rate": 0.00027822115384615385,
+      "loss": 0.2891,
+      "step": 5116
+    },
+    {
+      "epoch": 46.94,
+      "learning_rate": 0.0002781730769230769,
+      "loss": 0.3645,
+      "step": 5117
+    },
+    {
+      "epoch": 46.95,
+      "learning_rate": 0.000278125,
+      "loss": 0.3354,
+      "step": 5118
+    },
+    {
+      "epoch": 46.96,
+      "learning_rate": 0.00027807692307692304,
+      "loss": 0.2926,
+      "step": 5119
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 0.00027802884615384614,
+      "loss": 0.4076,
+      "step": 5120
+    },
+    {
+      "epoch": 46.98,
+      "learning_rate": 0.00027798076923076924,
+      "loss": 0.3333,
+      "step": 5121
+    },
+    {
+      "epoch": 46.99,
+      "learning_rate": 0.0002779326923076923,
+      "loss": 0.3572,
+      "step": 5122
+    },
+    {
+      "epoch": 47.0,
+      "learning_rate": 0.0002778846153846154,
+      "loss": 0.3971,
+      "step": 5123
+    },
+    {
+      "epoch": 47.01,
+      "learning_rate": 0.00027783653846153843,
+      "loss": 0.4441,
+      "step": 5124
+    },
+    {
+      "epoch": 47.02,
+      "learning_rate": 0.00027778846153846153,
+      "loss": 0.4264,
+      "step": 5125
+    },
+    {
+      "epoch": 47.03,
+      "learning_rate": 0.00027774038461538463,
+      "loss": 0.4171,
+      "step": 5126
+    },
+    {
+      "epoch": 47.04,
+      "learning_rate": 0.0002776923076923077,
+      "loss": 0.4525,
+      "step": 5127
+    },
+    {
+      "epoch": 47.05,
+      "learning_rate": 0.0002776442307692308,
+      "loss": 0.3863,
+      "step": 5128
+    },
+    {
+      "epoch": 47.06,
+      "learning_rate": 0.0002775961538461538,
+      "loss": 0.4749,
+      "step": 5129
+    },
+    {
+      "epoch": 47.06,
+      "learning_rate": 0.0002775480769230769,
+      "loss": 0.3417,
+      "step": 5130
+    },
+    {
+      "epoch": 47.07,
+      "learning_rate": 0.0002775,
+      "loss": 0.3778,
+      "step": 5131
+    },
+    {
+      "epoch": 47.08,
+      "learning_rate": 0.00027745192307692307,
+      "loss": 0.3095,
+      "step": 5132
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 0.00027740384615384617,
+      "loss": 0.3982,
+      "step": 5133
+    },
+    {
+      "epoch": 47.1,
+      "learning_rate": 0.0002773557692307692,
+      "loss": 0.4025,
+      "step": 5134
+    },
+    {
+      "epoch": 47.11,
+      "learning_rate": 0.0002773076923076923,
+      "loss": 0.3777,
+      "step": 5135
+    },
+    {
+      "epoch": 47.12,
+      "learning_rate": 0.0002772596153846154,
+      "loss": 0.2694,
+      "step": 5136
+    },
+    {
+      "epoch": 47.13,
+      "learning_rate": 0.00027721153846153846,
+      "loss": 0.433,
+      "step": 5137
+    },
+    {
+      "epoch": 47.14,
+      "learning_rate": 0.00027716346153846156,
+      "loss": 0.36,
+      "step": 5138
+    },
+    {
+      "epoch": 47.15,
+      "learning_rate": 0.0002771153846153846,
+      "loss": 0.2639,
+      "step": 5139
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 0.0002770673076923077,
+      "loss": 0.2779,
+      "step": 5140
+    },
+    {
+      "epoch": 47.17,
+      "learning_rate": 0.0002770192307692308,
+      "loss": 0.4339,
+      "step": 5141
+    },
+    {
+      "epoch": 47.17,
+      "learning_rate": 0.00027697115384615385,
+      "loss": 0.4061,
+      "step": 5142
+    },
+    {
+      "epoch": 47.18,
+      "learning_rate": 0.00027692307692307695,
+      "loss": 0.3951,
+      "step": 5143
+    },
+    {
+      "epoch": 47.19,
+      "learning_rate": 0.000276875,
+      "loss": 0.3808,
+      "step": 5144
+    },
+    {
+      "epoch": 47.2,
+      "learning_rate": 0.0002768269230769231,
+      "loss": 0.3595,
+      "step": 5145
+    },
+    {
+      "epoch": 47.21,
+      "learning_rate": 0.0002767788461538462,
+      "loss": 0.4527,
+      "step": 5146
+    },
+    {
+      "epoch": 47.22,
+      "learning_rate": 0.00027673076923076924,
+      "loss": 0.3315,
+      "step": 5147
+    },
+    {
+      "epoch": 47.23,
+      "learning_rate": 0.00027668269230769234,
+      "loss": 0.309,
+      "step": 5148
+    },
+    {
+      "epoch": 47.24,
+      "learning_rate": 0.0002766346153846154,
+      "loss": 0.3137,
+      "step": 5149
+    },
+    {
+      "epoch": 47.25,
+      "learning_rate": 0.0002765865384615385,
+      "loss": 0.344,
+      "step": 5150
+    },
+    {
+      "epoch": 47.26,
+      "learning_rate": 0.00027653846153846153,
+      "loss": 0.4731,
+      "step": 5151
+    },
+    {
+      "epoch": 47.27,
+      "learning_rate": 0.00027649038461538463,
+      "loss": 0.4763,
+      "step": 5152
+    },
+    {
+      "epoch": 47.28,
+      "learning_rate": 0.00027644230769230773,
+      "loss": 0.3989,
+      "step": 5153
+    },
+    {
+      "epoch": 47.28,
+      "learning_rate": 0.0002763942307692308,
+      "loss": 0.2878,
+      "step": 5154
+    },
+    {
+      "epoch": 47.29,
+      "learning_rate": 0.0002763461538461539,
+      "loss": 0.3623,
+      "step": 5155
+    },
+    {
+      "epoch": 47.3,
+      "learning_rate": 0.0002762980769230769,
+      "loss": 0.3572,
+      "step": 5156
+    },
+    {
+      "epoch": 47.31,
+      "learning_rate": 0.00027625,
+      "loss": 0.4535,
+      "step": 5157
+    },
+    {
+      "epoch": 47.32,
+      "learning_rate": 0.0002762019230769231,
+      "loss": 0.4066,
+      "step": 5158
+    },
+    {
+      "epoch": 47.33,
+      "learning_rate": 0.00027615384615384617,
+      "loss": 0.3367,
+      "step": 5159
+    },
+    {
+      "epoch": 47.34,
+      "learning_rate": 0.00027610576923076926,
+      "loss": 0.3882,
+      "step": 5160
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 0.0002760576923076923,
+      "loss": 0.3752,
+      "step": 5161
+    },
+    {
+      "epoch": 47.36,
+      "learning_rate": 0.0002760096153846154,
+      "loss": 0.3128,
+      "step": 5162
+    },
+    {
+      "epoch": 47.37,
+      "learning_rate": 0.0002759615384615385,
+      "loss": 0.3819,
+      "step": 5163
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 0.00027591346153846156,
+      "loss": 0.2931,
+      "step": 5164
+    },
+    {
+      "epoch": 47.39,
+      "learning_rate": 0.00027586538461538466,
+      "loss": 0.4542,
+      "step": 5165
+    },
+    {
+      "epoch": 47.39,
+      "learning_rate": 0.00027581730769230765,
+      "loss": 0.3749,
+      "step": 5166
+    },
+    {
+      "epoch": 47.4,
+      "learning_rate": 0.00027576923076923075,
+      "loss": 0.4341,
+      "step": 5167
+    },
+    {
+      "epoch": 47.41,
+      "learning_rate": 0.00027572115384615385,
+      "loss": 0.3215,
+      "step": 5168
+    },
+    {
+      "epoch": 47.42,
+      "learning_rate": 0.0002756730769230769,
+      "loss": 0.3259,
+      "step": 5169
+    },
+    {
+      "epoch": 47.43,
+      "learning_rate": 0.000275625,
+      "loss": 0.3161,
+      "step": 5170
+    },
+    {
+      "epoch": 47.44,
+      "learning_rate": 0.00027557692307692304,
+      "loss": 0.2934,
+      "step": 5171
+    },
+    {
+      "epoch": 47.45,
+      "learning_rate": 0.00027552884615384614,
+      "loss": 0.2444,
+      "step": 5172
+    },
+    {
+      "epoch": 47.46,
+      "learning_rate": 0.00027548076923076924,
+      "loss": 0.3765,
+      "step": 5173
+    },
+    {
+      "epoch": 47.47,
+      "learning_rate": 0.0002754326923076923,
+      "loss": 0.371,
+      "step": 5174
+    },
+    {
+      "epoch": 47.48,
+      "learning_rate": 0.0002753846153846154,
+      "loss": 0.4311,
+      "step": 5175
+    },
+    {
+      "epoch": 47.49,
+      "learning_rate": 0.0002753365384615384,
+      "loss": 0.2657,
+      "step": 5176
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 0.0002752884615384615,
+      "loss": 0.3452,
+      "step": 5177
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 0.0002752403846153846,
+      "loss": 0.4611,
+      "step": 5178
+    },
+    {
+      "epoch": 47.51,
+      "learning_rate": 0.00027519230769230767,
+      "loss": 0.4188,
+      "step": 5179
+    },
+    {
+      "epoch": 47.52,
+      "learning_rate": 0.00027514423076923077,
+      "loss": 0.3257,
+      "step": 5180
+    },
+    {
+      "epoch": 47.53,
+      "learning_rate": 0.0002750961538461538,
+      "loss": 0.3559,
+      "step": 5181
+    },
+    {
+      "epoch": 47.54,
+      "learning_rate": 0.0002750480769230769,
+      "loss": 0.3435,
+      "step": 5182
+    },
+    {
+      "epoch": 47.55,
+      "learning_rate": 0.000275,
+      "loss": 0.442,
+      "step": 5183
+    },
+    {
+      "epoch": 47.56,
+      "learning_rate": 0.00027495192307692306,
+      "loss": 0.4373,
+      "step": 5184
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 0.00027490384615384616,
+      "loss": 0.4561,
+      "step": 5185
+    },
+    {
+      "epoch": 47.58,
+      "learning_rate": 0.0002748557692307692,
+      "loss": 0.2792,
+      "step": 5186
+    },
+    {
+      "epoch": 47.59,
+      "learning_rate": 0.0002748076923076923,
+      "loss": 0.3448,
+      "step": 5187
+    },
+    {
+      "epoch": 47.6,
+      "learning_rate": 0.0002747596153846154,
+      "loss": 0.3885,
+      "step": 5188
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 0.00027471153846153845,
+      "loss": 0.3821,
+      "step": 5189
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 0.00027466346153846155,
+      "loss": 0.3969,
+      "step": 5190
+    },
+    {
+      "epoch": 47.62,
+      "learning_rate": 0.0002746153846153846,
+      "loss": 0.3937,
+      "step": 5191
+    },
+    {
+      "epoch": 47.63,
+      "learning_rate": 0.0002745673076923077,
+      "loss": 0.4549,
+      "step": 5192
+    },
+    {
+      "epoch": 47.64,
+      "learning_rate": 0.0002745192307692308,
+      "loss": 0.3509,
+      "step": 5193
+    },
+    {
+      "epoch": 47.65,
+      "learning_rate": 0.00027447115384615384,
+      "loss": 0.4421,
+      "step": 5194
+    },
+    {
+      "epoch": 47.66,
+      "learning_rate": 0.00027442307692307694,
+      "loss": 0.3282,
+      "step": 5195
+    },
+    {
+      "epoch": 47.67,
+      "learning_rate": 0.000274375,
+      "loss": 0.3749,
+      "step": 5196
+    },
+    {
+      "epoch": 47.68,
+      "learning_rate": 0.0002743269230769231,
+      "loss": 0.4063,
+      "step": 5197
+    },
+    {
+      "epoch": 47.69,
+      "learning_rate": 0.0002742788461538462,
+      "loss": 0.3883,
+      "step": 5198
+    },
+    {
+      "epoch": 47.7,
+      "learning_rate": 0.00027423076923076923,
+      "loss": 0.3765,
+      "step": 5199
+    },
+    {
+      "epoch": 47.71,
+      "learning_rate": 0.00027418269230769233,
+      "loss": 0.4622,
+      "step": 5200
+    },
+    {
+      "epoch": 47.72,
+      "learning_rate": 0.0002741346153846154,
+      "loss": 0.3738,
+      "step": 5201
+    },
+    {
+      "epoch": 47.72,
+      "learning_rate": 0.0002740865384615385,
+      "loss": 0.349,
+      "step": 5202
+    },
+    {
+      "epoch": 47.73,
+      "learning_rate": 0.0002740384615384616,
+      "loss": 0.3347,
+      "step": 5203
+    },
+    {
+      "epoch": 47.74,
+      "learning_rate": 0.0002739903846153846,
+      "loss": 0.2683,
+      "step": 5204
+    },
+    {
+      "epoch": 47.75,
+      "learning_rate": 0.0002739423076923077,
+      "loss": 0.3759,
+      "step": 5205
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 0.00027389423076923077,
+      "loss": 0.334,
+      "step": 5206
+    },
+    {
+      "epoch": 47.77,
+      "learning_rate": 0.00027384615384615387,
+      "loss": 0.3934,
+      "step": 5207
+    },
+    {
+      "epoch": 47.78,
+      "learning_rate": 0.00027379807692307697,
+      "loss": 0.3723,
+      "step": 5208
+    },
+    {
+      "epoch": 47.79,
+      "learning_rate": 0.00027375,
+      "loss": 0.4517,
+      "step": 5209
+    },
+    {
+      "epoch": 47.8,
+      "learning_rate": 0.0002737019230769231,
+      "loss": 0.4287,
+      "step": 5210
+    },
+    {
+      "epoch": 47.81,
+      "learning_rate": 0.00027365384615384616,
+      "loss": 0.2626,
+      "step": 5211
+    },
+    {
+      "epoch": 47.82,
+      "learning_rate": 0.00027360576923076926,
+      "loss": 0.4341,
+      "step": 5212
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 0.00027355769230769236,
+      "loss": 0.3206,
+      "step": 5213
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 0.0002735096153846154,
+      "loss": 0.3263,
+      "step": 5214
+    },
+    {
+      "epoch": 47.84,
+      "learning_rate": 0.0002734615384615385,
+      "loss": 0.4317,
+      "step": 5215
+    },
+    {
+      "epoch": 47.85,
+      "learning_rate": 0.00027341346153846155,
+      "loss": 0.4032,
+      "step": 5216
+    },
+    {
+      "epoch": 47.86,
+      "learning_rate": 0.00027336538461538465,
+      "loss": 0.3686,
+      "step": 5217
+    },
+    {
+      "epoch": 47.87,
+      "learning_rate": 0.0002733173076923077,
+      "loss": 0.3628,
+      "step": 5218
+    },
+    {
+      "epoch": 47.88,
+      "learning_rate": 0.0002732692307692308,
+      "loss": 0.2935,
+      "step": 5219
+    },
+    {
+      "epoch": 47.89,
+      "learning_rate": 0.0002732211538461539,
+      "loss": 0.3825,
+      "step": 5220
+    },
+    {
+      "epoch": 47.9,
+      "learning_rate": 0.00027317307692307694,
+      "loss": 0.3221,
+      "step": 5221
+    },
+    {
+      "epoch": 47.91,
+      "learning_rate": 0.00027312500000000004,
+      "loss": 0.2843,
+      "step": 5222
+    },
+    {
+      "epoch": 47.92,
+      "learning_rate": 0.00027307692307692303,
+      "loss": 0.4175,
+      "step": 5223
+    },
+    {
+      "epoch": 47.93,
+      "learning_rate": 0.00027302884615384613,
+      "loss": 0.37,
+      "step": 5224
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 0.00027298076923076923,
+      "loss": 0.326,
+      "step": 5225
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 0.0002729326923076923,
+      "loss": 0.3622,
+      "step": 5226
+    },
+    {
+      "epoch": 47.95,
+      "learning_rate": 0.0002728846153846154,
+      "loss": 0.3207,
+      "step": 5227
+    },
+    {
+      "epoch": 47.96,
+      "learning_rate": 0.0002728365384615384,
+      "loss": 0.3877,
+      "step": 5228
+    },
+    {
+      "epoch": 47.97,
+      "learning_rate": 0.0002727884615384615,
+      "loss": 0.3685,
+      "step": 5229
+    },
+    {
+      "epoch": 47.98,
+      "learning_rate": 0.0002727403846153846,
+      "loss": 0.3214,
+      "step": 5230
+    },
+    {
+      "epoch": 47.99,
+      "learning_rate": 0.00027269230769230767,
+      "loss": 0.441,
+      "step": 5231
+    },
+    {
+      "epoch": 48.0,
+      "learning_rate": 0.00027264423076923077,
+      "loss": 0.4489,
+      "step": 5232
+    },
+    {
+      "epoch": 48.01,
+      "learning_rate": 0.0002725961538461538,
+      "loss": 0.4952,
+      "step": 5233
+    },
+    {
+      "epoch": 48.02,
+      "learning_rate": 0.0002725480769230769,
+      "loss": 0.3835,
+      "step": 5234
+    },
+    {
+      "epoch": 48.03,
+      "learning_rate": 0.0002725,
+      "loss": 0.3878,
+      "step": 5235
+    },
+    {
+      "epoch": 48.04,
+      "learning_rate": 0.00027245192307692306,
+      "loss": 0.373,
+      "step": 5236
+    },
+    {
+      "epoch": 48.05,
+      "learning_rate": 0.00027240384615384616,
+      "loss": 0.3813,
+      "step": 5237
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 0.0002723557692307692,
+      "loss": 0.392,
+      "step": 5238
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 0.0002723076923076923,
+      "loss": 0.3388,
+      "step": 5239
+    },
+    {
+      "epoch": 48.07,
+      "learning_rate": 0.0002722596153846154,
+      "loss": 0.366,
+      "step": 5240
+    },
+    {
+      "epoch": 48.08,
+      "learning_rate": 0.00027221153846153845,
+      "loss": 0.3423,
+      "step": 5241
+    },
+    {
+      "epoch": 48.09,
+      "learning_rate": 0.00027216346153846155,
+      "loss": 0.3254,
+      "step": 5242
+    },
+    {
+      "epoch": 48.1,
+      "learning_rate": 0.0002721153846153846,
+      "loss": 0.4675,
+      "step": 5243
+    },
+    {
+      "epoch": 48.11,
+      "learning_rate": 0.0002720673076923077,
+      "loss": 0.3304,
+      "step": 5244
+    },
+    {
+      "epoch": 48.12,
+      "learning_rate": 0.0002720192307692308,
+      "loss": 0.3527,
+      "step": 5245
+    },
+    {
+      "epoch": 48.13,
+      "learning_rate": 0.00027197115384615384,
+      "loss": 0.4015,
+      "step": 5246
+    },
+    {
+      "epoch": 48.14,
+      "learning_rate": 0.00027192307692307694,
+      "loss": 0.3282,
+      "step": 5247
+    },
+    {
+      "epoch": 48.15,
+      "learning_rate": 0.000271875,
+      "loss": 0.3252,
+      "step": 5248
+    },
+    {
+      "epoch": 48.16,
+      "learning_rate": 0.0002718269230769231,
+      "loss": 0.2708,
+      "step": 5249
+    },
+    {
+      "epoch": 48.17,
+      "learning_rate": 0.0002717788461538462,
+      "loss": 0.4004,
+      "step": 5250
+    },
+    {
+      "epoch": 48.17,
+      "learning_rate": 0.0002717307692307692,
+      "loss": 0.2819,
+      "step": 5251
+    },
+    {
+      "epoch": 48.18,
+      "learning_rate": 0.0002716826923076923,
+      "loss": 0.4306,
+      "step": 5252
+    },
+    {
+      "epoch": 48.19,
+      "learning_rate": 0.00027163461538461537,
+      "loss": 0.3627,
+      "step": 5253
+    },
+    {
+      "epoch": 48.2,
+      "learning_rate": 0.00027158653846153847,
+      "loss": 0.3497,
+      "step": 5254
+    },
+    {
+      "epoch": 48.21,
+      "learning_rate": 0.00027153846153846157,
+      "loss": 0.3913,
+      "step": 5255
+    },
+    {
+      "epoch": 48.22,
+      "learning_rate": 0.0002714903846153846,
+      "loss": 0.407,
+      "step": 5256
+    },
+    {
+      "epoch": 48.23,
+      "learning_rate": 0.0002714423076923077,
+      "loss": 0.4931,
+      "step": 5257
+    },
+    {
+      "epoch": 48.24,
+      "learning_rate": 0.00027139423076923076,
+      "loss": 0.3406,
+      "step": 5258
+    },
+    {
+      "epoch": 48.25,
+      "learning_rate": 0.00027134615384615386,
+      "loss": 0.3391,
+      "step": 5259
+    },
+    {
+      "epoch": 48.26,
+      "learning_rate": 0.00027129807692307696,
+      "loss": 0.4726,
+      "step": 5260
+    },
+    {
+      "epoch": 48.27,
+      "learning_rate": 0.00027125,
+      "loss": 0.4847,
+      "step": 5261
+    },
+    {
+      "epoch": 48.28,
+      "learning_rate": 0.0002712019230769231,
+      "loss": 0.3525,
+      "step": 5262
+    },
+    {
+      "epoch": 48.28,
+      "learning_rate": 0.00027115384615384615,
+      "loss": 0.368,
+      "step": 5263
+    },
+    {
+      "epoch": 48.29,
+      "learning_rate": 0.00027110576923076925,
+      "loss": 0.3401,
+      "step": 5264
+    },
+    {
+      "epoch": 48.3,
+      "learning_rate": 0.00027105769230769235,
+      "loss": 0.4008,
+      "step": 5265
+    },
+    {
+      "epoch": 48.31,
+      "learning_rate": 0.0002710096153846154,
+      "loss": 0.4245,
+      "step": 5266
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 0.0002709615384615385,
+      "loss": 0.3147,
+      "step": 5267
+    },
+    {
+      "epoch": 48.33,
+      "learning_rate": 0.00027091346153846154,
+      "loss": 0.2993,
+      "step": 5268
+    },
+    {
+      "epoch": 48.34,
+      "learning_rate": 0.00027086538461538464,
+      "loss": 0.3741,
+      "step": 5269
+    },
+    {
+      "epoch": 48.35,
+      "learning_rate": 0.00027081730769230774,
+      "loss": 0.3571,
+      "step": 5270
+    },
+    {
+      "epoch": 48.36,
+      "learning_rate": 0.0002707692307692308,
+      "loss": 0.3472,
+      "step": 5271
+    },
+    {
+      "epoch": 48.37,
+      "learning_rate": 0.0002707211538461539,
+      "loss": 0.3166,
+      "step": 5272
+    },
+    {
+      "epoch": 48.38,
+      "learning_rate": 0.00027067307692307693,
+      "loss": 0.3184,
+      "step": 5273
+    },
+    {
+      "epoch": 48.39,
+      "learning_rate": 0.00027062500000000003,
+      "loss": 0.3136,
+      "step": 5274
+    },
+    {
+      "epoch": 48.39,
+      "learning_rate": 0.00027057692307692313,
+      "loss": 0.3517,
+      "step": 5275
+    },
+    {
+      "epoch": 48.4,
+      "learning_rate": 0.0002705288461538462,
+      "loss": 0.3414,
+      "step": 5276
+    },
+    {
+      "epoch": 48.41,
+      "learning_rate": 0.0002704807692307693,
+      "loss": 0.3009,
+      "step": 5277
+    },
+    {
+      "epoch": 48.42,
+      "learning_rate": 0.0002704326923076923,
+      "loss": 0.3711,
+      "step": 5278
+    },
+    {
+      "epoch": 48.43,
+      "learning_rate": 0.0002703846153846154,
+      "loss": 0.316,
+      "step": 5279
+    },
+    {
+      "epoch": 48.44,
+      "learning_rate": 0.00027033653846153847,
+      "loss": 0.3671,
+      "step": 5280
+    },
+    {
+      "epoch": 48.45,
+      "learning_rate": 0.0002702884615384615,
+      "loss": 0.3314,
+      "step": 5281
+    },
+    {
+      "epoch": 48.46,
+      "learning_rate": 0.0002702403846153846,
+      "loss": 0.2374,
+      "step": 5282
+    },
+    {
+      "epoch": 48.47,
+      "learning_rate": 0.00027019230769230766,
+      "loss": 0.4459,
+      "step": 5283
+    },
+    {
+      "epoch": 48.48,
+      "learning_rate": 0.00027014423076923076,
+      "loss": 0.3386,
+      "step": 5284
+    },
+    {
+      "epoch": 48.49,
+      "learning_rate": 0.0002700961538461538,
+      "loss": 0.3694,
+      "step": 5285
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 0.0002700480769230769,
+      "loss": 0.3307,
+      "step": 5286
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 0.00027,
+      "loss": 0.4459,
+      "step": 5287
+    },
+    {
+      "epoch": 48.51,
+      "learning_rate": 0.00026995192307692305,
+      "loss": 0.3871,
+      "step": 5288
+    },
+    {
+      "epoch": 48.52,
+      "learning_rate": 0.00026990384615384615,
+      "loss": 0.486,
+      "step": 5289
+    },
+    {
+      "epoch": 48.53,
+      "learning_rate": 0.0002698557692307692,
+      "loss": 0.3814,
+      "step": 5290
+    },
+    {
+      "epoch": 48.54,
+      "learning_rate": 0.0002698076923076923,
+      "loss": 0.3841,
+      "step": 5291
+    },
+    {
+      "epoch": 48.55,
+      "learning_rate": 0.0002697596153846154,
+      "loss": 0.3606,
+      "step": 5292
+    },
+    {
+      "epoch": 48.56,
+      "learning_rate": 0.00026971153846153844,
+      "loss": 0.2969,
+      "step": 5293
+    },
+    {
+      "epoch": 48.57,
+      "learning_rate": 0.00026966346153846154,
+      "loss": 0.3168,
+      "step": 5294
+    },
+    {
+      "epoch": 48.58,
+      "learning_rate": 0.0002696153846153846,
+      "loss": 0.4661,
+      "step": 5295
+    },
+    {
+      "epoch": 48.59,
+      "learning_rate": 0.0002695673076923077,
+      "loss": 0.37,
+      "step": 5296
+    },
+    {
+      "epoch": 48.6,
+      "learning_rate": 0.0002695192307692308,
+      "loss": 0.351,
+      "step": 5297
+    },
+    {
+      "epoch": 48.61,
+      "learning_rate": 0.00026947115384615383,
+      "loss": 0.3709,
+      "step": 5298
+    },
+    {
+      "epoch": 48.61,
+      "learning_rate": 0.00026942307692307693,
+      "loss": 0.4384,
+      "step": 5299
+    },
+    {
+      "epoch": 48.62,
+      "learning_rate": 0.000269375,
+      "loss": 0.3322,
+      "step": 5300
+    },
+    {
+      "epoch": 48.63,
+      "learning_rate": 0.0002693269230769231,
+      "loss": 0.3552,
+      "step": 5301
+    },
+    {
+      "epoch": 48.64,
+      "learning_rate": 0.0002692788461538462,
+      "loss": 0.3444,
+      "step": 5302
+    },
+    {
+      "epoch": 48.65,
+      "learning_rate": 0.0002692307692307692,
+      "loss": 0.244,
+      "step": 5303
+    },
+    {
+      "epoch": 48.66,
+      "learning_rate": 0.0002691826923076923,
+      "loss": 0.3176,
+      "step": 5304
+    },
+    {
+      "epoch": 48.67,
+      "learning_rate": 0.00026913461538461536,
+      "loss": 0.4157,
+      "step": 5305
+    },
+    {
+      "epoch": 48.68,
+      "learning_rate": 0.00026908653846153846,
+      "loss": 0.3408,
+      "step": 5306
+    },
+    {
+      "epoch": 48.69,
+      "learning_rate": 0.00026903846153846156,
+      "loss": 0.4071,
+      "step": 5307
+    },
+    {
+      "epoch": 48.7,
+      "learning_rate": 0.0002689903846153846,
+      "loss": 0.3582,
+      "step": 5308
+    },
+    {
+      "epoch": 48.71,
+      "learning_rate": 0.0002689423076923077,
+      "loss": 0.3899,
+      "step": 5309
+    },
+    {
+      "epoch": 48.72,
+      "learning_rate": 0.00026889423076923076,
+      "loss": 0.3287,
+      "step": 5310
+    },
+    {
+      "epoch": 48.72,
+      "learning_rate": 0.00026884615384615385,
+      "loss": 0.3203,
+      "step": 5311
+    },
+    {
+      "epoch": 48.73,
+      "learning_rate": 0.00026879807692307695,
+      "loss": 0.3695,
+      "step": 5312
+    },
+    {
+      "epoch": 48.74,
+      "learning_rate": 0.00026875,
+      "loss": 0.3471,
+      "step": 5313
+    },
+    {
+      "epoch": 48.75,
+      "learning_rate": 0.0002687019230769231,
+      "loss": 0.3904,
+      "step": 5314
+    },
+    {
+      "epoch": 48.76,
+      "learning_rate": 0.00026865384615384615,
+      "loss": 0.4579,
+      "step": 5315
+    },
+    {
+      "epoch": 48.77,
+      "learning_rate": 0.00026860576923076925,
+      "loss": 0.4093,
+      "step": 5316
+    },
+    {
+      "epoch": 48.78,
+      "learning_rate": 0.00026855769230769234,
+      "loss": 0.3287,
+      "step": 5317
+    },
+    {
+      "epoch": 48.79,
+      "learning_rate": 0.0002685096153846154,
+      "loss": 0.3656,
+      "step": 5318
+    },
+    {
+      "epoch": 48.8,
+      "learning_rate": 0.0002684615384615385,
+      "loss": 0.3457,
+      "step": 5319
+    },
+    {
+      "epoch": 48.81,
+      "learning_rate": 0.00026841346153846154,
+      "loss": 0.3845,
+      "step": 5320
+    },
+    {
+      "epoch": 48.82,
+      "learning_rate": 0.00026836538461538464,
+      "loss": 0.2929,
+      "step": 5321
+    },
+    {
+      "epoch": 48.83,
+      "learning_rate": 0.00026831730769230774,
+      "loss": 0.3222,
+      "step": 5322
+    },
+    {
+      "epoch": 48.83,
+      "learning_rate": 0.0002682692307692308,
+      "loss": 0.3729,
+      "step": 5323
+    },
+    {
+      "epoch": 48.84,
+      "learning_rate": 0.0002682211538461539,
+      "loss": 0.3257,
+      "step": 5324
+    },
+    {
+      "epoch": 48.85,
+      "learning_rate": 0.0002681730769230769,
+      "loss": 0.3844,
+      "step": 5325
+    },
+    {
+      "epoch": 48.86,
+      "learning_rate": 0.000268125,
+      "loss": 0.3667,
+      "step": 5326
+    },
+    {
+      "epoch": 48.87,
+      "learning_rate": 0.0002680769230769231,
+      "loss": 0.3627,
+      "step": 5327
+    },
+    {
+      "epoch": 48.88,
+      "learning_rate": 0.00026802884615384617,
+      "loss": 0.3646,
+      "step": 5328
+    },
+    {
+      "epoch": 48.89,
+      "learning_rate": 0.00026798076923076927,
+      "loss": 0.3945,
+      "step": 5329
+    },
+    {
+      "epoch": 48.9,
+      "learning_rate": 0.0002679326923076923,
+      "loss": 0.318,
+      "step": 5330
+    },
+    {
+      "epoch": 48.91,
+      "learning_rate": 0.0002678846153846154,
+      "loss": 0.3435,
+      "step": 5331
+    },
+    {
+      "epoch": 48.92,
+      "learning_rate": 0.0002678365384615385,
+      "loss": 0.3711,
+      "step": 5332
+    },
+    {
+      "epoch": 48.93,
+      "learning_rate": 0.00026778846153846156,
+      "loss": 0.5436,
+      "step": 5333
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 0.00026774038461538466,
+      "loss": 0.2536,
+      "step": 5334
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 0.0002676923076923077,
+      "loss": 0.3591,
+      "step": 5335
+    },
+    {
+      "epoch": 48.95,
+      "learning_rate": 0.00026764423076923075,
+      "loss": 0.2884,
+      "step": 5336
+    },
+    {
+      "epoch": 48.96,
+      "learning_rate": 0.00026759615384615385,
+      "loss": 0.3651,
+      "step": 5337
+    },
+    {
+      "epoch": 48.97,
+      "learning_rate": 0.0002675480769230769,
+      "loss": 0.389,
+      "step": 5338
+    },
+    {
+      "epoch": 48.98,
+      "learning_rate": 0.0002675,
+      "loss": 0.4454,
+      "step": 5339
+    },
+    {
+      "epoch": 48.99,
+      "learning_rate": 0.00026745192307692304,
+      "loss": 0.3131,
+      "step": 5340
+    },
+    {
+      "epoch": 49.0,
+      "learning_rate": 0.00026740384615384614,
+      "loss": 0.3728,
+      "step": 5341
+    },
+    {
+      "epoch": 49.01,
+      "learning_rate": 0.00026735576923076924,
+      "loss": 0.4842,
+      "step": 5342
+    },
+    {
+      "epoch": 49.02,
+      "learning_rate": 0.0002673076923076923,
+      "loss": 0.3755,
+      "step": 5343
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 0.0002672596153846154,
+      "loss": 0.3231,
+      "step": 5344
+    },
+    {
+      "epoch": 49.04,
+      "learning_rate": 0.00026721153846153843,
+      "loss": 0.3314,
+      "step": 5345
+    },
+    {
+      "epoch": 49.05,
+      "learning_rate": 0.00026716346153846153,
+      "loss": 0.4106,
+      "step": 5346
+    },
+    {
+      "epoch": 49.06,
+      "learning_rate": 0.0002671153846153846,
+      "loss": 0.3198,
+      "step": 5347
+    },
+    {
+      "epoch": 49.06,
+      "learning_rate": 0.0002670673076923077,
+      "loss": 0.3529,
+      "step": 5348
+    },
+    {
+      "epoch": 49.07,
+      "learning_rate": 0.0002670192307692308,
+      "loss": 0.3393,
+      "step": 5349
+    },
+    {
+      "epoch": 49.08,
+      "learning_rate": 0.0002669711538461538,
+      "loss": 0.4194,
+      "step": 5350
+    },
+    {
+      "epoch": 49.09,
+      "learning_rate": 0.0002669230769230769,
+      "loss": 0.4162,
+      "step": 5351
+    },
+    {
+      "epoch": 49.1,
+      "learning_rate": 0.00026687499999999997,
+      "loss": 0.3472,
+      "step": 5352
+    },
+    {
+      "epoch": 49.11,
+      "learning_rate": 0.00026682692307692307,
+      "loss": 0.3643,
+      "step": 5353
+    },
+    {
+      "epoch": 49.12,
+      "learning_rate": 0.00026677884615384617,
+      "loss": 0.3402,
+      "step": 5354
+    },
+    {
+      "epoch": 49.13,
+      "learning_rate": 0.0002667307692307692,
+      "loss": 0.3144,
+      "step": 5355
+    },
+    {
+      "epoch": 49.14,
+      "learning_rate": 0.0002666826923076923,
+      "loss": 0.3507,
+      "step": 5356
+    },
+    {
+      "epoch": 49.15,
+      "learning_rate": 0.00026663461538461536,
+      "loss": 0.36,
+      "step": 5357
+    },
+    {
+      "epoch": 49.16,
+      "learning_rate": 0.00026658653846153846,
+      "loss": 0.3656,
+      "step": 5358
+    },
+    {
+      "epoch": 49.17,
+      "learning_rate": 0.00026653846153846156,
+      "loss": 0.3186,
+      "step": 5359
+    },
+    {
+      "epoch": 49.17,
+      "learning_rate": 0.0002664903846153846,
+      "loss": 0.3029,
+      "step": 5360
+    },
+    {
+      "epoch": 49.18,
+      "learning_rate": 0.0002664423076923077,
+      "loss": 0.285,
+      "step": 5361
+    },
+    {
+      "epoch": 49.19,
+      "learning_rate": 0.00026639423076923075,
+      "loss": 0.2944,
+      "step": 5362
+    },
+    {
+      "epoch": 49.2,
+      "learning_rate": 0.00026634615384615385,
+      "loss": 0.3351,
+      "step": 5363
+    },
+    {
+      "epoch": 49.21,
+      "learning_rate": 0.00026629807692307695,
+      "loss": 0.394,
+      "step": 5364
+    },
+    {
+      "epoch": 49.22,
+      "learning_rate": 0.00026625,
+      "loss": 0.3852,
+      "step": 5365
+    },
+    {
+      "epoch": 49.23,
+      "learning_rate": 0.0002662019230769231,
+      "loss": 0.2428,
+      "step": 5366
+    },
+    {
+      "epoch": 49.24,
+      "learning_rate": 0.00026615384615384614,
+      "loss": 0.3578,
+      "step": 5367
+    },
+    {
+      "epoch": 49.25,
+      "learning_rate": 0.00026610576923076924,
+      "loss": 0.2753,
+      "step": 5368
+    },
+    {
+      "epoch": 49.26,
+      "learning_rate": 0.00026605769230769234,
+      "loss": 0.5079,
+      "step": 5369
+    },
+    {
+      "epoch": 49.27,
+      "learning_rate": 0.0002660096153846154,
+      "loss": 0.4113,
+      "step": 5370
+    },
+    {
+      "epoch": 49.28,
+      "learning_rate": 0.0002659615384615385,
+      "loss": 0.3609,
+      "step": 5371
+    },
+    {
+      "epoch": 49.28,
+      "learning_rate": 0.00026591346153846153,
+      "loss": 0.3407,
+      "step": 5372
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 0.00026586538461538463,
+      "loss": 0.4334,
+      "step": 5373
+    },
+    {
+      "epoch": 49.3,
+      "learning_rate": 0.00026581730769230773,
+      "loss": 0.3041,
+      "step": 5374
+    },
+    {
+      "epoch": 49.31,
+      "learning_rate": 0.0002657692307692308,
+      "loss": 0.4458,
+      "step": 5375
+    },
+    {
+      "epoch": 49.32,
+      "learning_rate": 0.0002657211538461539,
+      "loss": 0.345,
+      "step": 5376
+    },
+    {
+      "epoch": 49.33,
+      "learning_rate": 0.0002656730769230769,
+      "loss": 0.3034,
+      "step": 5377
+    },
+    {
+      "epoch": 49.34,
+      "learning_rate": 0.000265625,
+      "loss": 0.3059,
+      "step": 5378
+    },
+    {
+      "epoch": 49.35,
+      "learning_rate": 0.0002655769230769231,
+      "loss": 0.3622,
+      "step": 5379
+    },
+    {
+      "epoch": 49.36,
+      "learning_rate": 0.00026552884615384616,
+      "loss": 0.3741,
+      "step": 5380
+    },
+    {
+      "epoch": 49.37,
+      "learning_rate": 0.00026548076923076926,
+      "loss": 0.2422,
+      "step": 5381
+    },
+    {
+      "epoch": 49.38,
+      "learning_rate": 0.0002654326923076923,
+      "loss": 0.2212,
+      "step": 5382
+    },
+    {
+      "epoch": 49.39,
+      "learning_rate": 0.0002653846153846154,
+      "loss": 0.3059,
+      "step": 5383
+    },
+    {
+      "epoch": 49.39,
+      "learning_rate": 0.0002653365384615385,
+      "loss": 0.4369,
+      "step": 5384
+    },
+    {
+      "epoch": 49.4,
+      "learning_rate": 0.00026528846153846155,
+      "loss": 0.4014,
+      "step": 5385
+    },
+    {
+      "epoch": 49.41,
+      "learning_rate": 0.00026524038461538465,
+      "loss": 0.2971,
+      "step": 5386
+    },
+    {
+      "epoch": 49.42,
+      "learning_rate": 0.0002651923076923077,
+      "loss": 0.4145,
+      "step": 5387
+    },
+    {
+      "epoch": 49.43,
+      "learning_rate": 0.0002651442307692308,
+      "loss": 0.3591,
+      "step": 5388
+    },
+    {
+      "epoch": 49.44,
+      "learning_rate": 0.0002650961538461539,
+      "loss": 0.3213,
+      "step": 5389
+    },
+    {
+      "epoch": 49.45,
+      "learning_rate": 0.00026504807692307694,
+      "loss": 0.369,
+      "step": 5390
+    },
+    {
+      "epoch": 49.46,
+      "learning_rate": 0.00026500000000000004,
+      "loss": 0.3326,
+      "step": 5391
+    },
+    {
+      "epoch": 49.47,
+      "learning_rate": 0.00026495192307692304,
+      "loss": 0.2522,
+      "step": 5392
+    },
+    {
+      "epoch": 49.48,
+      "learning_rate": 0.00026490384615384614,
+      "loss": 0.4191,
+      "step": 5393
+    },
+    {
+      "epoch": 49.49,
+      "learning_rate": 0.00026485576923076924,
+      "loss": 0.2469,
+      "step": 5394
+    },
+    {
+      "epoch": 49.5,
+      "learning_rate": 0.0002648076923076923,
+      "loss": 0.3533,
+      "step": 5395
+    },
+    {
+      "epoch": 49.5,
+      "learning_rate": 0.0002647596153846154,
+      "loss": 0.3939,
+      "step": 5396
+    },
+    {
+      "epoch": 49.51,
+      "learning_rate": 0.0002647115384615384,
+      "loss": 0.3655,
+      "step": 5397
+    },
+    {
+      "epoch": 49.52,
+      "learning_rate": 0.0002646634615384615,
+      "loss": 0.4315,
+      "step": 5398
+    },
+    {
+      "epoch": 49.53,
+      "learning_rate": 0.0002646153846153846,
+      "loss": 0.4058,
+      "step": 5399
+    },
+    {
+      "epoch": 49.54,
+      "learning_rate": 0.00026456730769230767,
+      "loss": 0.3257,
+      "step": 5400
+    },
+    {
+      "epoch": 49.55,
+      "learning_rate": 0.00026451923076923077,
+      "loss": 0.5259,
+      "step": 5401
+    },
+    {
+      "epoch": 49.56,
+      "learning_rate": 0.0002644711538461538,
+      "loss": 0.365,
+      "step": 5402
+    },
+    {
+      "epoch": 49.57,
+      "learning_rate": 0.0002644230769230769,
+      "loss": 0.3599,
+      "step": 5403
+    },
+    {
+      "epoch": 49.58,
+      "learning_rate": 0.000264375,
+      "loss": 0.343,
+      "step": 5404
+    },
+    {
+      "epoch": 49.59,
+      "learning_rate": 0.00026432692307692306,
+      "loss": 0.2965,
+      "step": 5405
+    },
+    {
+      "epoch": 49.6,
+      "learning_rate": 0.00026427884615384616,
+      "loss": 0.3463,
+      "step": 5406
+    },
+    {
+      "epoch": 49.61,
+      "learning_rate": 0.0002642307692307692,
+      "loss": 0.3411,
+      "step": 5407
+    },
+    {
+      "epoch": 49.61,
+      "learning_rate": 0.0002641826923076923,
+      "loss": 0.3262,
+      "step": 5408
+    },
+    {
+      "epoch": 49.62,
+      "learning_rate": 0.0002641346153846154,
+      "loss": 0.3173,
+      "step": 5409
+    },
+    {
+      "epoch": 49.63,
+      "learning_rate": 0.00026408653846153845,
+      "loss": 0.3436,
+      "step": 5410
+    },
+    {
+      "epoch": 49.64,
+      "learning_rate": 0.00026403846153846155,
+      "loss": 0.3081,
+      "step": 5411
+    },
+    {
+      "epoch": 49.65,
+      "learning_rate": 0.0002639903846153846,
+      "loss": 0.3554,
+      "step": 5412
+    },
+    {
+      "epoch": 49.66,
+      "learning_rate": 0.0002639423076923077,
+      "loss": 0.3486,
+      "step": 5413
+    },
+    {
+      "epoch": 49.67,
+      "learning_rate": 0.00026389423076923074,
+      "loss": 0.3914,
+      "step": 5414
+    },
+    {
+      "epoch": 49.68,
+      "learning_rate": 0.00026384615384615384,
+      "loss": 0.3333,
+      "step": 5415
+    },
+    {
+      "epoch": 49.69,
+      "learning_rate": 0.00026379807692307694,
+      "loss": 0.3253,
+      "step": 5416
+    },
+    {
+      "epoch": 49.7,
+      "learning_rate": 0.00026375,
+      "loss": 0.4104,
+      "step": 5417
+    },
+    {
+      "epoch": 49.71,
+      "learning_rate": 0.0002637019230769231,
+      "loss": 0.433,
+      "step": 5418
+    },
+    {
+      "epoch": 49.72,
+      "learning_rate": 0.00026365384615384613,
+      "loss": 0.3334,
+      "step": 5419
+    },
+    {
+      "epoch": 49.72,
+      "learning_rate": 0.00026360576923076923,
+      "loss": 0.2953,
+      "step": 5420
+    },
+    {
+      "epoch": 49.73,
+      "learning_rate": 0.00026355769230769233,
+      "loss": 0.367,
+      "step": 5421
+    },
+    {
+      "epoch": 49.74,
+      "learning_rate": 0.0002635096153846154,
+      "loss": 0.3327,
+      "step": 5422
+    },
+    {
+      "epoch": 49.75,
+      "learning_rate": 0.0002634615384615385,
+      "loss": 0.4335,
+      "step": 5423
+    },
+    {
+      "epoch": 49.76,
+      "learning_rate": 0.0002634134615384615,
+      "loss": 0.3481,
+      "step": 5424
+    },
+    {
+      "epoch": 49.77,
+      "learning_rate": 0.0002633653846153846,
+      "loss": 0.4251,
+      "step": 5425
+    },
+    {
+      "epoch": 49.78,
+      "learning_rate": 0.0002633173076923077,
+      "loss": 0.3618,
+      "step": 5426
+    },
+    {
+      "epoch": 49.79,
+      "learning_rate": 0.00026326923076923077,
+      "loss": 0.313,
+      "step": 5427
+    },
+    {
+      "epoch": 49.8,
+      "learning_rate": 0.00026322115384615387,
+      "loss": 0.4029,
+      "step": 5428
+    },
+    {
+      "epoch": 49.81,
+      "learning_rate": 0.0002631730769230769,
+      "loss": 0.2829,
+      "step": 5429
+    },
+    {
+      "epoch": 49.82,
+      "learning_rate": 0.000263125,
+      "loss": 0.3279,
+      "step": 5430
+    },
+    {
+      "epoch": 49.83,
+      "learning_rate": 0.0002630769230769231,
+      "loss": 0.2915,
+      "step": 5431
+    },
+    {
+      "epoch": 49.83,
+      "learning_rate": 0.00026302884615384616,
+      "loss": 0.3336,
+      "step": 5432
+    },
+    {
+      "epoch": 49.84,
+      "learning_rate": 0.00026298076923076926,
+      "loss": 0.3781,
+      "step": 5433
+    },
+    {
+      "epoch": 49.85,
+      "learning_rate": 0.0002629326923076923,
+      "loss": 0.2805,
+      "step": 5434
+    },
+    {
+      "epoch": 49.86,
+      "learning_rate": 0.0002628846153846154,
+      "loss": 0.3616,
+      "step": 5435
+    },
+    {
+      "epoch": 49.87,
+      "learning_rate": 0.0002628365384615385,
+      "loss": 0.3278,
+      "step": 5436
+    },
+    {
+      "epoch": 49.88,
+      "learning_rate": 0.00026278846153846155,
+      "loss": 0.3751,
+      "step": 5437
+    },
+    {
+      "epoch": 49.89,
+      "learning_rate": 0.00026274038461538465,
+      "loss": 0.3551,
+      "step": 5438
+    },
+    {
+      "epoch": 49.9,
+      "learning_rate": 0.0002626923076923077,
+      "loss": 0.3164,
+      "step": 5439
+    },
+    {
+      "epoch": 49.91,
+      "learning_rate": 0.0002626442307692308,
+      "loss": 0.3177,
+      "step": 5440
+    },
+    {
+      "epoch": 49.92,
+      "learning_rate": 0.0002625961538461539,
+      "loss": 0.3121,
+      "step": 5441
+    },
+    {
+      "epoch": 49.93,
+      "learning_rate": 0.00026254807692307694,
+      "loss": 0.3186,
+      "step": 5442
+    },
+    {
+      "epoch": 49.94,
+      "learning_rate": 0.00026250000000000004,
+      "loss": 0.46,
+      "step": 5443
+    },
+    {
+      "epoch": 49.94,
+      "learning_rate": 0.0002624519230769231,
+      "loss": 0.2164,
+      "step": 5444
+    },
+    {
+      "epoch": 49.95,
+      "learning_rate": 0.0002624038461538462,
+      "loss": 0.363,
+      "step": 5445
+    },
+    {
+      "epoch": 49.96,
+      "learning_rate": 0.0002623557692307693,
+      "loss": 0.3798,
+      "step": 5446
+    },
+    {
+      "epoch": 49.97,
+      "learning_rate": 0.00026230769230769233,
+      "loss": 0.3829,
+      "step": 5447
+    },
+    {
+      "epoch": 49.98,
+      "learning_rate": 0.00026225961538461543,
+      "loss": 0.4207,
+      "step": 5448
+    },
+    {
+      "epoch": 49.99,
+      "learning_rate": 0.0002622115384615384,
+      "loss": 0.4039,
+      "step": 5449
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 0.0002621634615384615,
+      "loss": 0.3338,
+      "step": 5450
+    },
+    {
+      "epoch": 50.01,
+      "learning_rate": 0.0002621153846153846,
+      "loss": 0.5574,
+      "step": 5451
+    },
+    {
+      "epoch": 50.02,
+      "learning_rate": 0.00026206730769230766,
+      "loss": 0.3786,
+      "step": 5452
+    },
+    {
+      "epoch": 50.03,
+      "learning_rate": 0.00026201923076923076,
+      "loss": 0.4348,
+      "step": 5453
+    },
+    {
+      "epoch": 50.04,
+      "learning_rate": 0.0002619711538461538,
+      "loss": 0.337,
+      "step": 5454
+    },
+    {
+      "epoch": 50.05,
+      "learning_rate": 0.0002619230769230769,
+      "loss": 0.3355,
+      "step": 5455
+    },
+    {
+      "epoch": 50.06,
+      "learning_rate": 0.000261875,
+      "loss": 0.3104,
+      "step": 5456
+    },
+    {
+      "epoch": 50.06,
+      "learning_rate": 0.00026182692307692305,
+      "loss": 0.3516,
+      "step": 5457
+    },
+    {
+      "epoch": 50.07,
+      "learning_rate": 0.00026177884615384615,
+      "loss": 0.3735,
+      "step": 5458
+    },
+    {
+      "epoch": 50.08,
+      "learning_rate": 0.0002617307692307692,
+      "loss": 0.3838,
+      "step": 5459
+    },
+    {
+      "epoch": 50.09,
+      "learning_rate": 0.0002616826923076923,
+      "loss": 0.3471,
+      "step": 5460
+    },
+    {
+      "epoch": 50.1,
+      "learning_rate": 0.0002616346153846154,
+      "loss": 0.2926,
+      "step": 5461
+    },
+    {
+      "epoch": 50.11,
+      "learning_rate": 0.00026158653846153845,
+      "loss": 0.3124,
+      "step": 5462
+    },
+    {
+      "epoch": 50.12,
+      "learning_rate": 0.00026153846153846154,
+      "loss": 0.4281,
+      "step": 5463
+    },
+    {
+      "epoch": 50.13,
+      "learning_rate": 0.0002614903846153846,
+      "loss": 0.4085,
+      "step": 5464
+    },
+    {
+      "epoch": 50.14,
+      "learning_rate": 0.0002614423076923077,
+      "loss": 0.349,
+      "step": 5465
+    },
+    {
+      "epoch": 50.15,
+      "learning_rate": 0.0002613942307692308,
+      "loss": 0.443,
+      "step": 5466
+    },
+    {
+      "epoch": 50.16,
+      "learning_rate": 0.00026134615384615384,
+      "loss": 0.3384,
+      "step": 5467
+    },
+    {
+      "epoch": 50.17,
+      "learning_rate": 0.00026129807692307693,
+      "loss": 0.4641,
+      "step": 5468
+    },
+    {
+      "epoch": 50.17,
+      "learning_rate": 0.00026125,
+      "loss": 0.4994,
+      "step": 5469
+    },
+    {
+      "epoch": 50.18,
+      "learning_rate": 0.0002612019230769231,
+      "loss": 0.3344,
+      "step": 5470
+    },
+    {
+      "epoch": 50.19,
+      "learning_rate": 0.0002611538461538462,
+      "loss": 0.3814,
+      "step": 5471
+    },
+    {
+      "epoch": 50.2,
+      "learning_rate": 0.0002611057692307692,
+      "loss": 0.3348,
+      "step": 5472
+    },
+    {
+      "epoch": 50.21,
+      "learning_rate": 0.0002610576923076923,
+      "loss": 0.296,
+      "step": 5473
+    },
+    {
+      "epoch": 50.22,
+      "learning_rate": 0.00026100961538461537,
+      "loss": 0.4083,
+      "step": 5474
+    },
+    {
+      "epoch": 50.23,
+      "learning_rate": 0.00026096153846153847,
+      "loss": 0.4622,
+      "step": 5475
+    },
+    {
+      "epoch": 50.24,
+      "learning_rate": 0.0002609134615384615,
+      "loss": 0.4029,
+      "step": 5476
+    },
+    {
+      "epoch": 50.25,
+      "learning_rate": 0.0002608653846153846,
+      "loss": 0.3796,
+      "step": 5477
+    },
+    {
+      "epoch": 50.26,
+      "learning_rate": 0.0002608173076923077,
+      "loss": 0.4973,
+      "step": 5478
+    },
+    {
+      "epoch": 50.27,
+      "learning_rate": 0.00026076923076923076,
+      "loss": 0.2983,
+      "step": 5479
+    },
+    {
+      "epoch": 50.28,
+      "learning_rate": 0.00026072115384615386,
+      "loss": 0.3863,
+      "step": 5480
+    },
+    {
+      "epoch": 50.28,
+      "learning_rate": 0.0002606730769230769,
+      "loss": 0.3849,
+      "step": 5481
+    },
+    {
+      "epoch": 50.29,
+      "learning_rate": 0.000260625,
+      "loss": 0.3828,
+      "step": 5482
+    },
+    {
+      "epoch": 50.3,
+      "learning_rate": 0.0002605769230769231,
+      "loss": 0.4103,
+      "step": 5483
+    },
+    {
+      "epoch": 50.31,
+      "learning_rate": 0.00026052884615384615,
+      "loss": 0.3632,
+      "step": 5484
+    },
+    {
+      "epoch": 50.32,
+      "learning_rate": 0.00026048076923076925,
+      "loss": 0.3269,
+      "step": 5485
+    },
+    {
+      "epoch": 50.33,
+      "learning_rate": 0.0002604326923076923,
+      "loss": 0.2507,
+      "step": 5486
+    },
+    {
+      "epoch": 50.34,
+      "learning_rate": 0.0002603846153846154,
+      "loss": 0.3126,
+      "step": 5487
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 0.0002603365384615385,
+      "loss": 0.3204,
+      "step": 5488
+    },
+    {
+      "epoch": 50.36,
+      "learning_rate": 0.00026028846153846154,
+      "loss": 0.3265,
+      "step": 5489
+    },
+    {
+      "epoch": 50.37,
+      "learning_rate": 0.00026024038461538464,
+      "loss": 0.3247,
+      "step": 5490
+    },
+    {
+      "epoch": 50.38,
+      "learning_rate": 0.0002601923076923077,
+      "loss": 0.2945,
+      "step": 5491
+    },
+    {
+      "epoch": 50.39,
+      "learning_rate": 0.0002601442307692308,
+      "loss": 0.3727,
+      "step": 5492
+    },
+    {
+      "epoch": 50.39,
+      "learning_rate": 0.0002600961538461539,
+      "loss": 0.3332,
+      "step": 5493
+    },
+    {
+      "epoch": 50.4,
+      "learning_rate": 0.00026004807692307693,
+      "loss": 0.3431,
+      "step": 5494
+    },
+    {
+      "epoch": 50.41,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 0.4109,
+      "step": 5495
+    },
+    {
+      "epoch": 50.42,
+      "learning_rate": 0.0002599519230769231,
+      "loss": 0.3601,
+      "step": 5496
+    },
+    {
+      "epoch": 50.43,
+      "learning_rate": 0.0002599038461538462,
+      "loss": 0.4189,
+      "step": 5497
+    },
+    {
+      "epoch": 50.44,
+      "learning_rate": 0.0002598557692307693,
+      "loss": 0.4123,
+      "step": 5498
+    },
+    {
+      "epoch": 50.45,
+      "learning_rate": 0.0002598076923076923,
+      "loss": 0.3398,
+      "step": 5499
+    },
+    {
+      "epoch": 50.46,
+      "learning_rate": 0.0002597596153846154,
+      "loss": 0.3665,
+      "step": 5500
+    },
+    {
+      "epoch": 50.46,
+      "eval_cer": 0.16802350859832982,
+      "eval_loss": 0.7018359899520874,
+      "eval_runtime": 87.2276,
+      "eval_samples_per_second": 18.882,
+      "eval_steps_per_second": 2.362,
+      "eval_wer": 0.5662342968031866,
+      "step": 5500
+    },
+    {
+      "epoch": 50.47,
+      "learning_rate": 0.00025971153846153847,
+      "loss": 0.3501,
+      "step": 5501
+    },
+    {
+      "epoch": 50.48,
+      "learning_rate": 0.00025966346153846157,
+      "loss": 0.3998,
+      "step": 5502
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 0.00025961538461538467,
+      "loss": 0.3749,
+      "step": 5503
+    },
+    {
+      "epoch": 50.5,
+      "learning_rate": 0.0002595673076923077,
+      "loss": 0.3605,
+      "step": 5504
+    },
+    {
+      "epoch": 50.5,
+      "learning_rate": 0.0002595192307692308,
+      "loss": 0.4156,
+      "step": 5505
+    },
+    {
+      "epoch": 50.51,
+      "learning_rate": 0.0002594711538461538,
+      "loss": 0.3069,
+      "step": 5506
+    },
+    {
+      "epoch": 50.52,
+      "learning_rate": 0.0002594230769230769,
+      "loss": 0.3842,
+      "step": 5507
+    },
+    {
+      "epoch": 50.53,
+      "learning_rate": 0.000259375,
+      "loss": 0.3743,
+      "step": 5508
+    },
+    {
+      "epoch": 50.54,
+      "learning_rate": 0.00025932692307692305,
+      "loss": 0.5212,
+      "step": 5509
+    },
+    {
+      "epoch": 50.55,
+      "learning_rate": 0.00025927884615384615,
+      "loss": 0.3845,
+      "step": 5510
+    },
+    {
+      "epoch": 50.56,
+      "learning_rate": 0.0002592307692307692,
+      "loss": 0.3691,
+      "step": 5511
+    },
+    {
+      "epoch": 50.57,
+      "learning_rate": 0.0002591826923076923,
+      "loss": 0.3612,
+      "step": 5512
+    },
+    {
+      "epoch": 50.58,
+      "learning_rate": 0.0002591346153846154,
+      "loss": 0.362,
+      "step": 5513
+    },
+    {
+      "epoch": 50.59,
+      "learning_rate": 0.00025908653846153844,
+      "loss": 0.3468,
+      "step": 5514
+    },
+    {
+      "epoch": 50.6,
+      "learning_rate": 0.00025903846153846154,
+      "loss": 0.3286,
+      "step": 5515
+    },
+    {
+      "epoch": 50.61,
+      "learning_rate": 0.0002589903846153846,
+      "loss": 0.434,
+      "step": 5516
+    },
+    {
+      "epoch": 50.61,
+      "learning_rate": 0.0002589423076923077,
+      "loss": 0.305,
+      "step": 5517
+    },
+    {
+      "epoch": 50.62,
+      "learning_rate": 0.0002588942307692308,
+      "loss": 0.3174,
+      "step": 5518
+    },
+    {
+      "epoch": 50.63,
+      "learning_rate": 0.00025884615384615383,
+      "loss": 0.249,
+      "step": 5519
+    },
+    {
+      "epoch": 50.64,
+      "learning_rate": 0.00025879807692307693,
+      "loss": 0.4898,
+      "step": 5520
+    },
+    {
+      "epoch": 50.65,
+      "learning_rate": 0.00025875,
+      "loss": 0.3147,
+      "step": 5521
+    },
+    {
+      "epoch": 50.66,
+      "learning_rate": 0.0002587019230769231,
+      "loss": 0.2586,
+      "step": 5522
+    },
+    {
+      "epoch": 50.67,
+      "learning_rate": 0.0002586538461538462,
+      "loss": 0.3569,
+      "step": 5523
+    },
+    {
+      "epoch": 50.68,
+      "learning_rate": 0.0002586057692307692,
+      "loss": 0.3102,
+      "step": 5524
+    },
+    {
+      "epoch": 50.69,
+      "learning_rate": 0.0002585576923076923,
+      "loss": 0.3229,
+      "step": 5525
+    },
+    {
+      "epoch": 50.7,
+      "learning_rate": 0.00025850961538461536,
+      "loss": 0.2571,
+      "step": 5526
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 0.00025846153846153846,
+      "loss": 0.3058,
+      "step": 5527
+    },
+    {
+      "epoch": 50.72,
+      "learning_rate": 0.00025841346153846156,
+      "loss": 0.3082,
+      "step": 5528
+    },
+    {
+      "epoch": 50.72,
+      "learning_rate": 0.0002583653846153846,
+      "loss": 0.3875,
+      "step": 5529
+    },
+    {
+      "epoch": 50.73,
+      "learning_rate": 0.0002583173076923077,
+      "loss": 0.3641,
+      "step": 5530
+    },
+    {
+      "epoch": 50.74,
+      "learning_rate": 0.00025826923076923075,
+      "loss": 0.2229,
+      "step": 5531
+    },
+    {
+      "epoch": 50.75,
+      "learning_rate": 0.00025822115384615385,
+      "loss": 0.4225,
+      "step": 5532
+    },
+    {
+      "epoch": 50.76,
+      "learning_rate": 0.00025817307692307695,
+      "loss": 0.4102,
+      "step": 5533
+    },
+    {
+      "epoch": 50.77,
+      "learning_rate": 0.000258125,
+      "loss": 0.3995,
+      "step": 5534
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 0.0002580769230769231,
+      "loss": 0.3316,
+      "step": 5535
+    },
+    {
+      "epoch": 50.79,
+      "learning_rate": 0.00025802884615384614,
+      "loss": 0.3964,
+      "step": 5536
+    },
+    {
+      "epoch": 50.8,
+      "learning_rate": 0.00025798076923076924,
+      "loss": 0.3336,
+      "step": 5537
+    },
+    {
+      "epoch": 50.81,
+      "learning_rate": 0.00025793269230769234,
+      "loss": 0.3038,
+      "step": 5538
+    },
+    {
+      "epoch": 50.82,
+      "learning_rate": 0.0002578846153846154,
+      "loss": 0.3237,
+      "step": 5539
+    },
+    {
+      "epoch": 50.83,
+      "learning_rate": 0.0002578365384615385,
+      "loss": 0.2633,
+      "step": 5540
+    },
+    {
+      "epoch": 50.83,
+      "learning_rate": 0.00025778846153846153,
+      "loss": 0.3583,
+      "step": 5541
+    },
+    {
+      "epoch": 50.84,
+      "learning_rate": 0.00025774038461538463,
+      "loss": 0.38,
+      "step": 5542
+    },
+    {
+      "epoch": 50.85,
+      "learning_rate": 0.0002576923076923077,
+      "loss": 0.383,
+      "step": 5543
+    },
+    {
+      "epoch": 50.86,
+      "learning_rate": 0.0002576442307692308,
+      "loss": 0.3906,
+      "step": 5544
+    },
+    {
+      "epoch": 50.87,
+      "learning_rate": 0.0002575961538461539,
+      "loss": 0.297,
+      "step": 5545
+    },
+    {
+      "epoch": 50.88,
+      "learning_rate": 0.0002575480769230769,
+      "loss": 0.2724,
+      "step": 5546
+    },
+    {
+      "epoch": 50.89,
+      "learning_rate": 0.0002575,
+      "loss": 0.2461,
+      "step": 5547
+    },
+    {
+      "epoch": 50.9,
+      "learning_rate": 0.00025745192307692307,
+      "loss": 0.3794,
+      "step": 5548
+    },
+    {
+      "epoch": 50.91,
+      "learning_rate": 0.00025740384615384617,
+      "loss": 0.3711,
+      "step": 5549
+    },
+    {
+      "epoch": 50.92,
+      "learning_rate": 0.00025735576923076927,
+      "loss": 0.3301,
+      "step": 5550
+    },
+    {
+      "epoch": 50.93,
+      "learning_rate": 0.0002573076923076923,
+      "loss": 0.2903,
+      "step": 5551
+    },
+    {
+      "epoch": 50.94,
+      "learning_rate": 0.0002572596153846154,
+      "loss": 0.3568,
+      "step": 5552
+    },
+    {
+      "epoch": 50.94,
+      "learning_rate": 0.00025721153846153846,
+      "loss": 0.3805,
+      "step": 5553
+    },
+    {
+      "epoch": 50.95,
+      "learning_rate": 0.00025716346153846156,
+      "loss": 0.3467,
+      "step": 5554
+    },
+    {
+      "epoch": 50.96,
+      "learning_rate": 0.00025711538461538466,
+      "loss": 0.3761,
+      "step": 5555
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 0.0002570673076923077,
+      "loss": 0.3328,
+      "step": 5556
+    },
+    {
+      "epoch": 50.98,
+      "learning_rate": 0.0002570192307692308,
+      "loss": 0.3114,
+      "step": 5557
+    },
+    {
+      "epoch": 50.99,
+      "learning_rate": 0.00025697115384615385,
+      "loss": 0.2357,
+      "step": 5558
+    },
+    {
+      "epoch": 51.0,
+      "learning_rate": 0.00025692307692307695,
+      "loss": 0.3508,
+      "step": 5559
+    },
+    {
+      "epoch": 51.01,
+      "learning_rate": 0.00025687500000000005,
+      "loss": 0.3949,
+      "step": 5560
+    },
+    {
+      "epoch": 51.02,
+      "learning_rate": 0.0002568269230769231,
+      "loss": 0.418,
+      "step": 5561
+    },
+    {
+      "epoch": 51.03,
+      "learning_rate": 0.00025677884615384614,
+      "loss": 0.3645,
+      "step": 5562
+    },
+    {
+      "epoch": 51.04,
+      "learning_rate": 0.0002567307692307692,
+      "loss": 0.2905,
+      "step": 5563
+    },
+    {
+      "epoch": 51.05,
+      "learning_rate": 0.0002566826923076923,
+      "loss": 0.376,
+      "step": 5564
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 0.0002566346153846154,
+      "loss": 0.3665,
+      "step": 5565
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 0.00025658653846153843,
+      "loss": 0.3265,
+      "step": 5566
+    },
+    {
+      "epoch": 51.07,
+      "learning_rate": 0.00025653846153846153,
+      "loss": 0.3017,
+      "step": 5567
+    },
+    {
+      "epoch": 51.08,
+      "learning_rate": 0.0002564903846153846,
+      "loss": 0.3518,
+      "step": 5568
+    },
+    {
+      "epoch": 51.09,
+      "learning_rate": 0.0002564423076923077,
+      "loss": 0.3405,
+      "step": 5569
+    },
+    {
+      "epoch": 51.1,
+      "learning_rate": 0.0002563942307692308,
+      "loss": 0.2937,
+      "step": 5570
+    },
+    {
+      "epoch": 51.11,
+      "learning_rate": 0.0002563461538461538,
+      "loss": 0.3028,
+      "step": 5571
+    },
+    {
+      "epoch": 51.12,
+      "learning_rate": 0.0002562980769230769,
+      "loss": 0.3861,
+      "step": 5572
+    },
+    {
+      "epoch": 51.13,
+      "learning_rate": 0.00025624999999999997,
+      "loss": 0.3125,
+      "step": 5573
+    },
+    {
+      "epoch": 51.14,
+      "learning_rate": 0.00025620192307692307,
+      "loss": 0.2887,
+      "step": 5574
+    },
+    {
+      "epoch": 51.15,
+      "learning_rate": 0.00025615384615384617,
+      "loss": 0.2655,
+      "step": 5575
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 0.0002561057692307692,
+      "loss": 0.2526,
+      "step": 5576
+    },
+    {
+      "epoch": 51.17,
+      "learning_rate": 0.0002560576923076923,
+      "loss": 0.3307,
+      "step": 5577
+    },
+    {
+      "epoch": 51.17,
+      "learning_rate": 0.00025600961538461536,
+      "loss": 0.3058,
+      "step": 5578
+    },
+    {
+      "epoch": 51.18,
+      "learning_rate": 0.00025596153846153846,
+      "loss": 0.3552,
+      "step": 5579
+    },
+    {
+      "epoch": 51.19,
+      "learning_rate": 0.00025591346153846156,
+      "loss": 0.3438,
+      "step": 5580
+    },
+    {
+      "epoch": 51.2,
+      "learning_rate": 0.0002558653846153846,
+      "loss": 0.3502,
+      "step": 5581
+    },
+    {
+      "epoch": 51.21,
+      "learning_rate": 0.0002558173076923077,
+      "loss": 0.307,
+      "step": 5582
+    },
+    {
+      "epoch": 51.22,
+      "learning_rate": 0.00025576923076923075,
+      "loss": 0.3685,
+      "step": 5583
+    },
+    {
+      "epoch": 51.23,
+      "learning_rate": 0.00025572115384615385,
+      "loss": 0.4255,
+      "step": 5584
+    },
+    {
+      "epoch": 51.24,
+      "learning_rate": 0.00025567307692307695,
+      "loss": 0.3892,
+      "step": 5585
+    },
+    {
+      "epoch": 51.25,
+      "learning_rate": 0.000255625,
+      "loss": 0.3087,
+      "step": 5586
+    },
+    {
+      "epoch": 51.26,
+      "learning_rate": 0.0002555769230769231,
+      "loss": 0.4453,
+      "step": 5587
+    },
+    {
+      "epoch": 51.27,
+      "learning_rate": 0.00025552884615384614,
+      "loss": 0.483,
+      "step": 5588
+    },
+    {
+      "epoch": 51.28,
+      "learning_rate": 0.00025548076923076924,
+      "loss": 0.3262,
+      "step": 5589
+    },
+    {
+      "epoch": 51.28,
+      "learning_rate": 0.00025543269230769234,
+      "loss": 0.383,
+      "step": 5590
+    },
+    {
+      "epoch": 51.29,
+      "learning_rate": 0.0002553846153846154,
+      "loss": 0.3095,
+      "step": 5591
+    },
+    {
+      "epoch": 51.3,
+      "learning_rate": 0.0002553365384615385,
+      "loss": 0.408,
+      "step": 5592
+    },
+    {
+      "epoch": 51.31,
+      "learning_rate": 0.00025528846153846153,
+      "loss": 0.3235,
+      "step": 5593
+    },
+    {
+      "epoch": 51.32,
+      "learning_rate": 0.00025524038461538463,
+      "loss": 0.3207,
+      "step": 5594
+    },
+    {
+      "epoch": 51.33,
+      "learning_rate": 0.00025519230769230773,
+      "loss": 0.3076,
+      "step": 5595
+    },
+    {
+      "epoch": 51.34,
+      "learning_rate": 0.0002551442307692308,
+      "loss": 0.3269,
+      "step": 5596
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 0.0002550961538461539,
+      "loss": 0.3276,
+      "step": 5597
+    },
+    {
+      "epoch": 51.36,
+      "learning_rate": 0.0002550480769230769,
+      "loss": 0.3029,
+      "step": 5598
+    },
+    {
+      "epoch": 51.37,
+      "learning_rate": 0.000255,
+      "loss": 0.3748,
+      "step": 5599
+    },
+    {
+      "epoch": 51.38,
+      "learning_rate": 0.0002549519230769231,
+      "loss": 0.3343,
+      "step": 5600
+    },
+    {
+      "epoch": 51.39,
+      "learning_rate": 0.00025490384615384616,
+      "loss": 0.2884,
+      "step": 5601
+    },
+    {
+      "epoch": 51.39,
+      "learning_rate": 0.00025485576923076926,
+      "loss": 0.3867,
+      "step": 5602
+    },
+    {
+      "epoch": 51.4,
+      "learning_rate": 0.0002548076923076923,
+      "loss": 0.2441,
+      "step": 5603
+    },
+    {
+      "epoch": 51.41,
+      "learning_rate": 0.0002547596153846154,
+      "loss": 0.4847,
+      "step": 5604
+    },
+    {
+      "epoch": 51.42,
+      "learning_rate": 0.0002547115384615385,
+      "loss": 0.4458,
+      "step": 5605
+    },
+    {
+      "epoch": 51.43,
+      "learning_rate": 0.00025466346153846155,
+      "loss": 0.3863,
+      "step": 5606
+    },
+    {
+      "epoch": 51.44,
+      "learning_rate": 0.00025461538461538465,
+      "loss": 0.3045,
+      "step": 5607
+    },
+    {
+      "epoch": 51.45,
+      "learning_rate": 0.0002545673076923077,
+      "loss": 0.3044,
+      "step": 5608
+    },
+    {
+      "epoch": 51.46,
+      "learning_rate": 0.0002545192307692308,
+      "loss": 0.2963,
+      "step": 5609
+    },
+    {
+      "epoch": 51.47,
+      "learning_rate": 0.00025447115384615384,
+      "loss": 0.2896,
+      "step": 5610
+    },
+    {
+      "epoch": 51.48,
+      "learning_rate": 0.00025442307692307694,
+      "loss": 0.2563,
+      "step": 5611
+    },
+    {
+      "epoch": 51.49,
+      "learning_rate": 0.00025437500000000004,
+      "loss": 0.3073,
+      "step": 5612
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 0.0002543269230769231,
+      "loss": 0.2615,
+      "step": 5613
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 0.0002542788461538462,
+      "loss": 0.4385,
+      "step": 5614
+    },
+    {
+      "epoch": 51.51,
+      "learning_rate": 0.00025423076923076923,
+      "loss": 0.3542,
+      "step": 5615
+    },
+    {
+      "epoch": 51.52,
+      "learning_rate": 0.00025418269230769233,
+      "loss": 0.3805,
+      "step": 5616
+    },
+    {
+      "epoch": 51.53,
+      "learning_rate": 0.00025413461538461543,
+      "loss": 0.4165,
+      "step": 5617
+    },
+    {
+      "epoch": 51.54,
+      "learning_rate": 0.0002540865384615384,
+      "loss": 0.3191,
+      "step": 5618
+    },
+    {
+      "epoch": 51.55,
+      "learning_rate": 0.0002540384615384615,
+      "loss": 0.3177,
+      "step": 5619
+    },
+    {
+      "epoch": 51.56,
+      "learning_rate": 0.00025399038461538457,
+      "loss": 0.3843,
+      "step": 5620
+    },
+    {
+      "epoch": 51.57,
+      "learning_rate": 0.00025394230769230767,
+      "loss": 0.27,
+      "step": 5621
+    },
+    {
+      "epoch": 51.58,
+      "learning_rate": 0.00025389423076923077,
+      "loss": 0.2244,
+      "step": 5622
+    },
+    {
+      "epoch": 51.59,
+      "learning_rate": 0.0002538461538461538,
+      "loss": 0.3146,
+      "step": 5623
+    },
+    {
+      "epoch": 51.6,
+      "learning_rate": 0.0002537980769230769,
+      "loss": 0.4017,
+      "step": 5624
+    },
+    {
+      "epoch": 51.61,
+      "learning_rate": 0.00025374999999999996,
+      "loss": 0.3535,
+      "step": 5625
+    },
+    {
+      "epoch": 51.61,
+      "learning_rate": 0.00025370192307692306,
+      "loss": 0.3166,
+      "step": 5626
+    },
+    {
+      "epoch": 51.62,
+      "learning_rate": 0.00025365384615384616,
+      "loss": 0.2894,
+      "step": 5627
+    },
+    {
+      "epoch": 51.63,
+      "learning_rate": 0.0002536057692307692,
+      "loss": 0.3804,
+      "step": 5628
+    },
+    {
+      "epoch": 51.64,
+      "learning_rate": 0.0002535576923076923,
+      "loss": 0.3888,
+      "step": 5629
+    },
+    {
+      "epoch": 51.65,
+      "learning_rate": 0.00025350961538461535,
+      "loss": 0.3484,
+      "step": 5630
+    },
+    {
+      "epoch": 51.66,
+      "learning_rate": 0.00025346153846153845,
+      "loss": 0.4023,
+      "step": 5631
+    },
+    {
+      "epoch": 51.67,
+      "learning_rate": 0.00025341346153846155,
+      "loss": 0.3894,
+      "step": 5632
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 0.0002533653846153846,
+      "loss": 0.3517,
+      "step": 5633
+    },
+    {
+      "epoch": 51.69,
+      "learning_rate": 0.0002533173076923077,
+      "loss": 0.404,
+      "step": 5634
+    },
+    {
+      "epoch": 51.7,
+      "learning_rate": 0.00025326923076923074,
+      "loss": 0.3688,
+      "step": 5635
+    },
+    {
+      "epoch": 51.71,
+      "learning_rate": 0.00025322115384615384,
+      "loss": 0.2676,
+      "step": 5636
+    },
+    {
+      "epoch": 51.72,
+      "learning_rate": 0.00025317307692307694,
+      "loss": 0.2298,
+      "step": 5637
+    },
+    {
+      "epoch": 51.72,
+      "learning_rate": 0.000253125,
+      "loss": 0.2071,
+      "step": 5638
+    },
+    {
+      "epoch": 51.73,
+      "learning_rate": 0.0002530769230769231,
+      "loss": 0.361,
+      "step": 5639
+    },
+    {
+      "epoch": 51.74,
+      "learning_rate": 0.00025302884615384613,
+      "loss": 0.3587,
+      "step": 5640
+    },
+    {
+      "epoch": 51.75,
+      "learning_rate": 0.00025298076923076923,
+      "loss": 0.4527,
+      "step": 5641
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 0.00025293269230769233,
+      "loss": 0.397,
+      "step": 5642
+    },
+    {
+      "epoch": 51.77,
+      "learning_rate": 0.0002528846153846154,
+      "loss": 0.3943,
+      "step": 5643
+    },
+    {
+      "epoch": 51.78,
+      "learning_rate": 0.0002528365384615385,
+      "loss": 0.3159,
+      "step": 5644
+    },
+    {
+      "epoch": 51.79,
+      "learning_rate": 0.0002527884615384615,
+      "loss": 0.3726,
+      "step": 5645
+    },
+    {
+      "epoch": 51.8,
+      "learning_rate": 0.0002527403846153846,
+      "loss": 0.3352,
+      "step": 5646
+    },
+    {
+      "epoch": 51.81,
+      "learning_rate": 0.0002526923076923077,
+      "loss": 0.3585,
+      "step": 5647
+    },
+    {
+      "epoch": 51.82,
+      "learning_rate": 0.00025264423076923077,
+      "loss": 0.3214,
+      "step": 5648
+    },
+    {
+      "epoch": 51.83,
+      "learning_rate": 0.00025259615384615387,
+      "loss": 0.3326,
+      "step": 5649
+    },
+    {
+      "epoch": 51.83,
+      "learning_rate": 0.0002525480769230769,
+      "loss": 0.3247,
+      "step": 5650
+    },
+    {
+      "epoch": 51.84,
+      "learning_rate": 0.0002525,
+      "loss": 0.3808,
+      "step": 5651
+    },
+    {
+      "epoch": 51.85,
+      "learning_rate": 0.0002524519230769231,
+      "loss": 0.3293,
+      "step": 5652
+    },
+    {
+      "epoch": 51.86,
+      "learning_rate": 0.00025240384615384616,
+      "loss": 0.3465,
+      "step": 5653
+    },
+    {
+      "epoch": 51.87,
+      "learning_rate": 0.00025235576923076926,
+      "loss": 0.2927,
+      "step": 5654
+    },
+    {
+      "epoch": 51.88,
+      "learning_rate": 0.0002523076923076923,
+      "loss": 0.4123,
+      "step": 5655
+    },
+    {
+      "epoch": 51.89,
+      "learning_rate": 0.0002522596153846154,
+      "loss": 0.305,
+      "step": 5656
+    },
+    {
+      "epoch": 51.9,
+      "learning_rate": 0.0002522115384615385,
+      "loss": 0.3639,
+      "step": 5657
+    },
+    {
+      "epoch": 51.91,
+      "learning_rate": 0.00025216346153846155,
+      "loss": 0.3134,
+      "step": 5658
+    },
+    {
+      "epoch": 51.92,
+      "learning_rate": 0.00025211538461538465,
+      "loss": 0.4288,
+      "step": 5659
+    },
+    {
+      "epoch": 51.93,
+      "learning_rate": 0.0002520673076923077,
+      "loss": 0.3568,
+      "step": 5660
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 0.0002520192307692308,
+      "loss": 0.3395,
+      "step": 5661
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 0.0002519711538461539,
+      "loss": 0.2714,
+      "step": 5662
+    },
+    {
+      "epoch": 51.95,
+      "learning_rate": 0.00025192307692307694,
+      "loss": 0.4067,
+      "step": 5663
+    },
+    {
+      "epoch": 51.96,
+      "learning_rate": 0.00025187500000000004,
+      "loss": 0.3914,
+      "step": 5664
+    },
+    {
+      "epoch": 51.97,
+      "learning_rate": 0.0002518269230769231,
+      "loss": 0.4124,
+      "step": 5665
+    },
+    {
+      "epoch": 51.98,
+      "learning_rate": 0.0002517788461538462,
+      "loss": 0.3049,
+      "step": 5666
+    },
+    {
+      "epoch": 51.99,
+      "learning_rate": 0.0002517307692307693,
+      "loss": 0.2016,
+      "step": 5667
+    },
+    {
+      "epoch": 52.0,
+      "learning_rate": 0.00025168269230769233,
+      "loss": 0.2514,
+      "step": 5668
+    },
+    {
+      "epoch": 52.01,
+      "learning_rate": 0.00025163461538461543,
+      "loss": 0.4786,
+      "step": 5669
+    },
+    {
+      "epoch": 52.02,
+      "learning_rate": 0.0002515865384615385,
+      "loss": 0.3999,
+      "step": 5670
+    },
+    {
+      "epoch": 52.03,
+      "learning_rate": 0.00025153846153846157,
+      "loss": 0.3556,
+      "step": 5671
+    },
+    {
+      "epoch": 52.04,
+      "learning_rate": 0.0002514903846153846,
+      "loss": 0.3265,
+      "step": 5672
+    },
+    {
+      "epoch": 52.05,
+      "learning_rate": 0.0002514423076923077,
+      "loss": 0.2622,
+      "step": 5673
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 0.0002513942307692308,
+      "loss": 0.3625,
+      "step": 5674
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 0.0002513461538461538,
+      "loss": 0.33,
+      "step": 5675
+    },
+    {
+      "epoch": 52.07,
+      "learning_rate": 0.0002512980769230769,
+      "loss": 0.4006,
+      "step": 5676
+    },
+    {
+      "epoch": 52.08,
+      "learning_rate": 0.00025124999999999995,
+      "loss": 0.3761,
+      "step": 5677
+    },
+    {
+      "epoch": 52.09,
+      "learning_rate": 0.00025120192307692305,
+      "loss": 0.3106,
+      "step": 5678
+    },
+    {
+      "epoch": 52.1,
+      "learning_rate": 0.00025115384615384615,
+      "loss": 0.3059,
+      "step": 5679
+    },
+    {
+      "epoch": 52.11,
+      "learning_rate": 0.0002511057692307692,
+      "loss": 0.2826,
+      "step": 5680
+    },
+    {
+      "epoch": 52.12,
+      "learning_rate": 0.0002510576923076923,
+      "loss": 0.2483,
+      "step": 5681
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 0.00025100961538461534,
+      "loss": 0.3126,
+      "step": 5682
+    },
+    {
+      "epoch": 52.14,
+      "learning_rate": 0.00025096153846153844,
+      "loss": 0.2589,
+      "step": 5683
+    },
+    {
+      "epoch": 52.15,
+      "learning_rate": 0.00025091346153846154,
+      "loss": 0.2541,
+      "step": 5684
+    },
+    {
+      "epoch": 52.16,
+      "learning_rate": 0.0002508653846153846,
+      "loss": 0.2785,
+      "step": 5685
+    },
+    {
+      "epoch": 52.17,
+      "learning_rate": 0.0002508173076923077,
+      "loss": 0.3135,
+      "step": 5686
+    },
+    {
+      "epoch": 52.17,
+      "learning_rate": 0.00025076923076923073,
+      "loss": 0.4077,
+      "step": 5687
+    },
+    {
+      "epoch": 52.18,
+      "learning_rate": 0.00025072115384615383,
+      "loss": 0.323,
+      "step": 5688
+    },
+    {
+      "epoch": 52.19,
+      "learning_rate": 0.00025067307692307693,
+      "loss": 0.347,
+      "step": 5689
+    },
+    {
+      "epoch": 52.2,
+      "learning_rate": 0.000250625,
+      "loss": 0.4072,
+      "step": 5690
+    },
+    {
+      "epoch": 52.21,
+      "learning_rate": 0.0002505769230769231,
+      "loss": 0.2913,
+      "step": 5691
+    },
+    {
+      "epoch": 52.22,
+      "learning_rate": 0.0002505288461538461,
+      "loss": 0.3764,
+      "step": 5692
+    },
+    {
+      "epoch": 52.23,
+      "learning_rate": 0.0002504807692307692,
+      "loss": 0.332,
+      "step": 5693
+    },
+    {
+      "epoch": 52.24,
+      "learning_rate": 0.0002504326923076923,
+      "loss": 0.3629,
+      "step": 5694
+    },
+    {
+      "epoch": 52.25,
+      "learning_rate": 0.00025038461538461537,
+      "loss": 0.2823,
+      "step": 5695
+    },
+    {
+      "epoch": 52.26,
+      "learning_rate": 0.00025033653846153847,
+      "loss": 0.398,
+      "step": 5696
+    },
+    {
+      "epoch": 52.27,
+      "learning_rate": 0.0002502884615384615,
+      "loss": 0.3946,
+      "step": 5697
+    },
+    {
+      "epoch": 52.28,
+      "learning_rate": 0.0002502403846153846,
+      "loss": 0.3434,
+      "step": 5698
+    },
+    {
+      "epoch": 52.28,
+      "learning_rate": 0.0002501923076923077,
+      "loss": 0.2993,
+      "step": 5699
+    },
+    {
+      "epoch": 52.29,
+      "learning_rate": 0.00025014423076923076,
+      "loss": 0.3477,
+      "step": 5700
+    },
+    {
+      "epoch": 52.3,
+      "learning_rate": 0.00025009615384615386,
+      "loss": 0.3027,
+      "step": 5701
+    },
+    {
+      "epoch": 52.31,
+      "learning_rate": 0.0002500480769230769,
+      "loss": 0.2074,
+      "step": 5702
+    },
+    {
+      "epoch": 52.32,
+      "learning_rate": 0.00025,
+      "loss": 0.3091,
+      "step": 5703
+    },
+    {
+      "epoch": 52.33,
+      "learning_rate": 0.00024995192307692305,
+      "loss": 0.2948,
+      "step": 5704
+    },
+    {
+      "epoch": 52.34,
+      "learning_rate": 0.00024990384615384615,
+      "loss": 0.3565,
+      "step": 5705
+    },
+    {
+      "epoch": 52.35,
+      "learning_rate": 0.00024985576923076925,
+      "loss": 0.2966,
+      "step": 5706
+    },
+    {
+      "epoch": 52.36,
+      "learning_rate": 0.0002498076923076923,
+      "loss": 0.3822,
+      "step": 5707
+    },
+    {
+      "epoch": 52.37,
+      "learning_rate": 0.0002497596153846154,
+      "loss": 0.2698,
+      "step": 5708
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 0.00024971153846153844,
+      "loss": 0.322,
+      "step": 5709
+    },
+    {
+      "epoch": 52.39,
+      "learning_rate": 0.00024966346153846154,
+      "loss": 0.3726,
+      "step": 5710
+    },
+    {
+      "epoch": 52.39,
+      "learning_rate": 0.00024961538461538464,
+      "loss": 0.301,
+      "step": 5711
+    },
+    {
+      "epoch": 52.4,
+      "learning_rate": 0.0002495673076923077,
+      "loss": 0.4157,
+      "step": 5712
+    },
+    {
+      "epoch": 52.41,
+      "learning_rate": 0.0002495192307692308,
+      "loss": 0.3996,
+      "step": 5713
+    },
+    {
+      "epoch": 52.42,
+      "learning_rate": 0.00024947115384615383,
+      "loss": 0.3015,
+      "step": 5714
+    },
+    {
+      "epoch": 52.43,
+      "learning_rate": 0.00024942307692307693,
+      "loss": 0.3323,
+      "step": 5715
+    },
+    {
+      "epoch": 52.44,
+      "learning_rate": 0.00024937500000000003,
+      "loss": 0.361,
+      "step": 5716
+    },
+    {
+      "epoch": 52.45,
+      "learning_rate": 0.0002493269230769231,
+      "loss": 0.3362,
+      "step": 5717
+    },
+    {
+      "epoch": 52.46,
+      "learning_rate": 0.0002492788461538462,
+      "loss": 0.3792,
+      "step": 5718
+    },
+    {
+      "epoch": 52.47,
+      "learning_rate": 0.0002492307692307692,
+      "loss": 0.3524,
+      "step": 5719
+    },
+    {
+      "epoch": 52.48,
+      "learning_rate": 0.0002491826923076923,
+      "loss": 0.2483,
+      "step": 5720
+    },
+    {
+      "epoch": 52.49,
+      "learning_rate": 0.00024913461538461537,
+      "loss": 0.3195,
+      "step": 5721
+    },
+    {
+      "epoch": 52.5,
+      "learning_rate": 0.00024908653846153847,
+      "loss": 0.2507,
+      "step": 5722
+    },
+    {
+      "epoch": 52.5,
+      "learning_rate": 0.00024903846153846157,
+      "loss": 0.3685,
+      "step": 5723
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 0.0002489903846153846,
+      "loss": 0.3643,
+      "step": 5724
+    },
+    {
+      "epoch": 52.52,
+      "learning_rate": 0.0002489423076923077,
+      "loss": 0.4183,
+      "step": 5725
+    },
+    {
+      "epoch": 52.53,
+      "learning_rate": 0.00024889423076923076,
+      "loss": 0.3584,
+      "step": 5726
+    },
+    {
+      "epoch": 52.54,
+      "learning_rate": 0.00024884615384615386,
+      "loss": 0.2686,
+      "step": 5727
+    },
+    {
+      "epoch": 52.55,
+      "learning_rate": 0.00024879807692307696,
+      "loss": 0.3375,
+      "step": 5728
+    },
+    {
+      "epoch": 52.56,
+      "learning_rate": 0.00024875,
+      "loss": 0.2941,
+      "step": 5729
+    },
+    {
+      "epoch": 52.57,
+      "learning_rate": 0.0002487019230769231,
+      "loss": 0.3109,
+      "step": 5730
+    },
+    {
+      "epoch": 52.58,
+      "learning_rate": 0.00024865384615384615,
+      "loss": 0.3215,
+      "step": 5731
+    },
+    {
+      "epoch": 52.59,
+      "learning_rate": 0.00024860576923076925,
+      "loss": 0.3299,
+      "step": 5732
+    },
+    {
+      "epoch": 52.6,
+      "learning_rate": 0.0002485576923076923,
+      "loss": 0.3231,
+      "step": 5733
+    },
+    {
+      "epoch": 52.61,
+      "learning_rate": 0.0002485096153846154,
+      "loss": 0.3487,
+      "step": 5734
+    },
+    {
+      "epoch": 52.61,
+      "learning_rate": 0.00024846153846153844,
+      "loss": 0.3238,
+      "step": 5735
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 0.00024841346153846154,
+      "loss": 0.3829,
+      "step": 5736
+    },
+    {
+      "epoch": 52.63,
+      "learning_rate": 0.00024836538461538464,
+      "loss": 0.3532,
+      "step": 5737
+    },
+    {
+      "epoch": 52.64,
+      "learning_rate": 0.0002483173076923077,
+      "loss": 0.2312,
+      "step": 5738
+    },
+    {
+      "epoch": 52.65,
+      "learning_rate": 0.0002482692307692308,
+      "loss": 0.3255,
+      "step": 5739
+    },
+    {
+      "epoch": 52.66,
+      "learning_rate": 0.00024822115384615383,
+      "loss": 0.3621,
+      "step": 5740
+    },
+    {
+      "epoch": 52.67,
+      "learning_rate": 0.00024817307692307693,
+      "loss": 0.4138,
+      "step": 5741
+    },
+    {
+      "epoch": 52.68,
+      "learning_rate": 0.00024812500000000003,
+      "loss": 0.4041,
+      "step": 5742
+    },
+    {
+      "epoch": 52.69,
+      "learning_rate": 0.0002480769230769231,
+      "loss": 0.3862,
+      "step": 5743
+    },
+    {
+      "epoch": 52.7,
+      "learning_rate": 0.00024802884615384617,
+      "loss": 0.3249,
+      "step": 5744
+    },
+    {
+      "epoch": 52.71,
+      "learning_rate": 0.0002479807692307692,
+      "loss": 0.4402,
+      "step": 5745
+    },
+    {
+      "epoch": 52.72,
+      "learning_rate": 0.0002479326923076923,
+      "loss": 0.3719,
+      "step": 5746
+    },
+    {
+      "epoch": 52.72,
+      "learning_rate": 0.0002478846153846154,
+      "loss": 0.328,
+      "step": 5747
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 0.00024783653846153846,
+      "loss": 0.3491,
+      "step": 5748
+    },
+    {
+      "epoch": 52.74,
+      "learning_rate": 0.00024778846153846156,
+      "loss": 0.2222,
+      "step": 5749
+    },
+    {
+      "epoch": 52.75,
+      "learning_rate": 0.0002477403846153846,
+      "loss": 0.3935,
+      "step": 5750
+    },
+    {
+      "epoch": 52.76,
+      "learning_rate": 0.0002476923076923077,
+      "loss": 0.3967,
+      "step": 5751
+    },
+    {
+      "epoch": 52.77,
+      "learning_rate": 0.00024764423076923075,
+      "loss": 0.456,
+      "step": 5752
+    },
+    {
+      "epoch": 52.78,
+      "learning_rate": 0.00024759615384615385,
+      "loss": 0.3525,
+      "step": 5753
+    },
+    {
+      "epoch": 52.79,
+      "learning_rate": 0.00024754807692307695,
+      "loss": 0.4589,
+      "step": 5754
+    },
+    {
+      "epoch": 52.8,
+      "learning_rate": 0.0002475,
+      "loss": 0.3629,
+      "step": 5755
+    },
+    {
+      "epoch": 52.81,
+      "learning_rate": 0.0002474519230769231,
+      "loss": 0.3881,
+      "step": 5756
+    },
+    {
+      "epoch": 52.82,
+      "learning_rate": 0.00024740384615384614,
+      "loss": 0.3771,
+      "step": 5757
+    },
+    {
+      "epoch": 52.83,
+      "learning_rate": 0.00024735576923076924,
+      "loss": 0.329,
+      "step": 5758
+    },
+    {
+      "epoch": 52.83,
+      "learning_rate": 0.00024730769230769234,
+      "loss": 0.3562,
+      "step": 5759
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 0.0002472596153846154,
+      "loss": 0.3421,
+      "step": 5760
+    },
+    {
+      "epoch": 52.85,
+      "learning_rate": 0.00024721153846153843,
+      "loss": 0.256,
+      "step": 5761
+    },
+    {
+      "epoch": 52.86,
+      "learning_rate": 0.00024716346153846153,
+      "loss": 0.3341,
+      "step": 5762
+    },
+    {
+      "epoch": 52.87,
+      "learning_rate": 0.00024711538461538463,
+      "loss": 0.3429,
+      "step": 5763
+    },
+    {
+      "epoch": 52.88,
+      "learning_rate": 0.0002470673076923077,
+      "loss": 0.4404,
+      "step": 5764
+    },
+    {
+      "epoch": 52.89,
+      "learning_rate": 0.0002470192307692308,
+      "loss": 0.4481,
+      "step": 5765
+    },
+    {
+      "epoch": 52.9,
+      "learning_rate": 0.0002469711538461538,
+      "loss": 0.2681,
+      "step": 5766
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 0.0002469230769230769,
+      "loss": 0.3548,
+      "step": 5767
+    },
+    {
+      "epoch": 52.92,
+      "learning_rate": 0.000246875,
+      "loss": 0.3342,
+      "step": 5768
+    },
+    {
+      "epoch": 52.93,
+      "learning_rate": 0.00024682692307692307,
+      "loss": 0.3222,
+      "step": 5769
+    },
+    {
+      "epoch": 52.94,
+      "learning_rate": 0.00024677884615384617,
+      "loss": 0.348,
+      "step": 5770
+    },
+    {
+      "epoch": 52.94,
+      "learning_rate": 0.0002467307692307692,
+      "loss": 0.2884,
+      "step": 5771
+    },
+    {
+      "epoch": 52.95,
+      "learning_rate": 0.0002466826923076923,
+      "loss": 0.19,
+      "step": 5772
+    },
+    {
+      "epoch": 52.96,
+      "learning_rate": 0.0002466346153846154,
+      "loss": 0.4438,
+      "step": 5773
+    },
+    {
+      "epoch": 52.97,
+      "learning_rate": 0.00024658653846153846,
+      "loss": 0.3512,
+      "step": 5774
+    },
+    {
+      "epoch": 52.98,
+      "learning_rate": 0.00024653846153846156,
+      "loss": 0.3341,
+      "step": 5775
+    },
+    {
+      "epoch": 52.99,
+      "learning_rate": 0.0002464903846153846,
+      "loss": 0.2382,
+      "step": 5776
+    },
+    {
+      "epoch": 53.0,
+      "learning_rate": 0.0002464423076923077,
+      "loss": 0.3677,
+      "step": 5777
+    },
+    {
+      "epoch": 53.01,
+      "learning_rate": 0.0002463942307692308,
+      "loss": 0.4254,
+      "step": 5778
+    },
+    {
+      "epoch": 53.02,
+      "learning_rate": 0.00024634615384615385,
+      "loss": 0.3803,
+      "step": 5779
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 0.00024629807692307695,
+      "loss": 0.3078,
+      "step": 5780
+    },
+    {
+      "epoch": 53.04,
+      "learning_rate": 0.00024625,
+      "loss": 0.403,
+      "step": 5781
+    },
+    {
+      "epoch": 53.05,
+      "learning_rate": 0.0002462019230769231,
+      "loss": 0.4219,
+      "step": 5782
+    },
+    {
+      "epoch": 53.06,
+      "learning_rate": 0.0002461538461538462,
+      "loss": 0.3278,
+      "step": 5783
+    },
+    {
+      "epoch": 53.06,
+      "learning_rate": 0.00024610576923076924,
+      "loss": 0.332,
+      "step": 5784
+    },
+    {
+      "epoch": 53.07,
+      "learning_rate": 0.00024605769230769234,
+      "loss": 0.2756,
+      "step": 5785
+    },
+    {
+      "epoch": 53.08,
+      "learning_rate": 0.0002460096153846154,
+      "loss": 0.3323,
+      "step": 5786
+    },
+    {
+      "epoch": 53.09,
+      "learning_rate": 0.0002459615384615385,
+      "loss": 0.3311,
+      "step": 5787
+    },
+    {
+      "epoch": 53.1,
+      "learning_rate": 0.00024591346153846153,
+      "loss": 0.2914,
+      "step": 5788
+    },
+    {
+      "epoch": 53.11,
+      "learning_rate": 0.00024586538461538463,
+      "loss": 0.3101,
+      "step": 5789
+    },
+    {
+      "epoch": 53.12,
+      "learning_rate": 0.0002458173076923077,
+      "loss": 0.3035,
+      "step": 5790
+    },
+    {
+      "epoch": 53.13,
+      "learning_rate": 0.0002457692307692308,
+      "loss": 0.2834,
+      "step": 5791
+    },
+    {
+      "epoch": 53.14,
+      "learning_rate": 0.0002457211538461538,
+      "loss": 0.2877,
+      "step": 5792
+    },
+    {
+      "epoch": 53.15,
+      "learning_rate": 0.0002456730769230769,
+      "loss": 0.3781,
+      "step": 5793
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 0.000245625,
+      "loss": 0.3451,
+      "step": 5794
+    },
+    {
+      "epoch": 53.17,
+      "learning_rate": 0.00024557692307692307,
+      "loss": 0.228,
+      "step": 5795
+    },
+    {
+      "epoch": 53.17,
+      "learning_rate": 0.00024552884615384617,
+      "loss": 0.3079,
+      "step": 5796
+    },
+    {
+      "epoch": 53.18,
+      "learning_rate": 0.0002454807692307692,
+      "loss": 0.2762,
+      "step": 5797
+    },
+    {
+      "epoch": 53.19,
+      "learning_rate": 0.0002454326923076923,
+      "loss": 0.3318,
+      "step": 5798
+    },
+    {
+      "epoch": 53.2,
+      "learning_rate": 0.0002453846153846154,
+      "loss": 0.3322,
+      "step": 5799
+    },
+    {
+      "epoch": 53.21,
+      "learning_rate": 0.00024533653846153846,
+      "loss": 0.2125,
+      "step": 5800
+    },
+    {
+      "epoch": 53.22,
+      "learning_rate": 0.00024528846153846156,
+      "loss": 0.2393,
+      "step": 5801
+    },
+    {
+      "epoch": 53.23,
+      "learning_rate": 0.0002452403846153846,
+      "loss": 0.2993,
+      "step": 5802
+    },
+    {
+      "epoch": 53.24,
+      "learning_rate": 0.0002451923076923077,
+      "loss": 0.2957,
+      "step": 5803
+    },
+    {
+      "epoch": 53.25,
+      "learning_rate": 0.0002451442307692308,
+      "loss": 0.2108,
+      "step": 5804
+    },
+    {
+      "epoch": 53.26,
+      "learning_rate": 0.00024509615384615385,
+      "loss": 0.4256,
+      "step": 5805
+    },
+    {
+      "epoch": 53.27,
+      "learning_rate": 0.00024504807692307695,
+      "loss": 0.345,
+      "step": 5806
+    },
+    {
+      "epoch": 53.28,
+      "learning_rate": 0.000245,
+      "loss": 0.3007,
+      "step": 5807
+    },
+    {
+      "epoch": 53.28,
+      "learning_rate": 0.0002449519230769231,
+      "loss": 0.3694,
+      "step": 5808
+    },
+    {
+      "epoch": 53.29,
+      "learning_rate": 0.0002449038461538462,
+      "loss": 0.2972,
+      "step": 5809
+    },
+    {
+      "epoch": 53.3,
+      "learning_rate": 0.00024485576923076924,
+      "loss": 0.3604,
+      "step": 5810
+    },
+    {
+      "epoch": 53.31,
+      "learning_rate": 0.00024480769230769234,
+      "loss": 0.3796,
+      "step": 5811
+    },
+    {
+      "epoch": 53.32,
+      "learning_rate": 0.0002447596153846154,
+      "loss": 0.3353,
+      "step": 5812
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 0.0002447115384615385,
+      "loss": 0.3132,
+      "step": 5813
+    },
+    {
+      "epoch": 53.34,
+      "learning_rate": 0.0002446634615384616,
+      "loss": 0.287,
+      "step": 5814
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 0.00024461538461538463,
+      "loss": 0.2565,
+      "step": 5815
+    },
+    {
+      "epoch": 53.36,
+      "learning_rate": 0.0002445673076923077,
+      "loss": 0.2299,
+      "step": 5816
+    },
+    {
+      "epoch": 53.37,
+      "learning_rate": 0.00024451923076923077,
+      "loss": 0.3264,
+      "step": 5817
+    },
+    {
+      "epoch": 53.38,
+      "learning_rate": 0.0002444711538461538,
+      "loss": 0.2438,
+      "step": 5818
+    },
+    {
+      "epoch": 53.39,
+      "learning_rate": 0.0002444230769230769,
+      "loss": 0.2695,
+      "step": 5819
+    },
+    {
+      "epoch": 53.39,
+      "learning_rate": 0.000244375,
+      "loss": 0.2749,
+      "step": 5820
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 0.00024432692307692306,
+      "loss": 0.3656,
+      "step": 5821
+    },
+    {
+      "epoch": 53.41,
+      "learning_rate": 0.00024427884615384616,
+      "loss": 0.2366,
+      "step": 5822
+    },
+    {
+      "epoch": 53.42,
+      "learning_rate": 0.0002442307692307692,
+      "loss": 0.3336,
+      "step": 5823
+    },
+    {
+      "epoch": 53.43,
+      "learning_rate": 0.0002441826923076923,
+      "loss": 0.2902,
+      "step": 5824
+    },
+    {
+      "epoch": 53.44,
+      "learning_rate": 0.0002441346153846154,
+      "loss": 0.2663,
+      "step": 5825
+    },
+    {
+      "epoch": 53.45,
+      "learning_rate": 0.00024408653846153848,
+      "loss": 0.3618,
+      "step": 5826
+    },
+    {
+      "epoch": 53.46,
+      "learning_rate": 0.00024403846153846155,
+      "loss": 0.2831,
+      "step": 5827
+    },
+    {
+      "epoch": 53.47,
+      "learning_rate": 0.00024399038461538463,
+      "loss": 0.2404,
+      "step": 5828
+    },
+    {
+      "epoch": 53.48,
+      "learning_rate": 0.0002439423076923077,
+      "loss": 0.3227,
+      "step": 5829
+    },
+    {
+      "epoch": 53.49,
+      "learning_rate": 0.0002438942307692308,
+      "loss": 0.2805,
+      "step": 5830
+    },
+    {
+      "epoch": 53.5,
+      "learning_rate": 0.00024384615384615384,
+      "loss": 0.2276,
+      "step": 5831
+    },
+    {
+      "epoch": 53.5,
+      "learning_rate": 0.00024379807692307692,
+      "loss": 0.3063,
+      "step": 5832
+    },
+    {
+      "epoch": 53.51,
+      "learning_rate": 0.00024375,
+      "loss": 0.3542,
+      "step": 5833
+    },
+    {
+      "epoch": 53.52,
+      "learning_rate": 0.00024370192307692306,
+      "loss": 0.3258,
+      "step": 5834
+    },
+    {
+      "epoch": 53.53,
+      "learning_rate": 0.00024365384615384616,
+      "loss": 0.2997,
+      "step": 5835
+    },
+    {
+      "epoch": 53.54,
+      "learning_rate": 0.00024360576923076923,
+      "loss": 0.3815,
+      "step": 5836
+    },
+    {
+      "epoch": 53.55,
+      "learning_rate": 0.0002435576923076923,
+      "loss": 0.3441,
+      "step": 5837
+    },
+    {
+      "epoch": 53.56,
+      "learning_rate": 0.00024350961538461538,
+      "loss": 0.2549,
+      "step": 5838
+    },
+    {
+      "epoch": 53.57,
+      "learning_rate": 0.00024346153846153845,
+      "loss": 0.3424,
+      "step": 5839
+    },
+    {
+      "epoch": 53.58,
+      "learning_rate": 0.00024341346153846155,
+      "loss": 0.2917,
+      "step": 5840
+    },
+    {
+      "epoch": 53.59,
+      "learning_rate": 0.00024336538461538462,
+      "loss": 0.3897,
+      "step": 5841
+    },
+    {
+      "epoch": 53.6,
+      "learning_rate": 0.0002433173076923077,
+      "loss": 0.3296,
+      "step": 5842
+    },
+    {
+      "epoch": 53.61,
+      "learning_rate": 0.00024326923076923077,
+      "loss": 0.2724,
+      "step": 5843
+    },
+    {
+      "epoch": 53.61,
+      "learning_rate": 0.00024322115384615384,
+      "loss": 0.1861,
+      "step": 5844
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 0.00024317307692307694,
+      "loss": 0.4202,
+      "step": 5845
+    },
+    {
+      "epoch": 53.63,
+      "learning_rate": 0.00024312500000000001,
+      "loss": 0.3052,
+      "step": 5846
+    },
+    {
+      "epoch": 53.64,
+      "learning_rate": 0.0002430769230769231,
+      "loss": 0.3068,
+      "step": 5847
+    },
+    {
+      "epoch": 53.65,
+      "learning_rate": 0.00024302884615384616,
+      "loss": 0.2598,
+      "step": 5848
+    },
+    {
+      "epoch": 53.66,
+      "learning_rate": 0.00024298076923076923,
+      "loss": 0.2119,
+      "step": 5849
+    },
+    {
+      "epoch": 53.67,
+      "learning_rate": 0.0002429326923076923,
+      "loss": 0.4712,
+      "step": 5850
+    },
+    {
+      "epoch": 53.68,
+      "learning_rate": 0.0002428846153846154,
+      "loss": 0.3094,
+      "step": 5851
+    },
+    {
+      "epoch": 53.69,
+      "learning_rate": 0.00024283653846153848,
+      "loss": 0.2813,
+      "step": 5852
+    },
+    {
+      "epoch": 53.7,
+      "learning_rate": 0.00024278846153846155,
+      "loss": 0.4106,
+      "step": 5853
+    },
+    {
+      "epoch": 53.71,
+      "learning_rate": 0.00024274038461538462,
+      "loss": 0.2666,
+      "step": 5854
+    },
+    {
+      "epoch": 53.72,
+      "learning_rate": 0.0002426923076923077,
+      "loss": 0.2772,
+      "step": 5855
+    },
+    {
+      "epoch": 53.72,
+      "learning_rate": 0.0002426442307692308,
+      "loss": 0.2667,
+      "step": 5856
+    },
+    {
+      "epoch": 53.73,
+      "learning_rate": 0.00024259615384615387,
+      "loss": 0.2896,
+      "step": 5857
+    },
+    {
+      "epoch": 53.74,
+      "learning_rate": 0.00024254807692307694,
+      "loss": 0.252,
+      "step": 5858
+    },
+    {
+      "epoch": 53.75,
+      "learning_rate": 0.00024249999999999999,
+      "loss": 0.3746,
+      "step": 5859
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 0.00024245192307692306,
+      "loss": 0.3656,
+      "step": 5860
+    },
+    {
+      "epoch": 53.77,
+      "learning_rate": 0.00024240384615384616,
+      "loss": 0.3162,
+      "step": 5861
+    },
+    {
+      "epoch": 53.78,
+      "learning_rate": 0.00024235576923076923,
+      "loss": 0.3588,
+      "step": 5862
+    },
+    {
+      "epoch": 53.79,
+      "learning_rate": 0.0002423076923076923,
+      "loss": 0.2029,
+      "step": 5863
+    },
+    {
+      "epoch": 53.8,
+      "learning_rate": 0.00024225961538461538,
+      "loss": 0.3628,
+      "step": 5864
+    },
+    {
+      "epoch": 53.81,
+      "learning_rate": 0.00024221153846153845,
+      "loss": 0.2897,
+      "step": 5865
+    },
+    {
+      "epoch": 53.82,
+      "learning_rate": 0.00024216346153846155,
+      "loss": 0.2996,
+      "step": 5866
+    },
+    {
+      "epoch": 53.83,
+      "learning_rate": 0.00024211538461538462,
+      "loss": 0.2641,
+      "step": 5867
+    },
+    {
+      "epoch": 53.83,
+      "learning_rate": 0.0002420673076923077,
+      "loss": 0.2523,
+      "step": 5868
+    },
+    {
+      "epoch": 53.84,
+      "learning_rate": 0.00024201923076923077,
+      "loss": 0.2227,
+      "step": 5869
+    },
+    {
+      "epoch": 53.85,
+      "learning_rate": 0.00024197115384615384,
+      "loss": 0.2792,
+      "step": 5870
+    },
+    {
+      "epoch": 53.86,
+      "learning_rate": 0.00024192307692307694,
+      "loss": 0.2589,
+      "step": 5871
+    },
+    {
+      "epoch": 53.87,
+      "learning_rate": 0.000241875,
+      "loss": 0.2322,
+      "step": 5872
+    },
+    {
+      "epoch": 53.88,
+      "learning_rate": 0.00024182692307692308,
+      "loss": 0.4007,
+      "step": 5873
+    },
+    {
+      "epoch": 53.89,
+      "learning_rate": 0.00024177884615384616,
+      "loss": 0.2541,
+      "step": 5874
+    },
+    {
+      "epoch": 53.9,
+      "learning_rate": 0.00024173076923076923,
+      "loss": 0.2783,
+      "step": 5875
+    },
+    {
+      "epoch": 53.91,
+      "learning_rate": 0.00024168269230769233,
+      "loss": 0.3417,
+      "step": 5876
+    },
+    {
+      "epoch": 53.92,
+      "learning_rate": 0.0002416346153846154,
+      "loss": 0.2645,
+      "step": 5877
+    },
+    {
+      "epoch": 53.93,
+      "learning_rate": 0.00024158653846153847,
+      "loss": 0.2757,
+      "step": 5878
+    },
+    {
+      "epoch": 53.94,
+      "learning_rate": 0.00024153846153846155,
+      "loss": 0.3695,
+      "step": 5879
+    },
+    {
+      "epoch": 53.94,
+      "learning_rate": 0.00024149038461538462,
+      "loss": 0.3452,
+      "step": 5880
+    },
+    {
+      "epoch": 53.95,
+      "learning_rate": 0.00024144230769230772,
+      "loss": 0.3231,
+      "step": 5881
+    },
+    {
+      "epoch": 53.96,
+      "learning_rate": 0.0002413942307692308,
+      "loss": 0.326,
+      "step": 5882
+    },
+    {
+      "epoch": 53.97,
+      "learning_rate": 0.00024134615384615386,
+      "loss": 0.2077,
+      "step": 5883
+    },
+    {
+      "epoch": 53.98,
+      "learning_rate": 0.00024129807692307694,
+      "loss": 0.3319,
+      "step": 5884
+    },
+    {
+      "epoch": 53.99,
+      "learning_rate": 0.00024125,
+      "loss": 0.251,
+      "step": 5885
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 0.00024120192307692308,
+      "loss": 0.2378,
+      "step": 5886
+    },
+    {
+      "epoch": 54.01,
+      "learning_rate": 0.00024115384615384615,
+      "loss": 0.4534,
+      "step": 5887
+    },
+    {
+      "epoch": 54.02,
+      "learning_rate": 0.00024110576923076923,
+      "loss": 0.2873,
+      "step": 5888
+    },
+    {
+      "epoch": 54.03,
+      "learning_rate": 0.0002410576923076923,
+      "loss": 0.3285,
+      "step": 5889
+    },
+    {
+      "epoch": 54.04,
+      "learning_rate": 0.00024100961538461537,
+      "loss": 0.3332,
+      "step": 5890
+    },
+    {
+      "epoch": 54.05,
+      "learning_rate": 0.00024096153846153845,
+      "loss": 0.3505,
+      "step": 5891
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 0.00024091346153846154,
+      "loss": 0.3434,
+      "step": 5892
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 0.00024086538461538462,
+      "loss": 0.2942,
+      "step": 5893
+    },
+    {
+      "epoch": 54.07,
+      "learning_rate": 0.0002408173076923077,
+      "loss": 0.4063,
+      "step": 5894
+    },
+    {
+      "epoch": 54.08,
+      "learning_rate": 0.00024076923076923076,
+      "loss": 0.3497,
+      "step": 5895
+    },
+    {
+      "epoch": 54.09,
+      "learning_rate": 0.00024072115384615384,
+      "loss": 0.3288,
+      "step": 5896
+    },
+    {
+      "epoch": 54.1,
+      "learning_rate": 0.00024067307692307694,
+      "loss": 0.3523,
+      "step": 5897
+    },
+    {
+      "epoch": 54.11,
+      "learning_rate": 0.000240625,
+      "loss": 0.3689,
+      "step": 5898
+    },
+    {
+      "epoch": 54.12,
+      "learning_rate": 0.00024057692307692308,
+      "loss": 0.2493,
+      "step": 5899
+    },
+    {
+      "epoch": 54.13,
+      "learning_rate": 0.00024052884615384615,
+      "loss": 0.2916,
+      "step": 5900
+    },
+    {
+      "epoch": 54.14,
+      "learning_rate": 0.00024048076923076923,
+      "loss": 0.3839,
+      "step": 5901
+    },
+    {
+      "epoch": 54.15,
+      "learning_rate": 0.00024043269230769233,
+      "loss": 0.3692,
+      "step": 5902
+    },
+    {
+      "epoch": 54.16,
+      "learning_rate": 0.0002403846153846154,
+      "loss": 0.4589,
+      "step": 5903
+    },
+    {
+      "epoch": 54.17,
+      "learning_rate": 0.00024033653846153847,
+      "loss": 0.3704,
+      "step": 5904
+    },
+    {
+      "epoch": 54.17,
+      "learning_rate": 0.00024028846153846154,
+      "loss": 0.3027,
+      "step": 5905
+    },
+    {
+      "epoch": 54.18,
+      "learning_rate": 0.00024024038461538462,
+      "loss": 0.2921,
+      "step": 5906
+    },
+    {
+      "epoch": 54.19,
+      "learning_rate": 0.00024019230769230772,
+      "loss": 0.3339,
+      "step": 5907
+    },
+    {
+      "epoch": 54.2,
+      "learning_rate": 0.0002401442307692308,
+      "loss": 0.3078,
+      "step": 5908
+    },
+    {
+      "epoch": 54.21,
+      "learning_rate": 0.00024009615384615386,
+      "loss": 0.3785,
+      "step": 5909
+    },
+    {
+      "epoch": 54.22,
+      "learning_rate": 0.00024004807692307693,
+      "loss": 0.2018,
+      "step": 5910
+    },
+    {
+      "epoch": 54.23,
+      "learning_rate": 0.00024,
+      "loss": 0.2895,
+      "step": 5911
+    },
+    {
+      "epoch": 54.24,
+      "learning_rate": 0.0002399519230769231,
+      "loss": 0.254,
+      "step": 5912
+    },
+    {
+      "epoch": 54.25,
+      "learning_rate": 0.00023990384615384618,
+      "loss": 0.378,
+      "step": 5913
+    },
+    {
+      "epoch": 54.26,
+      "learning_rate": 0.00023985576923076925,
+      "loss": 0.4423,
+      "step": 5914
+    },
+    {
+      "epoch": 54.27,
+      "learning_rate": 0.0002398076923076923,
+      "loss": 0.375,
+      "step": 5915
+    },
+    {
+      "epoch": 54.28,
+      "learning_rate": 0.00023975961538461537,
+      "loss": 0.2715,
+      "step": 5916
+    },
+    {
+      "epoch": 54.28,
+      "learning_rate": 0.00023971153846153844,
+      "loss": 0.2313,
+      "step": 5917
+    },
+    {
+      "epoch": 54.29,
+      "learning_rate": 0.00023966346153846154,
+      "loss": 0.3859,
+      "step": 5918
+    },
+    {
+      "epoch": 54.3,
+      "learning_rate": 0.00023961538461538461,
+      "loss": 0.3044,
+      "step": 5919
+    },
+    {
+      "epoch": 54.31,
+      "learning_rate": 0.0002395673076923077,
+      "loss": 0.2959,
+      "step": 5920
+    },
+    {
+      "epoch": 54.32,
+      "learning_rate": 0.00023951923076923076,
+      "loss": 0.2428,
+      "step": 5921
+    },
+    {
+      "epoch": 54.33,
+      "learning_rate": 0.00023947115384615383,
+      "loss": 0.3032,
+      "step": 5922
+    },
+    {
+      "epoch": 54.34,
+      "learning_rate": 0.00023942307692307693,
+      "loss": 0.3602,
+      "step": 5923
+    },
+    {
+      "epoch": 54.35,
+      "learning_rate": 0.000239375,
+      "loss": 0.2677,
+      "step": 5924
+    },
+    {
+      "epoch": 54.36,
+      "learning_rate": 0.00023932692307692308,
+      "loss": 0.3522,
+      "step": 5925
+    },
+    {
+      "epoch": 54.37,
+      "learning_rate": 0.00023927884615384615,
+      "loss": 0.3096,
+      "step": 5926
+    },
+    {
+      "epoch": 54.38,
+      "learning_rate": 0.00023923076923076922,
+      "loss": 0.2519,
+      "step": 5927
+    },
+    {
+      "epoch": 54.39,
+      "learning_rate": 0.00023918269230769232,
+      "loss": 0.4111,
+      "step": 5928
+    },
+    {
+      "epoch": 54.39,
+      "learning_rate": 0.0002391346153846154,
+      "loss": 0.3616,
+      "step": 5929
+    },
+    {
+      "epoch": 54.4,
+      "learning_rate": 0.00023908653846153847,
+      "loss": 0.2201,
+      "step": 5930
+    },
+    {
+      "epoch": 54.41,
+      "learning_rate": 0.00023903846153846154,
+      "loss": 0.3212,
+      "step": 5931
+    },
+    {
+      "epoch": 54.42,
+      "learning_rate": 0.0002389903846153846,
+      "loss": 0.3158,
+      "step": 5932
+    },
+    {
+      "epoch": 54.43,
+      "learning_rate": 0.0002389423076923077,
+      "loss": 0.356,
+      "step": 5933
+    },
+    {
+      "epoch": 54.44,
+      "learning_rate": 0.00023889423076923078,
+      "loss": 0.3119,
+      "step": 5934
+    },
+    {
+      "epoch": 54.45,
+      "learning_rate": 0.00023884615384615386,
+      "loss": 0.3477,
+      "step": 5935
+    },
+    {
+      "epoch": 54.46,
+      "learning_rate": 0.00023879807692307693,
+      "loss": 0.4011,
+      "step": 5936
+    },
+    {
+      "epoch": 54.47,
+      "learning_rate": 0.00023875,
+      "loss": 0.2853,
+      "step": 5937
+    },
+    {
+      "epoch": 54.48,
+      "learning_rate": 0.0002387019230769231,
+      "loss": 0.3025,
+      "step": 5938
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 0.00023865384615384618,
+      "loss": 0.1441,
+      "step": 5939
+    },
+    {
+      "epoch": 54.5,
+      "learning_rate": 0.00023860576923076925,
+      "loss": 0.174,
+      "step": 5940
+    },
+    {
+      "epoch": 54.5,
+      "learning_rate": 0.00023855769230769232,
+      "loss": 0.4382,
+      "step": 5941
+    },
+    {
+      "epoch": 54.51,
+      "learning_rate": 0.0002385096153846154,
+      "loss": 0.3755,
+      "step": 5942
+    },
+    {
+      "epoch": 54.52,
+      "learning_rate": 0.0002384615384615385,
+      "loss": 0.3487,
+      "step": 5943
+    },
+    {
+      "epoch": 54.53,
+      "learning_rate": 0.00023841346153846154,
+      "loss": 0.3583,
+      "step": 5944
+    },
+    {
+      "epoch": 54.54,
+      "learning_rate": 0.0002383653846153846,
+      "loss": 0.2977,
+      "step": 5945
+    },
+    {
+      "epoch": 54.55,
+      "learning_rate": 0.00023831730769230768,
+      "loss": 0.3134,
+      "step": 5946
+    },
+    {
+      "epoch": 54.56,
+      "learning_rate": 0.00023826923076923076,
+      "loss": 0.3647,
+      "step": 5947
+    },
+    {
+      "epoch": 54.57,
+      "learning_rate": 0.00023822115384615383,
+      "loss": 0.3084,
+      "step": 5948
+    },
+    {
+      "epoch": 54.58,
+      "learning_rate": 0.00023817307692307693,
+      "loss": 0.2659,
+      "step": 5949
+    },
+    {
+      "epoch": 54.59,
+      "learning_rate": 0.000238125,
+      "loss": 0.3258,
+      "step": 5950
+    },
+    {
+      "epoch": 54.6,
+      "learning_rate": 0.00023807692307692307,
+      "loss": 0.2811,
+      "step": 5951
+    },
+    {
+      "epoch": 54.61,
+      "learning_rate": 0.00023802884615384615,
+      "loss": 0.3579,
+      "step": 5952
+    },
+    {
+      "epoch": 54.61,
+      "learning_rate": 0.00023798076923076922,
+      "loss": 0.2508,
+      "step": 5953
+    },
+    {
+      "epoch": 54.62,
+      "learning_rate": 0.00023793269230769232,
+      "loss": 0.3566,
+      "step": 5954
+    },
+    {
+      "epoch": 54.63,
+      "learning_rate": 0.0002378846153846154,
+      "loss": 0.3119,
+      "step": 5955
+    },
+    {
+      "epoch": 54.64,
+      "learning_rate": 0.00023783653846153846,
+      "loss": 0.3224,
+      "step": 5956
+    },
+    {
+      "epoch": 54.65,
+      "learning_rate": 0.00023778846153846154,
+      "loss": 0.2439,
+      "step": 5957
+    },
+    {
+      "epoch": 54.66,
+      "learning_rate": 0.0002377403846153846,
+      "loss": 0.2913,
+      "step": 5958
+    },
+    {
+      "epoch": 54.67,
+      "learning_rate": 0.0002376923076923077,
+      "loss": 0.32,
+      "step": 5959
+    },
+    {
+      "epoch": 54.68,
+      "learning_rate": 0.00023764423076923078,
+      "loss": 0.3043,
+      "step": 5960
+    },
+    {
+      "epoch": 54.69,
+      "learning_rate": 0.00023759615384615385,
+      "loss": 0.3103,
+      "step": 5961
+    },
+    {
+      "epoch": 54.7,
+      "learning_rate": 0.00023754807692307693,
+      "loss": 0.226,
+      "step": 5962
+    },
+    {
+      "epoch": 54.71,
+      "learning_rate": 0.0002375,
+      "loss": 0.2066,
+      "step": 5963
+    },
+    {
+      "epoch": 54.72,
+      "learning_rate": 0.0002374519230769231,
+      "loss": 0.2626,
+      "step": 5964
+    },
+    {
+      "epoch": 54.72,
+      "learning_rate": 0.00023740384615384617,
+      "loss": 0.2616,
+      "step": 5965
+    },
+    {
+      "epoch": 54.73,
+      "learning_rate": 0.00023735576923076924,
+      "loss": 0.3323,
+      "step": 5966
+    },
+    {
+      "epoch": 54.74,
+      "learning_rate": 0.00023730769230769232,
+      "loss": 0.2668,
+      "step": 5967
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 0.0002372596153846154,
+      "loss": 0.3493,
+      "step": 5968
+    },
+    {
+      "epoch": 54.76,
+      "learning_rate": 0.0002372115384615385,
+      "loss": 0.3018,
+      "step": 5969
+    },
+    {
+      "epoch": 54.77,
+      "learning_rate": 0.00023716346153846156,
+      "loss": 0.3154,
+      "step": 5970
+    },
+    {
+      "epoch": 54.78,
+      "learning_rate": 0.00023711538461538463,
+      "loss": 0.2642,
+      "step": 5971
+    },
+    {
+      "epoch": 54.79,
+      "learning_rate": 0.00023706730769230768,
+      "loss": 0.3152,
+      "step": 5972
+    },
+    {
+      "epoch": 54.8,
+      "learning_rate": 0.00023701923076923075,
+      "loss": 0.3139,
+      "step": 5973
+    },
+    {
+      "epoch": 54.81,
+      "learning_rate": 0.00023697115384615385,
+      "loss": 0.3655,
+      "step": 5974
+    },
+    {
+      "epoch": 54.82,
+      "learning_rate": 0.00023692307692307693,
+      "loss": 0.3652,
+      "step": 5975
+    },
+    {
+      "epoch": 54.83,
+      "learning_rate": 0.000236875,
+      "loss": 0.3014,
+      "step": 5976
+    },
+    {
+      "epoch": 54.83,
+      "learning_rate": 0.00023682692307692307,
+      "loss": 0.3363,
+      "step": 5977
+    },
+    {
+      "epoch": 54.84,
+      "learning_rate": 0.00023677884615384614,
+      "loss": 0.346,
+      "step": 5978
+    },
+    {
+      "epoch": 54.85,
+      "learning_rate": 0.00023673076923076924,
+      "loss": 0.3004,
+      "step": 5979
+    },
+    {
+      "epoch": 54.86,
+      "learning_rate": 0.00023668269230769232,
+      "loss": 0.3222,
+      "step": 5980
+    },
+    {
+      "epoch": 54.87,
+      "learning_rate": 0.0002366346153846154,
+      "loss": 0.3422,
+      "step": 5981
+    },
+    {
+      "epoch": 54.88,
+      "learning_rate": 0.00023658653846153846,
+      "loss": 0.3694,
+      "step": 5982
+    },
+    {
+      "epoch": 54.89,
+      "learning_rate": 0.00023653846153846153,
+      "loss": 0.3906,
+      "step": 5983
+    },
+    {
+      "epoch": 54.9,
+      "learning_rate": 0.0002364903846153846,
+      "loss": 0.2453,
+      "step": 5984
+    },
+    {
+      "epoch": 54.91,
+      "learning_rate": 0.0002364423076923077,
+      "loss": 0.3363,
+      "step": 5985
+    },
+    {
+      "epoch": 54.92,
+      "learning_rate": 0.00023639423076923078,
+      "loss": 0.3809,
+      "step": 5986
+    },
+    {
+      "epoch": 54.93,
+      "learning_rate": 0.00023634615384615385,
+      "loss": 0.3914,
+      "step": 5987
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 0.00023629807692307692,
+      "loss": 0.3467,
+      "step": 5988
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 0.00023625,
+      "loss": 0.2271,
+      "step": 5989
+    },
+    {
+      "epoch": 54.95,
+      "learning_rate": 0.0002362019230769231,
+      "loss": 0.3333,
+      "step": 5990
+    },
+    {
+      "epoch": 54.96,
+      "learning_rate": 0.00023615384615384617,
+      "loss": 0.2946,
+      "step": 5991
+    },
+    {
+      "epoch": 54.97,
+      "learning_rate": 0.00023610576923076924,
+      "loss": 0.2414,
+      "step": 5992
+    },
+    {
+      "epoch": 54.98,
+      "learning_rate": 0.00023605769230769231,
+      "loss": 0.3007,
+      "step": 5993
+    },
+    {
+      "epoch": 54.99,
+      "learning_rate": 0.0002360096153846154,
+      "loss": 0.2457,
+      "step": 5994
+    },
+    {
+      "epoch": 55.0,
+      "learning_rate": 0.00023596153846153849,
+      "loss": 0.3252,
+      "step": 5995
+    },
+    {
+      "epoch": 55.01,
+      "learning_rate": 0.00023591346153846156,
+      "loss": 0.4127,
+      "step": 5996
+    },
+    {
+      "epoch": 55.02,
+      "learning_rate": 0.00023586538461538463,
+      "loss": 0.3484,
+      "step": 5997
+    },
+    {
+      "epoch": 55.03,
+      "learning_rate": 0.0002358173076923077,
+      "loss": 0.3225,
+      "step": 5998
+    },
+    {
+      "epoch": 55.04,
+      "learning_rate": 0.00023576923076923078,
+      "loss": 0.2797,
+      "step": 5999
+    },
+    {
+      "epoch": 55.05,
+      "learning_rate": 0.00023572115384615388,
+      "loss": 0.3914,
+      "step": 6000
+    },
+    {
+      "epoch": 55.05,
+      "eval_cer": 0.16313587909863717,
+      "eval_loss": 0.71206134557724,
+      "eval_runtime": 88.1935,
+      "eval_samples_per_second": 18.675,
+      "eval_steps_per_second": 2.336,
+      "eval_wer": 0.5552037585537739,
+      "step": 6000
+    },
+    {
+      "epoch": 55.06,
+      "learning_rate": 0.00023567307692307692,
+      "loss": 0.3344,
+      "step": 6001
+    },
+    {
+      "epoch": 55.06,
+      "learning_rate": 0.000235625,
+      "loss": 0.3724,
+      "step": 6002
+    },
+    {
+      "epoch": 55.07,
+      "learning_rate": 0.00023557692307692307,
+      "loss": 0.323,
+      "step": 6003
+    },
+    {
+      "epoch": 55.08,
+      "learning_rate": 0.00023552884615384614,
+      "loss": 0.3024,
+      "step": 6004
+    },
+    {
+      "epoch": 55.09,
+      "learning_rate": 0.00023548076923076924,
+      "loss": 0.328,
+      "step": 6005
+    },
+    {
+      "epoch": 55.1,
+      "learning_rate": 0.0002354326923076923,
+      "loss": 0.3875,
+      "step": 6006
+    },
+    {
+      "epoch": 55.11,
+      "learning_rate": 0.00023538461538461538,
+      "loss": 0.3639,
+      "step": 6007
+    },
+    {
+      "epoch": 55.12,
+      "learning_rate": 0.00023533653846153846,
+      "loss": 0.2635,
+      "step": 6008
+    },
+    {
+      "epoch": 55.13,
+      "learning_rate": 0.00023528846153846153,
+      "loss": 0.3014,
+      "step": 6009
+    },
+    {
+      "epoch": 55.14,
+      "learning_rate": 0.00023524038461538463,
+      "loss": 0.3213,
+      "step": 6010
+    },
+    {
+      "epoch": 55.15,
+      "learning_rate": 0.0002351923076923077,
+      "loss": 0.3369,
+      "step": 6011
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 0.00023514423076923078,
+      "loss": 0.3638,
+      "step": 6012
+    },
+    {
+      "epoch": 55.17,
+      "learning_rate": 0.00023509615384615385,
+      "loss": 0.3247,
+      "step": 6013
+    },
+    {
+      "epoch": 55.17,
+      "learning_rate": 0.00023504807692307692,
+      "loss": 0.3455,
+      "step": 6014
+    },
+    {
+      "epoch": 55.18,
+      "learning_rate": 0.000235,
+      "loss": 0.3028,
+      "step": 6015
+    },
+    {
+      "epoch": 55.19,
+      "learning_rate": 0.0002349519230769231,
+      "loss": 0.2993,
+      "step": 6016
+    },
+    {
+      "epoch": 55.2,
+      "learning_rate": 0.00023490384615384617,
+      "loss": 0.2813,
+      "step": 6017
+    },
+    {
+      "epoch": 55.21,
+      "learning_rate": 0.00023485576923076924,
+      "loss": 0.2247,
+      "step": 6018
+    },
+    {
+      "epoch": 55.22,
+      "learning_rate": 0.0002348076923076923,
+      "loss": 0.2157,
+      "step": 6019
+    },
+    {
+      "epoch": 55.23,
+      "learning_rate": 0.00023475961538461538,
+      "loss": 0.2671,
+      "step": 6020
+    },
+    {
+      "epoch": 55.24,
+      "learning_rate": 0.00023471153846153848,
+      "loss": 0.1923,
+      "step": 6021
+    },
+    {
+      "epoch": 55.25,
+      "learning_rate": 0.00023466346153846156,
+      "loss": 0.2496,
+      "step": 6022
+    },
+    {
+      "epoch": 55.26,
+      "learning_rate": 0.00023461538461538463,
+      "loss": 0.4264,
+      "step": 6023
+    },
+    {
+      "epoch": 55.27,
+      "learning_rate": 0.0002345673076923077,
+      "loss": 0.3282,
+      "step": 6024
+    },
+    {
+      "epoch": 55.28,
+      "learning_rate": 0.00023451923076923077,
+      "loss": 0.3483,
+      "step": 6025
+    },
+    {
+      "epoch": 55.28,
+      "learning_rate": 0.00023447115384615387,
+      "loss": 0.3016,
+      "step": 6026
+    },
+    {
+      "epoch": 55.29,
+      "learning_rate": 0.00023442307692307695,
+      "loss": 0.2962,
+      "step": 6027
+    },
+    {
+      "epoch": 55.3,
+      "learning_rate": 0.000234375,
+      "loss": 0.2199,
+      "step": 6028
+    },
+    {
+      "epoch": 55.31,
+      "learning_rate": 0.00023432692307692306,
+      "loss": 0.4017,
+      "step": 6029
+    },
+    {
+      "epoch": 55.32,
+      "learning_rate": 0.00023427884615384614,
+      "loss": 0.2929,
+      "step": 6030
+    },
+    {
+      "epoch": 55.33,
+      "learning_rate": 0.00023423076923076924,
+      "loss": 0.3405,
+      "step": 6031
+    },
+    {
+      "epoch": 55.34,
+      "learning_rate": 0.0002341826923076923,
+      "loss": 0.2728,
+      "step": 6032
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 0.00023413461538461538,
+      "loss": 0.3537,
+      "step": 6033
+    },
+    {
+      "epoch": 55.36,
+      "learning_rate": 0.00023408653846153845,
+      "loss": 0.2695,
+      "step": 6034
+    },
+    {
+      "epoch": 55.37,
+      "learning_rate": 0.00023403846153846153,
+      "loss": 0.2931,
+      "step": 6035
+    },
+    {
+      "epoch": 55.38,
+      "learning_rate": 0.00023399038461538463,
+      "loss": 0.3465,
+      "step": 6036
+    },
+    {
+      "epoch": 55.39,
+      "learning_rate": 0.0002339423076923077,
+      "loss": 0.3294,
+      "step": 6037
+    },
+    {
+      "epoch": 55.39,
+      "learning_rate": 0.00023389423076923077,
+      "loss": 0.2603,
+      "step": 6038
+    },
+    {
+      "epoch": 55.4,
+      "learning_rate": 0.00023384615384615384,
+      "loss": 0.2908,
+      "step": 6039
+    },
+    {
+      "epoch": 55.41,
+      "learning_rate": 0.00023379807692307692,
+      "loss": 0.2259,
+      "step": 6040
+    },
+    {
+      "epoch": 55.42,
+      "learning_rate": 0.00023375000000000002,
+      "loss": 0.2708,
+      "step": 6041
+    },
+    {
+      "epoch": 55.43,
+      "learning_rate": 0.0002337019230769231,
+      "loss": 0.3378,
+      "step": 6042
+    },
+    {
+      "epoch": 55.44,
+      "learning_rate": 0.00023365384615384616,
+      "loss": 0.3043,
+      "step": 6043
+    },
+    {
+      "epoch": 55.45,
+      "learning_rate": 0.00023360576923076923,
+      "loss": 0.2732,
+      "step": 6044
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 0.0002335576923076923,
+      "loss": 0.3722,
+      "step": 6045
+    },
+    {
+      "epoch": 55.47,
+      "learning_rate": 0.00023350961538461538,
+      "loss": 0.2269,
+      "step": 6046
+    },
+    {
+      "epoch": 55.48,
+      "learning_rate": 0.00023346153846153848,
+      "loss": 0.1652,
+      "step": 6047
+    },
+    {
+      "epoch": 55.49,
+      "learning_rate": 0.00023341346153846155,
+      "loss": 0.3067,
+      "step": 6048
+    },
+    {
+      "epoch": 55.5,
+      "learning_rate": 0.00023336538461538462,
+      "loss": 0.1473,
+      "step": 6049
+    },
+    {
+      "epoch": 55.5,
+      "learning_rate": 0.0002333173076923077,
+      "loss": 0.4359,
+      "step": 6050
+    },
+    {
+      "epoch": 55.51,
+      "learning_rate": 0.00023326923076923077,
+      "loss": 0.3426,
+      "step": 6051
+    },
+    {
+      "epoch": 55.52,
+      "learning_rate": 0.00023322115384615387,
+      "loss": 0.3071,
+      "step": 6052
+    },
+    {
+      "epoch": 55.53,
+      "learning_rate": 0.00023317307692307694,
+      "loss": 0.3825,
+      "step": 6053
+    },
+    {
+      "epoch": 55.54,
+      "learning_rate": 0.00023312500000000002,
+      "loss": 0.2591,
+      "step": 6054
+    },
+    {
+      "epoch": 55.55,
+      "learning_rate": 0.0002330769230769231,
+      "loss": 0.3375,
+      "step": 6055
+    },
+    {
+      "epoch": 55.56,
+      "learning_rate": 0.00023302884615384613,
+      "loss": 0.282,
+      "step": 6056
+    },
+    {
+      "epoch": 55.57,
+      "learning_rate": 0.00023298076923076923,
+      "loss": 0.2065,
+      "step": 6057
+    },
+    {
+      "epoch": 55.58,
+      "learning_rate": 0.0002329326923076923,
+      "loss": 0.3012,
+      "step": 6058
+    },
+    {
+      "epoch": 55.59,
+      "learning_rate": 0.00023288461538461538,
+      "loss": 0.3274,
+      "step": 6059
+    },
+    {
+      "epoch": 55.6,
+      "learning_rate": 0.00023283653846153845,
+      "loss": 0.2837,
+      "step": 6060
+    },
+    {
+      "epoch": 55.61,
+      "learning_rate": 0.00023278846153846152,
+      "loss": 0.2933,
+      "step": 6061
+    },
+    {
+      "epoch": 55.61,
+      "learning_rate": 0.00023274038461538462,
+      "loss": 0.2526,
+      "step": 6062
+    },
+    {
+      "epoch": 55.62,
+      "learning_rate": 0.0002326923076923077,
+      "loss": 0.3099,
+      "step": 6063
+    },
+    {
+      "epoch": 55.63,
+      "learning_rate": 0.00023264423076923077,
+      "loss": 0.3282,
+      "step": 6064
+    },
+    {
+      "epoch": 55.64,
+      "learning_rate": 0.00023259615384615384,
+      "loss": 0.2263,
+      "step": 6065
+    },
+    {
+      "epoch": 55.65,
+      "learning_rate": 0.00023254807692307691,
+      "loss": 0.2965,
+      "step": 6066
+    },
+    {
+      "epoch": 55.66,
+      "learning_rate": 0.0002325,
+      "loss": 0.2475,
+      "step": 6067
+    },
+    {
+      "epoch": 55.67,
+      "learning_rate": 0.00023245192307692309,
+      "loss": 0.2669,
+      "step": 6068
+    },
+    {
+      "epoch": 55.68,
+      "learning_rate": 0.00023240384615384616,
+      "loss": 0.2716,
+      "step": 6069
+    },
+    {
+      "epoch": 55.69,
+      "learning_rate": 0.00023235576923076923,
+      "loss": 0.2523,
+      "step": 6070
+    },
+    {
+      "epoch": 55.7,
+      "learning_rate": 0.0002323076923076923,
+      "loss": 0.3011,
+      "step": 6071
+    },
+    {
+      "epoch": 55.71,
+      "learning_rate": 0.0002322596153846154,
+      "loss": 0.2027,
+      "step": 6072
+    },
+    {
+      "epoch": 55.72,
+      "learning_rate": 0.00023221153846153848,
+      "loss": 0.2116,
+      "step": 6073
+    },
+    {
+      "epoch": 55.72,
+      "learning_rate": 0.00023216346153846155,
+      "loss": 0.2796,
+      "step": 6074
+    },
+    {
+      "epoch": 55.73,
+      "learning_rate": 0.00023211538461538462,
+      "loss": 0.2941,
+      "step": 6075
+    },
+    {
+      "epoch": 55.74,
+      "learning_rate": 0.0002320673076923077,
+      "loss": 0.246,
+      "step": 6076
+    },
+    {
+      "epoch": 55.75,
+      "learning_rate": 0.00023201923076923077,
+      "loss": 0.3469,
+      "step": 6077
+    },
+    {
+      "epoch": 55.76,
+      "learning_rate": 0.00023197115384615387,
+      "loss": 0.3647,
+      "step": 6078
+    },
+    {
+      "epoch": 55.77,
+      "learning_rate": 0.00023192307692307694,
+      "loss": 0.3465,
+      "step": 6079
+    },
+    {
+      "epoch": 55.78,
+      "learning_rate": 0.000231875,
+      "loss": 0.394,
+      "step": 6080
+    },
+    {
+      "epoch": 55.79,
+      "learning_rate": 0.00023182692307692308,
+      "loss": 0.3788,
+      "step": 6081
+    },
+    {
+      "epoch": 55.8,
+      "learning_rate": 0.00023177884615384616,
+      "loss": 0.2942,
+      "step": 6082
+    },
+    {
+      "epoch": 55.81,
+      "learning_rate": 0.00023173076923076926,
+      "loss": 0.3606,
+      "step": 6083
+    },
+    {
+      "epoch": 55.82,
+      "learning_rate": 0.00023168269230769233,
+      "loss": 0.2413,
+      "step": 6084
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 0.00023163461538461538,
+      "loss": 0.301,
+      "step": 6085
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 0.00023158653846153845,
+      "loss": 0.3587,
+      "step": 6086
+    },
+    {
+      "epoch": 55.84,
+      "learning_rate": 0.00023153846153846152,
+      "loss": 0.3122,
+      "step": 6087
+    },
+    {
+      "epoch": 55.85,
+      "learning_rate": 0.00023149038461538462,
+      "loss": 0.2758,
+      "step": 6088
+    },
+    {
+      "epoch": 55.86,
+      "learning_rate": 0.0002314423076923077,
+      "loss": 0.263,
+      "step": 6089
+    },
+    {
+      "epoch": 55.87,
+      "learning_rate": 0.00023139423076923077,
+      "loss": 0.3281,
+      "step": 6090
+    },
+    {
+      "epoch": 55.88,
+      "learning_rate": 0.00023134615384615384,
+      "loss": 0.2729,
+      "step": 6091
+    },
+    {
+      "epoch": 55.89,
+      "learning_rate": 0.0002312980769230769,
+      "loss": 0.2503,
+      "step": 6092
+    },
+    {
+      "epoch": 55.9,
+      "learning_rate": 0.00023125,
+      "loss": 0.3594,
+      "step": 6093
+    },
+    {
+      "epoch": 55.91,
+      "learning_rate": 0.00023120192307692308,
+      "loss": 0.2501,
+      "step": 6094
+    },
+    {
+      "epoch": 55.92,
+      "learning_rate": 0.00023115384615384616,
+      "loss": 0.2833,
+      "step": 6095
+    },
+    {
+      "epoch": 55.93,
+      "learning_rate": 0.00023110576923076923,
+      "loss": 0.5096,
+      "step": 6096
+    },
+    {
+      "epoch": 55.94,
+      "learning_rate": 0.0002310576923076923,
+      "loss": 0.3131,
+      "step": 6097
+    },
+    {
+      "epoch": 55.94,
+      "learning_rate": 0.0002310096153846154,
+      "loss": 0.3604,
+      "step": 6098
+    },
+    {
+      "epoch": 55.95,
+      "learning_rate": 0.00023096153846153847,
+      "loss": 0.2559,
+      "step": 6099
+    },
+    {
+      "epoch": 55.96,
+      "learning_rate": 0.00023091346153846155,
+      "loss": 0.2522,
+      "step": 6100
+    },
+    {
+      "epoch": 55.97,
+      "learning_rate": 0.00023086538461538462,
+      "loss": 0.3146,
+      "step": 6101
+    },
+    {
+      "epoch": 55.98,
+      "learning_rate": 0.0002308173076923077,
+      "loss": 0.2944,
+      "step": 6102
+    },
+    {
+      "epoch": 55.99,
+      "learning_rate": 0.0002307692307692308,
+      "loss": 0.1308,
+      "step": 6103
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 0.00023072115384615386,
+      "loss": 0.3924,
+      "step": 6104
+    },
+    {
+      "epoch": 56.01,
+      "learning_rate": 0.00023067307692307694,
+      "loss": 0.3756,
+      "step": 6105
+    },
+    {
+      "epoch": 56.02,
+      "learning_rate": 0.000230625,
+      "loss": 0.3066,
+      "step": 6106
+    },
+    {
+      "epoch": 56.03,
+      "learning_rate": 0.00023057692307692308,
+      "loss": 0.3139,
+      "step": 6107
+    },
+    {
+      "epoch": 56.04,
+      "learning_rate": 0.00023052884615384618,
+      "loss": 0.3295,
+      "step": 6108
+    },
+    {
+      "epoch": 56.05,
+      "learning_rate": 0.00023048076923076925,
+      "loss": 0.3227,
+      "step": 6109
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 0.00023043269230769233,
+      "loss": 0.2993,
+      "step": 6110
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 0.0002303846153846154,
+      "loss": 0.2744,
+      "step": 6111
+    },
+    {
+      "epoch": 56.07,
+      "learning_rate": 0.00023033653846153847,
+      "loss": 0.3639,
+      "step": 6112
+    },
+    {
+      "epoch": 56.08,
+      "learning_rate": 0.00023028846153846152,
+      "loss": 0.3001,
+      "step": 6113
+    },
+    {
+      "epoch": 56.09,
+      "learning_rate": 0.00023024038461538462,
+      "loss": 0.3516,
+      "step": 6114
+    },
+    {
+      "epoch": 56.1,
+      "learning_rate": 0.0002301923076923077,
+      "loss": 0.2762,
+      "step": 6115
+    },
+    {
+      "epoch": 56.11,
+      "learning_rate": 0.00023014423076923076,
+      "loss": 0.3348,
+      "step": 6116
+    },
+    {
+      "epoch": 56.12,
+      "learning_rate": 0.00023009615384615383,
+      "loss": 0.2394,
+      "step": 6117
+    },
+    {
+      "epoch": 56.13,
+      "learning_rate": 0.0002300480769230769,
+      "loss": 0.2462,
+      "step": 6118
+    },
+    {
+      "epoch": 56.14,
+      "learning_rate": 0.00023,
+      "loss": 0.2741,
+      "step": 6119
+    },
+    {
+      "epoch": 56.15,
+      "learning_rate": 0.00022995192307692308,
+      "loss": 0.3267,
+      "step": 6120
+    },
+    {
+      "epoch": 56.16,
+      "learning_rate": 0.00022990384615384615,
+      "loss": 0.3479,
+      "step": 6121
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 0.00022985576923076922,
+      "loss": 0.358,
+      "step": 6122
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 0.0002298076923076923,
+      "loss": 0.2691,
+      "step": 6123
+    },
+    {
+      "epoch": 56.18,
+      "learning_rate": 0.0002297596153846154,
+      "loss": 0.292,
+      "step": 6124
+    },
+    {
+      "epoch": 56.19,
+      "learning_rate": 0.00022971153846153847,
+      "loss": 0.2784,
+      "step": 6125
+    },
+    {
+      "epoch": 56.2,
+      "learning_rate": 0.00022966346153846154,
+      "loss": 0.3655,
+      "step": 6126
+    },
+    {
+      "epoch": 56.21,
+      "learning_rate": 0.00022961538461538462,
+      "loss": 0.3922,
+      "step": 6127
+    },
+    {
+      "epoch": 56.22,
+      "learning_rate": 0.0002295673076923077,
+      "loss": 0.2547,
+      "step": 6128
+    },
+    {
+      "epoch": 56.23,
+      "learning_rate": 0.0002295192307692308,
+      "loss": 0.2862,
+      "step": 6129
+    },
+    {
+      "epoch": 56.24,
+      "learning_rate": 0.00022947115384615386,
+      "loss": 0.3635,
+      "step": 6130
+    },
+    {
+      "epoch": 56.25,
+      "learning_rate": 0.00022942307692307693,
+      "loss": 0.1777,
+      "step": 6131
+    },
+    {
+      "epoch": 56.26,
+      "learning_rate": 0.000229375,
+      "loss": 0.3444,
+      "step": 6132
+    },
+    {
+      "epoch": 56.27,
+      "learning_rate": 0.00022932692307692308,
+      "loss": 0.3227,
+      "step": 6133
+    },
+    {
+      "epoch": 56.28,
+      "learning_rate": 0.00022927884615384618,
+      "loss": 0.3683,
+      "step": 6134
+    },
+    {
+      "epoch": 56.28,
+      "learning_rate": 0.00022923076923076925,
+      "loss": 0.3486,
+      "step": 6135
+    },
+    {
+      "epoch": 56.29,
+      "learning_rate": 0.00022918269230769232,
+      "loss": 0.3458,
+      "step": 6136
+    },
+    {
+      "epoch": 56.3,
+      "learning_rate": 0.0002291346153846154,
+      "loss": 0.319,
+      "step": 6137
+    },
+    {
+      "epoch": 56.31,
+      "learning_rate": 0.00022908653846153847,
+      "loss": 0.3792,
+      "step": 6138
+    },
+    {
+      "epoch": 56.32,
+      "learning_rate": 0.00022903846153846157,
+      "loss": 0.3016,
+      "step": 6139
+    },
+    {
+      "epoch": 56.33,
+      "learning_rate": 0.00022899038461538464,
+      "loss": 0.2234,
+      "step": 6140
+    },
+    {
+      "epoch": 56.34,
+      "learning_rate": 0.0002289423076923077,
+      "loss": 0.379,
+      "step": 6141
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 0.00022889423076923076,
+      "loss": 0.2598,
+      "step": 6142
+    },
+    {
+      "epoch": 56.36,
+      "learning_rate": 0.00022884615384615383,
+      "loss": 0.3267,
+      "step": 6143
+    },
+    {
+      "epoch": 56.37,
+      "learning_rate": 0.0002287980769230769,
+      "loss": 0.3248,
+      "step": 6144
+    },
+    {
+      "epoch": 56.38,
+      "learning_rate": 0.00022875,
+      "loss": 0.3205,
+      "step": 6145
+    },
+    {
+      "epoch": 56.39,
+      "learning_rate": 0.00022870192307692308,
+      "loss": 0.3045,
+      "step": 6146
+    },
+    {
+      "epoch": 56.39,
+      "learning_rate": 0.00022865384615384615,
+      "loss": 0.3121,
+      "step": 6147
+    },
+    {
+      "epoch": 56.4,
+      "learning_rate": 0.00022860576923076922,
+      "loss": 0.357,
+      "step": 6148
+    },
+    {
+      "epoch": 56.41,
+      "learning_rate": 0.0002285576923076923,
+      "loss": 0.2883,
+      "step": 6149
+    },
+    {
+      "epoch": 56.42,
+      "learning_rate": 0.0002285096153846154,
+      "loss": 0.3628,
+      "step": 6150
+    },
+    {
+      "epoch": 56.43,
+      "learning_rate": 0.00022846153846153847,
+      "loss": 0.2997,
+      "step": 6151
+    },
+    {
+      "epoch": 56.44,
+      "learning_rate": 0.00022841346153846154,
+      "loss": 0.3538,
+      "step": 6152
+    },
+    {
+      "epoch": 56.45,
+      "learning_rate": 0.0002283653846153846,
+      "loss": 0.3749,
+      "step": 6153
+    },
+    {
+      "epoch": 56.46,
+      "learning_rate": 0.00022831730769230768,
+      "loss": 0.3349,
+      "step": 6154
+    },
+    {
+      "epoch": 56.47,
+      "learning_rate": 0.00022826923076923078,
+      "loss": 0.2448,
+      "step": 6155
+    },
+    {
+      "epoch": 56.48,
+      "learning_rate": 0.00022822115384615386,
+      "loss": 0.3002,
+      "step": 6156
+    },
+    {
+      "epoch": 56.49,
+      "learning_rate": 0.00022817307692307693,
+      "loss": 0.254,
+      "step": 6157
+    },
+    {
+      "epoch": 56.5,
+      "learning_rate": 0.000228125,
+      "loss": 0.3087,
+      "step": 6158
+    },
+    {
+      "epoch": 56.5,
+      "learning_rate": 0.00022807692307692307,
+      "loss": 0.3919,
+      "step": 6159
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 0.00022802884615384617,
+      "loss": 0.3678,
+      "step": 6160
+    },
+    {
+      "epoch": 56.52,
+      "learning_rate": 0.00022798076923076925,
+      "loss": 0.3494,
+      "step": 6161
+    },
+    {
+      "epoch": 56.53,
+      "learning_rate": 0.00022793269230769232,
+      "loss": 0.3199,
+      "step": 6162
+    },
+    {
+      "epoch": 56.54,
+      "learning_rate": 0.0002278846153846154,
+      "loss": 0.2707,
+      "step": 6163
+    },
+    {
+      "epoch": 56.55,
+      "learning_rate": 0.00022783653846153846,
+      "loss": 0.2547,
+      "step": 6164
+    },
+    {
+      "epoch": 56.56,
+      "learning_rate": 0.00022778846153846156,
+      "loss": 0.3768,
+      "step": 6165
+    },
+    {
+      "epoch": 56.57,
+      "learning_rate": 0.00022774038461538464,
+      "loss": 0.3715,
+      "step": 6166
+    },
+    {
+      "epoch": 56.58,
+      "learning_rate": 0.0002276923076923077,
+      "loss": 0.3623,
+      "step": 6167
+    },
+    {
+      "epoch": 56.59,
+      "learning_rate": 0.00022764423076923078,
+      "loss": 0.4305,
+      "step": 6168
+    },
+    {
+      "epoch": 56.6,
+      "learning_rate": 0.00022759615384615386,
+      "loss": 0.3686,
+      "step": 6169
+    },
+    {
+      "epoch": 56.61,
+      "learning_rate": 0.00022754807692307693,
+      "loss": 0.2469,
+      "step": 6170
+    },
+    {
+      "epoch": 56.61,
+      "learning_rate": 0.0002275,
+      "loss": 0.3072,
+      "step": 6171
+    },
+    {
+      "epoch": 56.62,
+      "learning_rate": 0.00022745192307692307,
+      "loss": 0.28,
+      "step": 6172
+    },
+    {
+      "epoch": 56.63,
+      "learning_rate": 0.00022740384615384615,
+      "loss": 0.3259,
+      "step": 6173
+    },
+    {
+      "epoch": 56.64,
+      "learning_rate": 0.00022735576923076922,
+      "loss": 0.3745,
+      "step": 6174
+    },
+    {
+      "epoch": 56.65,
+      "learning_rate": 0.0002273076923076923,
+      "loss": 0.3366,
+      "step": 6175
+    },
+    {
+      "epoch": 56.66,
+      "learning_rate": 0.0002272596153846154,
+      "loss": 0.3325,
+      "step": 6176
+    },
+    {
+      "epoch": 56.67,
+      "learning_rate": 0.00022721153846153846,
+      "loss": 0.2643,
+      "step": 6177
+    },
+    {
+      "epoch": 56.68,
+      "learning_rate": 0.00022716346153846154,
+      "loss": 0.3604,
+      "step": 6178
+    },
+    {
+      "epoch": 56.69,
+      "learning_rate": 0.0002271153846153846,
+      "loss": 0.3533,
+      "step": 6179
+    },
+    {
+      "epoch": 56.7,
+      "learning_rate": 0.00022706730769230768,
+      "loss": 0.4156,
+      "step": 6180
+    },
+    {
+      "epoch": 56.71,
+      "learning_rate": 0.00022701923076923078,
+      "loss": 0.2525,
+      "step": 6181
+    },
+    {
+      "epoch": 56.72,
+      "learning_rate": 0.00022697115384615385,
+      "loss": 0.3069,
+      "step": 6182
+    },
+    {
+      "epoch": 56.72,
+      "learning_rate": 0.00022692307692307693,
+      "loss": 0.3408,
+      "step": 6183
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 0.000226875,
+      "loss": 0.3091,
+      "step": 6184
+    },
+    {
+      "epoch": 56.74,
+      "learning_rate": 0.00022682692307692307,
+      "loss": 0.235,
+      "step": 6185
+    },
+    {
+      "epoch": 56.75,
+      "learning_rate": 0.00022677884615384617,
+      "loss": 0.3878,
+      "step": 6186
+    },
+    {
+      "epoch": 56.76,
+      "learning_rate": 0.00022673076923076924,
+      "loss": 0.3632,
+      "step": 6187
+    },
+    {
+      "epoch": 56.77,
+      "learning_rate": 0.00022668269230769232,
+      "loss": 0.3065,
+      "step": 6188
+    },
+    {
+      "epoch": 56.78,
+      "learning_rate": 0.0002266346153846154,
+      "loss": 0.2916,
+      "step": 6189
+    },
+    {
+      "epoch": 56.79,
+      "learning_rate": 0.00022658653846153846,
+      "loss": 0.3588,
+      "step": 6190
+    },
+    {
+      "epoch": 56.8,
+      "learning_rate": 0.00022653846153846156,
+      "loss": 0.2494,
+      "step": 6191
+    },
+    {
+      "epoch": 56.81,
+      "learning_rate": 0.00022649038461538463,
+      "loss": 0.3146,
+      "step": 6192
+    },
+    {
+      "epoch": 56.82,
+      "learning_rate": 0.0002264423076923077,
+      "loss": 0.3186,
+      "step": 6193
+    },
+    {
+      "epoch": 56.83,
+      "learning_rate": 0.00022639423076923078,
+      "loss": 0.3144,
+      "step": 6194
+    },
+    {
+      "epoch": 56.83,
+      "learning_rate": 0.00022634615384615385,
+      "loss": 0.3156,
+      "step": 6195
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 0.00022629807692307695,
+      "loss": 0.2812,
+      "step": 6196
+    },
+    {
+      "epoch": 56.85,
+      "learning_rate": 0.00022625000000000002,
+      "loss": 0.2701,
+      "step": 6197
+    },
+    {
+      "epoch": 56.86,
+      "learning_rate": 0.00022620192307692307,
+      "loss": 0.2959,
+      "step": 6198
+    },
+    {
+      "epoch": 56.87,
+      "learning_rate": 0.00022615384615384614,
+      "loss": 0.2773,
+      "step": 6199
+    },
+    {
+      "epoch": 56.88,
+      "learning_rate": 0.00022610576923076922,
+      "loss": 0.3175,
+      "step": 6200
+    },
+    {
+      "epoch": 56.89,
+      "learning_rate": 0.00022605769230769231,
+      "loss": 0.353,
+      "step": 6201
+    },
+    {
+      "epoch": 56.9,
+      "learning_rate": 0.0002260096153846154,
+      "loss": 0.3266,
+      "step": 6202
+    },
+    {
+      "epoch": 56.91,
+      "learning_rate": 0.00022596153846153846,
+      "loss": 0.2402,
+      "step": 6203
+    },
+    {
+      "epoch": 56.92,
+      "learning_rate": 0.00022591346153846153,
+      "loss": 0.3387,
+      "step": 6204
+    },
+    {
+      "epoch": 56.93,
+      "learning_rate": 0.0002258653846153846,
+      "loss": 0.2607,
+      "step": 6205
+    },
+    {
+      "epoch": 56.94,
+      "learning_rate": 0.0002258173076923077,
+      "loss": 0.3463,
+      "step": 6206
+    },
+    {
+      "epoch": 56.94,
+      "learning_rate": 0.00022576923076923078,
+      "loss": 0.2706,
+      "step": 6207
+    },
+    {
+      "epoch": 56.95,
+      "learning_rate": 0.00022572115384615385,
+      "loss": 0.297,
+      "step": 6208
+    },
+    {
+      "epoch": 56.96,
+      "learning_rate": 0.00022567307692307692,
+      "loss": 0.2306,
+      "step": 6209
+    },
+    {
+      "epoch": 56.97,
+      "learning_rate": 0.000225625,
+      "loss": 0.2133,
+      "step": 6210
+    },
+    {
+      "epoch": 56.98,
+      "learning_rate": 0.00022557692307692307,
+      "loss": 0.2232,
+      "step": 6211
+    },
+    {
+      "epoch": 56.99,
+      "learning_rate": 0.00022552884615384617,
+      "loss": 0.354,
+      "step": 6212
+    },
+    {
+      "epoch": 57.0,
+      "learning_rate": 0.00022548076923076924,
+      "loss": 0.386,
+      "step": 6213
+    },
+    {
+      "epoch": 57.01,
+      "learning_rate": 0.0002254326923076923,
+      "loss": 0.3908,
+      "step": 6214
+    },
+    {
+      "epoch": 57.02,
+      "learning_rate": 0.00022538461538461539,
+      "loss": 0.3677,
+      "step": 6215
+    },
+    {
+      "epoch": 57.03,
+      "learning_rate": 0.00022533653846153846,
+      "loss": 0.3563,
+      "step": 6216
+    },
+    {
+      "epoch": 57.04,
+      "learning_rate": 0.00022528846153846156,
+      "loss": 0.253,
+      "step": 6217
+    },
+    {
+      "epoch": 57.05,
+      "learning_rate": 0.00022524038461538463,
+      "loss": 0.2994,
+      "step": 6218
+    },
+    {
+      "epoch": 57.06,
+      "learning_rate": 0.0002251923076923077,
+      "loss": 0.3328,
+      "step": 6219
+    },
+    {
+      "epoch": 57.06,
+      "learning_rate": 0.00022514423076923078,
+      "loss": 0.3488,
+      "step": 6220
+    },
+    {
+      "epoch": 57.07,
+      "learning_rate": 0.00022509615384615385,
+      "loss": 0.2667,
+      "step": 6221
+    },
+    {
+      "epoch": 57.08,
+      "learning_rate": 0.00022504807692307695,
+      "loss": 0.3001,
+      "step": 6222
+    },
+    {
+      "epoch": 57.09,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 0.2605,
+      "step": 6223
+    },
+    {
+      "epoch": 57.1,
+      "learning_rate": 0.0002249519230769231,
+      "loss": 0.2561,
+      "step": 6224
+    },
+    {
+      "epoch": 57.11,
+      "learning_rate": 0.00022490384615384617,
+      "loss": 0.2774,
+      "step": 6225
+    },
+    {
+      "epoch": 57.12,
+      "learning_rate": 0.0002248557692307692,
+      "loss": 0.2566,
+      "step": 6226
+    },
+    {
+      "epoch": 57.13,
+      "learning_rate": 0.0002248076923076923,
+      "loss": 0.2247,
+      "step": 6227
+    },
+    {
+      "epoch": 57.14,
+      "learning_rate": 0.00022475961538461538,
+      "loss": 0.2692,
+      "step": 6228
+    },
+    {
+      "epoch": 57.15,
+      "learning_rate": 0.00022471153846153846,
+      "loss": 0.2499,
+      "step": 6229
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 0.00022466346153846153,
+      "loss": 0.2565,
+      "step": 6230
+    },
+    {
+      "epoch": 57.17,
+      "learning_rate": 0.0002246153846153846,
+      "loss": 0.2577,
+      "step": 6231
+    },
+    {
+      "epoch": 57.17,
+      "learning_rate": 0.0002245673076923077,
+      "loss": 0.281,
+      "step": 6232
+    },
+    {
+      "epoch": 57.18,
+      "learning_rate": 0.00022451923076923077,
+      "loss": 0.3019,
+      "step": 6233
+    },
+    {
+      "epoch": 57.19,
+      "learning_rate": 0.00022447115384615385,
+      "loss": 0.2929,
+      "step": 6234
+    },
+    {
+      "epoch": 57.2,
+      "learning_rate": 0.00022442307692307692,
+      "loss": 0.2904,
+      "step": 6235
+    },
+    {
+      "epoch": 57.21,
+      "learning_rate": 0.000224375,
+      "loss": 0.2889,
+      "step": 6236
+    },
+    {
+      "epoch": 57.22,
+      "learning_rate": 0.0002243269230769231,
+      "loss": 0.2468,
+      "step": 6237
+    },
+    {
+      "epoch": 57.23,
+      "learning_rate": 0.00022427884615384616,
+      "loss": 0.3063,
+      "step": 6238
+    },
+    {
+      "epoch": 57.24,
+      "learning_rate": 0.00022423076923076924,
+      "loss": 0.3085,
+      "step": 6239
+    },
+    {
+      "epoch": 57.25,
+      "learning_rate": 0.0002241826923076923,
+      "loss": 0.1772,
+      "step": 6240
+    },
+    {
+      "epoch": 57.26,
+      "learning_rate": 0.00022413461538461538,
+      "loss": 0.4217,
+      "step": 6241
+    },
+    {
+      "epoch": 57.27,
+      "learning_rate": 0.00022408653846153846,
+      "loss": 0.3723,
+      "step": 6242
+    },
+    {
+      "epoch": 57.28,
+      "learning_rate": 0.00022403846153846155,
+      "loss": 0.3141,
+      "step": 6243
+    },
+    {
+      "epoch": 57.28,
+      "learning_rate": 0.00022399038461538463,
+      "loss": 0.3552,
+      "step": 6244
+    },
+    {
+      "epoch": 57.29,
+      "learning_rate": 0.0002239423076923077,
+      "loss": 0.2699,
+      "step": 6245
+    },
+    {
+      "epoch": 57.3,
+      "learning_rate": 0.00022389423076923077,
+      "loss": 0.2404,
+      "step": 6246
+    },
+    {
+      "epoch": 57.31,
+      "learning_rate": 0.00022384615384615385,
+      "loss": 0.2895,
+      "step": 6247
+    },
+    {
+      "epoch": 57.32,
+      "learning_rate": 0.00022379807692307695,
+      "loss": 0.2628,
+      "step": 6248
+    },
+    {
+      "epoch": 57.33,
+      "learning_rate": 0.00022375000000000002,
+      "loss": 0.3444,
+      "step": 6249
+    },
+    {
+      "epoch": 57.34,
+      "learning_rate": 0.0002237019230769231,
+      "loss": 0.2649,
+      "step": 6250
+    },
+    {
+      "epoch": 57.35,
+      "learning_rate": 0.00022365384615384616,
+      "loss": 0.2906,
+      "step": 6251
+    },
+    {
+      "epoch": 57.36,
+      "learning_rate": 0.00022360576923076924,
+      "loss": 0.3929,
+      "step": 6252
+    },
+    {
+      "epoch": 57.37,
+      "learning_rate": 0.00022355769230769234,
+      "loss": 0.3038,
+      "step": 6253
+    },
+    {
+      "epoch": 57.38,
+      "learning_rate": 0.0002235096153846154,
+      "loss": 0.33,
+      "step": 6254
+    },
+    {
+      "epoch": 57.39,
+      "learning_rate": 0.00022346153846153845,
+      "loss": 0.3335,
+      "step": 6255
+    },
+    {
+      "epoch": 57.39,
+      "learning_rate": 0.00022341346153846153,
+      "loss": 0.4305,
+      "step": 6256
+    },
+    {
+      "epoch": 57.4,
+      "learning_rate": 0.0002233653846153846,
+      "loss": 0.2368,
+      "step": 6257
+    },
+    {
+      "epoch": 57.41,
+      "learning_rate": 0.0002233173076923077,
+      "loss": 0.1646,
+      "step": 6258
+    },
+    {
+      "epoch": 57.42,
+      "learning_rate": 0.00022326923076923077,
+      "loss": 0.3231,
+      "step": 6259
+    },
+    {
+      "epoch": 57.43,
+      "learning_rate": 0.00022322115384615384,
+      "loss": 0.2389,
+      "step": 6260
+    },
+    {
+      "epoch": 57.44,
+      "learning_rate": 0.00022317307692307692,
+      "loss": 0.3072,
+      "step": 6261
+    },
+    {
+      "epoch": 57.45,
+      "learning_rate": 0.000223125,
+      "loss": 0.3632,
+      "step": 6262
+    },
+    {
+      "epoch": 57.46,
+      "learning_rate": 0.0002230769230769231,
+      "loss": 0.2827,
+      "step": 6263
+    },
+    {
+      "epoch": 57.47,
+      "learning_rate": 0.00022302884615384616,
+      "loss": 0.2945,
+      "step": 6264
+    },
+    {
+      "epoch": 57.48,
+      "learning_rate": 0.00022298076923076923,
+      "loss": 0.3357,
+      "step": 6265
+    },
+    {
+      "epoch": 57.49,
+      "learning_rate": 0.0002229326923076923,
+      "loss": 0.3851,
+      "step": 6266
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 0.00022288461538461538,
+      "loss": 0.1535,
+      "step": 6267
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 0.00022283653846153848,
+      "loss": 0.4469,
+      "step": 6268
+    },
+    {
+      "epoch": 57.51,
+      "learning_rate": 0.00022278846153846155,
+      "loss": 0.3048,
+      "step": 6269
+    },
+    {
+      "epoch": 57.52,
+      "learning_rate": 0.00022274038461538462,
+      "loss": 0.3561,
+      "step": 6270
+    },
+    {
+      "epoch": 57.53,
+      "learning_rate": 0.0002226923076923077,
+      "loss": 0.4269,
+      "step": 6271
+    },
+    {
+      "epoch": 57.54,
+      "learning_rate": 0.00022264423076923077,
+      "loss": 0.331,
+      "step": 6272
+    },
+    {
+      "epoch": 57.55,
+      "learning_rate": 0.00022259615384615384,
+      "loss": 0.3606,
+      "step": 6273
+    },
+    {
+      "epoch": 57.56,
+      "learning_rate": 0.00022254807692307694,
+      "loss": 0.3458,
+      "step": 6274
+    },
+    {
+      "epoch": 57.57,
+      "learning_rate": 0.00022250000000000001,
+      "loss": 0.2753,
+      "step": 6275
+    },
+    {
+      "epoch": 57.58,
+      "learning_rate": 0.0002224519230769231,
+      "loss": 0.3051,
+      "step": 6276
+    },
+    {
+      "epoch": 57.59,
+      "learning_rate": 0.00022240384615384616,
+      "loss": 0.3832,
+      "step": 6277
+    },
+    {
+      "epoch": 57.6,
+      "learning_rate": 0.00022235576923076923,
+      "loss": 0.2729,
+      "step": 6278
+    },
+    {
+      "epoch": 57.61,
+      "learning_rate": 0.00022230769230769233,
+      "loss": 0.34,
+      "step": 6279
+    },
+    {
+      "epoch": 57.61,
+      "learning_rate": 0.0002222596153846154,
+      "loss": 0.352,
+      "step": 6280
+    },
+    {
+      "epoch": 57.62,
+      "learning_rate": 0.00022221153846153848,
+      "loss": 0.3148,
+      "step": 6281
+    },
+    {
+      "epoch": 57.63,
+      "learning_rate": 0.00022216346153846155,
+      "loss": 0.3488,
+      "step": 6282
+    },
+    {
+      "epoch": 57.64,
+      "learning_rate": 0.0002221153846153846,
+      "loss": 0.3901,
+      "step": 6283
+    },
+    {
+      "epoch": 57.65,
+      "learning_rate": 0.0002220673076923077,
+      "loss": 0.3943,
+      "step": 6284
+    },
+    {
+      "epoch": 57.66,
+      "learning_rate": 0.00022201923076923077,
+      "loss": 0.3059,
+      "step": 6285
+    },
+    {
+      "epoch": 57.67,
+      "learning_rate": 0.00022197115384615384,
+      "loss": 0.2494,
+      "step": 6286
+    },
+    {
+      "epoch": 57.68,
+      "learning_rate": 0.0002219230769230769,
+      "loss": 0.3234,
+      "step": 6287
+    },
+    {
+      "epoch": 57.69,
+      "learning_rate": 0.00022187499999999999,
+      "loss": 0.2813,
+      "step": 6288
+    },
+    {
+      "epoch": 57.7,
+      "learning_rate": 0.00022182692307692309,
+      "loss": 0.2845,
+      "step": 6289
+    },
+    {
+      "epoch": 57.71,
+      "learning_rate": 0.00022177884615384616,
+      "loss": 0.2632,
+      "step": 6290
+    },
+    {
+      "epoch": 57.72,
+      "learning_rate": 0.00022173076923076923,
+      "loss": 0.2554,
+      "step": 6291
+    },
+    {
+      "epoch": 57.72,
+      "learning_rate": 0.0002216826923076923,
+      "loss": 0.2241,
+      "step": 6292
+    },
+    {
+      "epoch": 57.73,
+      "learning_rate": 0.00022163461538461538,
+      "loss": 0.2099,
+      "step": 6293
+    },
+    {
+      "epoch": 57.74,
+      "learning_rate": 0.00022158653846153848,
+      "loss": 0.2695,
+      "step": 6294
+    },
+    {
+      "epoch": 57.75,
+      "learning_rate": 0.00022153846153846155,
+      "loss": 0.4385,
+      "step": 6295
+    },
+    {
+      "epoch": 57.76,
+      "learning_rate": 0.00022149038461538462,
+      "loss": 0.3258,
+      "step": 6296
+    },
+    {
+      "epoch": 57.77,
+      "learning_rate": 0.0002214423076923077,
+      "loss": 0.3464,
+      "step": 6297
+    },
+    {
+      "epoch": 57.78,
+      "learning_rate": 0.00022139423076923077,
+      "loss": 0.3512,
+      "step": 6298
+    },
+    {
+      "epoch": 57.79,
+      "learning_rate": 0.00022134615384615387,
+      "loss": 0.3789,
+      "step": 6299
+    },
+    {
+      "epoch": 57.8,
+      "learning_rate": 0.00022129807692307694,
+      "loss": 0.3051,
+      "step": 6300
+    },
+    {
+      "epoch": 57.81,
+      "learning_rate": 0.00022125,
+      "loss": 0.295,
+      "step": 6301
+    },
+    {
+      "epoch": 57.82,
+      "learning_rate": 0.00022120192307692308,
+      "loss": 0.2801,
+      "step": 6302
+    },
+    {
+      "epoch": 57.83,
+      "learning_rate": 0.00022115384615384616,
+      "loss": 0.355,
+      "step": 6303
+    },
+    {
+      "epoch": 57.83,
+      "learning_rate": 0.00022110576923076926,
+      "loss": 0.2182,
+      "step": 6304
+    },
+    {
+      "epoch": 57.84,
+      "learning_rate": 0.00022105769230769233,
+      "loss": 0.329,
+      "step": 6305
+    },
+    {
+      "epoch": 57.85,
+      "learning_rate": 0.0002210096153846154,
+      "loss": 0.3144,
+      "step": 6306
+    },
+    {
+      "epoch": 57.86,
+      "learning_rate": 0.00022096153846153847,
+      "loss": 0.278,
+      "step": 6307
+    },
+    {
+      "epoch": 57.87,
+      "learning_rate": 0.00022091346153846155,
+      "loss": 0.2203,
+      "step": 6308
+    },
+    {
+      "epoch": 57.88,
+      "learning_rate": 0.00022086538461538462,
+      "loss": 0.2946,
+      "step": 6309
+    },
+    {
+      "epoch": 57.89,
+      "learning_rate": 0.00022081730769230772,
+      "loss": 0.3462,
+      "step": 6310
+    },
+    {
+      "epoch": 57.9,
+      "learning_rate": 0.00022076923076923076,
+      "loss": 0.2477,
+      "step": 6311
+    },
+    {
+      "epoch": 57.91,
+      "learning_rate": 0.00022072115384615384,
+      "loss": 0.2321,
+      "step": 6312
+    },
+    {
+      "epoch": 57.92,
+      "learning_rate": 0.0002206730769230769,
+      "loss": 0.3921,
+      "step": 6313
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 0.00022062499999999998,
+      "loss": 0.3428,
+      "step": 6314
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 0.00022057692307692308,
+      "loss": 0.2667,
+      "step": 6315
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 0.00022052884615384615,
+      "loss": 0.2838,
+      "step": 6316
+    },
+    {
+      "epoch": 57.95,
+      "learning_rate": 0.00022048076923076923,
+      "loss": 0.4016,
+      "step": 6317
+    },
+    {
+      "epoch": 57.96,
+      "learning_rate": 0.0002204326923076923,
+      "loss": 0.2556,
+      "step": 6318
+    },
+    {
+      "epoch": 57.97,
+      "learning_rate": 0.00022038461538461537,
+      "loss": 0.4576,
+      "step": 6319
+    },
+    {
+      "epoch": 57.98,
+      "learning_rate": 0.00022033653846153847,
+      "loss": 0.2755,
+      "step": 6320
+    },
+    {
+      "epoch": 57.99,
+      "learning_rate": 0.00022028846153846154,
+      "loss": 0.2315,
+      "step": 6321
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 0.00022024038461538462,
+      "loss": 0.3088,
+      "step": 6322
+    },
+    {
+      "epoch": 58.01,
+      "learning_rate": 0.0002201923076923077,
+      "loss": 0.3544,
+      "step": 6323
+    },
+    {
+      "epoch": 58.02,
+      "learning_rate": 0.00022014423076923076,
+      "loss": 0.2799,
+      "step": 6324
+    },
+    {
+      "epoch": 58.03,
+      "learning_rate": 0.00022009615384615386,
+      "loss": 0.2921,
+      "step": 6325
+    },
+    {
+      "epoch": 58.04,
+      "learning_rate": 0.00022004807692307694,
+      "loss": 0.2611,
+      "step": 6326
+    },
+    {
+      "epoch": 58.05,
+      "learning_rate": 0.00022,
+      "loss": 0.3341,
+      "step": 6327
+    },
+    {
+      "epoch": 58.06,
+      "learning_rate": 0.00021995192307692308,
+      "loss": 0.3028,
+      "step": 6328
+    },
+    {
+      "epoch": 58.06,
+      "learning_rate": 0.00021990384615384615,
+      "loss": 0.2434,
+      "step": 6329
+    },
+    {
+      "epoch": 58.07,
+      "learning_rate": 0.00021985576923076925,
+      "loss": 0.2736,
+      "step": 6330
+    },
+    {
+      "epoch": 58.08,
+      "learning_rate": 0.00021980769230769233,
+      "loss": 0.2575,
+      "step": 6331
+    },
+    {
+      "epoch": 58.09,
+      "learning_rate": 0.0002197596153846154,
+      "loss": 0.2199,
+      "step": 6332
+    },
+    {
+      "epoch": 58.1,
+      "learning_rate": 0.00021971153846153847,
+      "loss": 0.228,
+      "step": 6333
+    },
+    {
+      "epoch": 58.11,
+      "learning_rate": 0.00021966346153846154,
+      "loss": 0.244,
+      "step": 6334
+    },
+    {
+      "epoch": 58.12,
+      "learning_rate": 0.00021961538461538464,
+      "loss": 0.2923,
+      "step": 6335
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 0.00021956730769230772,
+      "loss": 0.1965,
+      "step": 6336
+    },
+    {
+      "epoch": 58.14,
+      "learning_rate": 0.0002195192307692308,
+      "loss": 0.3051,
+      "step": 6337
+    },
+    {
+      "epoch": 58.15,
+      "learning_rate": 0.00021947115384615386,
+      "loss": 0.3287,
+      "step": 6338
+    },
+    {
+      "epoch": 58.16,
+      "learning_rate": 0.0002194230769230769,
+      "loss": 0.3358,
+      "step": 6339
+    },
+    {
+      "epoch": 58.17,
+      "learning_rate": 0.00021937499999999998,
+      "loss": 0.2089,
+      "step": 6340
+    },
+    {
+      "epoch": 58.17,
+      "learning_rate": 0.00021932692307692308,
+      "loss": 0.254,
+      "step": 6341
+    },
+    {
+      "epoch": 58.18,
+      "learning_rate": 0.00021927884615384615,
+      "loss": 0.2743,
+      "step": 6342
+    },
+    {
+      "epoch": 58.19,
+      "learning_rate": 0.00021923076923076922,
+      "loss": 0.3248,
+      "step": 6343
+    },
+    {
+      "epoch": 58.2,
+      "learning_rate": 0.0002191826923076923,
+      "loss": 0.3282,
+      "step": 6344
+    },
+    {
+      "epoch": 58.21,
+      "learning_rate": 0.00021913461538461537,
+      "loss": 0.1682,
+      "step": 6345
+    },
+    {
+      "epoch": 58.22,
+      "learning_rate": 0.00021908653846153847,
+      "loss": 0.2986,
+      "step": 6346
+    },
+    {
+      "epoch": 58.23,
+      "learning_rate": 0.00021903846153846154,
+      "loss": 0.2509,
+      "step": 6347
+    },
+    {
+      "epoch": 58.24,
+      "learning_rate": 0.00021899038461538461,
+      "loss": 0.2619,
+      "step": 6348
+    },
+    {
+      "epoch": 58.25,
+      "learning_rate": 0.0002189423076923077,
+      "loss": 0.1386,
+      "step": 6349
+    },
+    {
+      "epoch": 58.26,
+      "learning_rate": 0.00021889423076923076,
+      "loss": 0.4813,
+      "step": 6350
+    },
+    {
+      "epoch": 58.27,
+      "learning_rate": 0.00021884615384615386,
+      "loss": 0.3662,
+      "step": 6351
+    },
+    {
+      "epoch": 58.28,
+      "learning_rate": 0.00021879807692307693,
+      "loss": 0.3026,
+      "step": 6352
+    },
+    {
+      "epoch": 58.28,
+      "learning_rate": 0.00021875,
+      "loss": 0.3346,
+      "step": 6353
+    },
+    {
+      "epoch": 58.29,
+      "learning_rate": 0.00021870192307692308,
+      "loss": 0.2812,
+      "step": 6354
+    },
+    {
+      "epoch": 58.3,
+      "learning_rate": 0.00021865384615384615,
+      "loss": 0.2927,
+      "step": 6355
+    },
+    {
+      "epoch": 58.31,
+      "learning_rate": 0.00021860576923076925,
+      "loss": 0.3214,
+      "step": 6356
+    },
+    {
+      "epoch": 58.32,
+      "learning_rate": 0.00021855769230769232,
+      "loss": 0.3578,
+      "step": 6357
+    },
+    {
+      "epoch": 58.33,
+      "learning_rate": 0.0002185096153846154,
+      "loss": 0.2811,
+      "step": 6358
+    },
+    {
+      "epoch": 58.34,
+      "learning_rate": 0.00021846153846153847,
+      "loss": 0.2982,
+      "step": 6359
+    },
+    {
+      "epoch": 58.35,
+      "learning_rate": 0.00021841346153846154,
+      "loss": 0.3093,
+      "step": 6360
+    },
+    {
+      "epoch": 58.36,
+      "learning_rate": 0.00021836538461538464,
+      "loss": 0.3234,
+      "step": 6361
+    },
+    {
+      "epoch": 58.37,
+      "learning_rate": 0.0002183173076923077,
+      "loss": 0.2703,
+      "step": 6362
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 0.00021826923076923079,
+      "loss": 0.3064,
+      "step": 6363
+    },
+    {
+      "epoch": 58.39,
+      "learning_rate": 0.00021822115384615386,
+      "loss": 0.2626,
+      "step": 6364
+    },
+    {
+      "epoch": 58.39,
+      "learning_rate": 0.00021817307692307693,
+      "loss": 0.3155,
+      "step": 6365
+    },
+    {
+      "epoch": 58.4,
+      "learning_rate": 0.00021812500000000003,
+      "loss": 0.2119,
+      "step": 6366
+    },
+    {
+      "epoch": 58.41,
+      "learning_rate": 0.0002180769230769231,
+      "loss": 0.3175,
+      "step": 6367
+    },
+    {
+      "epoch": 58.42,
+      "learning_rate": 0.00021802884615384615,
+      "loss": 0.3308,
+      "step": 6368
+    },
+    {
+      "epoch": 58.43,
+      "learning_rate": 0.00021798076923076922,
+      "loss": 0.2835,
+      "step": 6369
+    },
+    {
+      "epoch": 58.44,
+      "learning_rate": 0.0002179326923076923,
+      "loss": 0.2501,
+      "step": 6370
+    },
+    {
+      "epoch": 58.45,
+      "learning_rate": 0.00021788461538461537,
+      "loss": 0.2587,
+      "step": 6371
+    },
+    {
+      "epoch": 58.46,
+      "learning_rate": 0.00021783653846153847,
+      "loss": 0.3046,
+      "step": 6372
+    },
+    {
+      "epoch": 58.47,
+      "learning_rate": 0.00021778846153846154,
+      "loss": 0.2549,
+      "step": 6373
+    },
+    {
+      "epoch": 58.48,
+      "learning_rate": 0.0002177403846153846,
+      "loss": 0.3256,
+      "step": 6374
+    },
+    {
+      "epoch": 58.49,
+      "learning_rate": 0.00021769230769230768,
+      "loss": 0.2916,
+      "step": 6375
+    },
+    {
+      "epoch": 58.5,
+      "learning_rate": 0.00021764423076923076,
+      "loss": 0.2528,
+      "step": 6376
+    },
+    {
+      "epoch": 58.5,
+      "learning_rate": 0.00021759615384615386,
+      "loss": 0.4508,
+      "step": 6377
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 0.00021754807692307693,
+      "loss": 0.3211,
+      "step": 6378
+    },
+    {
+      "epoch": 58.52,
+      "learning_rate": 0.0002175,
+      "loss": 0.3385,
+      "step": 6379
+    },
+    {
+      "epoch": 58.53,
+      "learning_rate": 0.00021745192307692307,
+      "loss": 0.2957,
+      "step": 6380
+    },
+    {
+      "epoch": 58.54,
+      "learning_rate": 0.00021740384615384615,
+      "loss": 0.3925,
+      "step": 6381
+    },
+    {
+      "epoch": 58.55,
+      "learning_rate": 0.00021735576923076925,
+      "loss": 0.3732,
+      "step": 6382
+    },
+    {
+      "epoch": 58.56,
+      "learning_rate": 0.00021730769230769232,
+      "loss": 0.3073,
+      "step": 6383
+    },
+    {
+      "epoch": 58.57,
+      "learning_rate": 0.0002172596153846154,
+      "loss": 0.3031,
+      "step": 6384
+    },
+    {
+      "epoch": 58.58,
+      "learning_rate": 0.00021721153846153846,
+      "loss": 0.3247,
+      "step": 6385
+    },
+    {
+      "epoch": 58.59,
+      "learning_rate": 0.00021716346153846154,
+      "loss": 0.2988,
+      "step": 6386
+    },
+    {
+      "epoch": 58.6,
+      "learning_rate": 0.00021711538461538464,
+      "loss": 0.2992,
+      "step": 6387
+    },
+    {
+      "epoch": 58.61,
+      "learning_rate": 0.0002170673076923077,
+      "loss": 0.3132,
+      "step": 6388
+    },
+    {
+      "epoch": 58.61,
+      "learning_rate": 0.00021701923076923078,
+      "loss": 0.197,
+      "step": 6389
+    },
+    {
+      "epoch": 58.62,
+      "learning_rate": 0.00021697115384615385,
+      "loss": 0.3616,
+      "step": 6390
+    },
+    {
+      "epoch": 58.63,
+      "learning_rate": 0.00021692307692307693,
+      "loss": 0.3805,
+      "step": 6391
+    },
+    {
+      "epoch": 58.64,
+      "learning_rate": 0.00021687500000000003,
+      "loss": 0.2599,
+      "step": 6392
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 0.0002168269230769231,
+      "loss": 0.2516,
+      "step": 6393
+    },
+    {
+      "epoch": 58.66,
+      "learning_rate": 0.00021677884615384617,
+      "loss": 0.2038,
+      "step": 6394
+    },
+    {
+      "epoch": 58.67,
+      "learning_rate": 0.00021673076923076924,
+      "loss": 0.3281,
+      "step": 6395
+    },
+    {
+      "epoch": 58.68,
+      "learning_rate": 0.0002166826923076923,
+      "loss": 0.267,
+      "step": 6396
+    },
+    {
+      "epoch": 58.69,
+      "learning_rate": 0.0002166346153846154,
+      "loss": 0.2951,
+      "step": 6397
+    },
+    {
+      "epoch": 58.7,
+      "learning_rate": 0.00021658653846153846,
+      "loss": 0.221,
+      "step": 6398
+    },
+    {
+      "epoch": 58.71,
+      "learning_rate": 0.00021653846153846154,
+      "loss": 0.2731,
+      "step": 6399
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 0.0002164903846153846,
+      "loss": 0.3046,
+      "step": 6400
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 0.00021644230769230768,
+      "loss": 0.3731,
+      "step": 6401
+    },
+    {
+      "epoch": 58.73,
+      "learning_rate": 0.00021639423076923075,
+      "loss": 0.2622,
+      "step": 6402
+    },
+    {
+      "epoch": 58.74,
+      "learning_rate": 0.00021634615384615385,
+      "loss": 0.2117,
+      "step": 6403
+    },
+    {
+      "epoch": 58.75,
+      "learning_rate": 0.00021629807692307693,
+      "loss": 0.3447,
+      "step": 6404
+    },
+    {
+      "epoch": 58.76,
+      "learning_rate": 0.00021625,
+      "loss": 0.3701,
+      "step": 6405
+    },
+    {
+      "epoch": 58.77,
+      "learning_rate": 0.00021620192307692307,
+      "loss": 0.2929,
+      "step": 6406
+    },
+    {
+      "epoch": 58.78,
+      "learning_rate": 0.00021615384615384614,
+      "loss": 0.3269,
+      "step": 6407
+    },
+    {
+      "epoch": 58.79,
+      "learning_rate": 0.00021610576923076924,
+      "loss": 0.3807,
+      "step": 6408
+    },
+    {
+      "epoch": 58.8,
+      "learning_rate": 0.00021605769230769232,
+      "loss": 0.3675,
+      "step": 6409
+    },
+    {
+      "epoch": 58.81,
+      "learning_rate": 0.0002160096153846154,
+      "loss": 0.2861,
+      "step": 6410
+    },
+    {
+      "epoch": 58.82,
+      "learning_rate": 0.00021596153846153846,
+      "loss": 0.2894,
+      "step": 6411
+    },
+    {
+      "epoch": 58.83,
+      "learning_rate": 0.00021591346153846153,
+      "loss": 0.3489,
+      "step": 6412
+    },
+    {
+      "epoch": 58.83,
+      "learning_rate": 0.00021586538461538463,
+      "loss": 0.3252,
+      "step": 6413
+    },
+    {
+      "epoch": 58.84,
+      "learning_rate": 0.0002158173076923077,
+      "loss": 0.3573,
+      "step": 6414
+    },
+    {
+      "epoch": 58.85,
+      "learning_rate": 0.00021576923076923078,
+      "loss": 0.2516,
+      "step": 6415
+    },
+    {
+      "epoch": 58.86,
+      "learning_rate": 0.00021572115384615385,
+      "loss": 0.3022,
+      "step": 6416
+    },
+    {
+      "epoch": 58.87,
+      "learning_rate": 0.00021567307692307692,
+      "loss": 0.3312,
+      "step": 6417
+    },
+    {
+      "epoch": 58.88,
+      "learning_rate": 0.00021562500000000002,
+      "loss": 0.3248,
+      "step": 6418
+    },
+    {
+      "epoch": 58.89,
+      "learning_rate": 0.0002155769230769231,
+      "loss": 0.2878,
+      "step": 6419
+    },
+    {
+      "epoch": 58.9,
+      "learning_rate": 0.00021552884615384617,
+      "loss": 0.3026,
+      "step": 6420
+    },
+    {
+      "epoch": 58.91,
+      "learning_rate": 0.00021548076923076924,
+      "loss": 0.2207,
+      "step": 6421
+    },
+    {
+      "epoch": 58.92,
+      "learning_rate": 0.00021543269230769231,
+      "loss": 0.3713,
+      "step": 6422
+    },
+    {
+      "epoch": 58.93,
+      "learning_rate": 0.00021538461538461541,
+      "loss": 0.3171,
+      "step": 6423
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 0.00021533653846153846,
+      "loss": 0.3145,
+      "step": 6424
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 0.00021528846153846153,
+      "loss": 0.2641,
+      "step": 6425
+    },
+    {
+      "epoch": 58.95,
+      "learning_rate": 0.0002152403846153846,
+      "loss": 0.2594,
+      "step": 6426
+    },
+    {
+      "epoch": 58.96,
+      "learning_rate": 0.00021519230769230768,
+      "loss": 0.2271,
+      "step": 6427
+    },
+    {
+      "epoch": 58.97,
+      "learning_rate": 0.00021514423076923078,
+      "loss": 0.2856,
+      "step": 6428
+    },
+    {
+      "epoch": 58.98,
+      "learning_rate": 0.00021509615384615385,
+      "loss": 0.2711,
+      "step": 6429
+    },
+    {
+      "epoch": 58.99,
+      "learning_rate": 0.00021504807692307692,
+      "loss": 0.2611,
+      "step": 6430
+    },
+    {
+      "epoch": 59.0,
+      "learning_rate": 0.000215,
+      "loss": 0.2409,
+      "step": 6431
+    },
+    {
+      "epoch": 59.01,
+      "learning_rate": 0.00021495192307692307,
+      "loss": 0.3267,
+      "step": 6432
+    },
+    {
+      "epoch": 59.02,
+      "learning_rate": 0.00021490384615384617,
+      "loss": 0.2827,
+      "step": 6433
+    },
+    {
+      "epoch": 59.03,
+      "learning_rate": 0.00021485576923076924,
+      "loss": 0.3348,
+      "step": 6434
+    },
+    {
+      "epoch": 59.04,
+      "learning_rate": 0.0002148076923076923,
+      "loss": 0.3188,
+      "step": 6435
+    },
+    {
+      "epoch": 59.05,
+      "learning_rate": 0.00021475961538461538,
+      "loss": 0.3275,
+      "step": 6436
+    },
+    {
+      "epoch": 59.06,
+      "learning_rate": 0.00021471153846153846,
+      "loss": 0.3037,
+      "step": 6437
+    },
+    {
+      "epoch": 59.06,
+      "learning_rate": 0.00021466346153846153,
+      "loss": 0.2505,
+      "step": 6438
+    },
+    {
+      "epoch": 59.07,
+      "learning_rate": 0.00021461538461538463,
+      "loss": 0.2888,
+      "step": 6439
+    },
+    {
+      "epoch": 59.08,
+      "learning_rate": 0.0002145673076923077,
+      "loss": 0.3144,
+      "step": 6440
+    },
+    {
+      "epoch": 59.09,
+      "learning_rate": 0.00021451923076923078,
+      "loss": 0.3061,
+      "step": 6441
+    },
+    {
+      "epoch": 59.1,
+      "learning_rate": 0.00021447115384615385,
+      "loss": 0.2807,
+      "step": 6442
+    },
+    {
+      "epoch": 59.11,
+      "learning_rate": 0.00021442307692307692,
+      "loss": 0.2706,
+      "step": 6443
+    },
+    {
+      "epoch": 59.12,
+      "learning_rate": 0.00021437500000000002,
+      "loss": 0.2887,
+      "step": 6444
+    },
+    {
+      "epoch": 59.13,
+      "learning_rate": 0.0002143269230769231,
+      "loss": 0.2999,
+      "step": 6445
+    },
+    {
+      "epoch": 59.14,
+      "learning_rate": 0.00021427884615384617,
+      "loss": 0.2144,
+      "step": 6446
+    },
+    {
+      "epoch": 59.15,
+      "learning_rate": 0.00021423076923076924,
+      "loss": 0.2297,
+      "step": 6447
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 0.0002141826923076923,
+      "loss": 0.2637,
+      "step": 6448
+    },
+    {
+      "epoch": 59.17,
+      "learning_rate": 0.0002141346153846154,
+      "loss": 0.1603,
+      "step": 6449
+    },
+    {
+      "epoch": 59.17,
+      "learning_rate": 0.00021408653846153848,
+      "loss": 0.2211,
+      "step": 6450
+    },
+    {
+      "epoch": 59.18,
+      "learning_rate": 0.00021403846153846156,
+      "loss": 0.2356,
+      "step": 6451
+    },
+    {
+      "epoch": 59.19,
+      "learning_rate": 0.0002139903846153846,
+      "loss": 0.1994,
+      "step": 6452
+    },
+    {
+      "epoch": 59.2,
+      "learning_rate": 0.00021394230769230767,
+      "loss": 0.3502,
+      "step": 6453
+    },
+    {
+      "epoch": 59.21,
+      "learning_rate": 0.00021389423076923077,
+      "loss": 0.2728,
+      "step": 6454
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 0.00021384615384615385,
+      "loss": 0.3228,
+      "step": 6455
+    },
+    {
+      "epoch": 59.23,
+      "learning_rate": 0.00021379807692307692,
+      "loss": 0.272,
+      "step": 6456
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 0.00021375,
+      "loss": 0.2558,
+      "step": 6457
+    },
+    {
+      "epoch": 59.25,
+      "learning_rate": 0.00021370192307692306,
+      "loss": 0.1732,
+      "step": 6458
+    },
+    {
+      "epoch": 59.26,
+      "learning_rate": 0.00021365384615384616,
+      "loss": 0.3542,
+      "step": 6459
+    },
+    {
+      "epoch": 59.27,
+      "learning_rate": 0.00021360576923076924,
+      "loss": 0.4072,
+      "step": 6460
+    },
+    {
+      "epoch": 59.28,
+      "learning_rate": 0.0002135576923076923,
+      "loss": 0.3022,
+      "step": 6461
+    },
+    {
+      "epoch": 59.28,
+      "learning_rate": 0.00021350961538461538,
+      "loss": 0.3054,
+      "step": 6462
+    },
+    {
+      "epoch": 59.29,
+      "learning_rate": 0.00021346153846153845,
+      "loss": 0.2693,
+      "step": 6463
+    },
+    {
+      "epoch": 59.3,
+      "learning_rate": 0.00021341346153846155,
+      "loss": 0.354,
+      "step": 6464
+    },
+    {
+      "epoch": 59.31,
+      "learning_rate": 0.00021336538461538463,
+      "loss": 0.2952,
+      "step": 6465
+    },
+    {
+      "epoch": 59.32,
+      "learning_rate": 0.0002133173076923077,
+      "loss": 0.351,
+      "step": 6466
+    },
+    {
+      "epoch": 59.33,
+      "learning_rate": 0.00021326923076923077,
+      "loss": 0.3171,
+      "step": 6467
+    },
+    {
+      "epoch": 59.34,
+      "learning_rate": 0.00021322115384615384,
+      "loss": 0.2411,
+      "step": 6468
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 0.00021317307692307692,
+      "loss": 0.2764,
+      "step": 6469
+    },
+    {
+      "epoch": 59.36,
+      "learning_rate": 0.00021312500000000002,
+      "loss": 0.2473,
+      "step": 6470
+    },
+    {
+      "epoch": 59.37,
+      "learning_rate": 0.0002130769230769231,
+      "loss": 0.2518,
+      "step": 6471
+    },
+    {
+      "epoch": 59.38,
+      "learning_rate": 0.00021302884615384616,
+      "loss": 0.2903,
+      "step": 6472
+    },
+    {
+      "epoch": 59.39,
+      "learning_rate": 0.00021298076923076923,
+      "loss": 0.2986,
+      "step": 6473
+    },
+    {
+      "epoch": 59.39,
+      "learning_rate": 0.0002129326923076923,
+      "loss": 0.3285,
+      "step": 6474
+    },
+    {
+      "epoch": 59.4,
+      "learning_rate": 0.0002128846153846154,
+      "loss": 0.2469,
+      "step": 6475
+    },
+    {
+      "epoch": 59.41,
+      "learning_rate": 0.00021283653846153848,
+      "loss": 0.2704,
+      "step": 6476
+    },
+    {
+      "epoch": 59.42,
+      "learning_rate": 0.00021278846153846155,
+      "loss": 0.2267,
+      "step": 6477
+    },
+    {
+      "epoch": 59.43,
+      "learning_rate": 0.00021274038461538463,
+      "loss": 0.302,
+      "step": 6478
+    },
+    {
+      "epoch": 59.44,
+      "learning_rate": 0.0002126923076923077,
+      "loss": 0.2481,
+      "step": 6479
+    },
+    {
+      "epoch": 59.45,
+      "learning_rate": 0.0002126442307692308,
+      "loss": 0.2556,
+      "step": 6480
+    },
+    {
+      "epoch": 59.46,
+      "learning_rate": 0.00021259615384615384,
+      "loss": 0.1903,
+      "step": 6481
+    },
+    {
+      "epoch": 59.47,
+      "learning_rate": 0.00021254807692307692,
+      "loss": 0.3289,
+      "step": 6482
+    },
+    {
+      "epoch": 59.48,
+      "learning_rate": 0.0002125,
+      "loss": 0.3327,
+      "step": 6483
+    },
+    {
+      "epoch": 59.49,
+      "learning_rate": 0.00021245192307692306,
+      "loss": 0.2841,
+      "step": 6484
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 0.00021240384615384616,
+      "loss": 0.2758,
+      "step": 6485
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 0.00021235576923076923,
+      "loss": 0.4025,
+      "step": 6486
+    },
+    {
+      "epoch": 59.51,
+      "learning_rate": 0.0002123076923076923,
+      "loss": 0.2709,
+      "step": 6487
+    },
+    {
+      "epoch": 59.52,
+      "learning_rate": 0.00021225961538461538,
+      "loss": 0.322,
+      "step": 6488
+    },
+    {
+      "epoch": 59.53,
+      "learning_rate": 0.00021221153846153845,
+      "loss": 0.3627,
+      "step": 6489
+    },
+    {
+      "epoch": 59.54,
+      "learning_rate": 0.00021216346153846155,
+      "loss": 0.2987,
+      "step": 6490
+    },
+    {
+      "epoch": 59.55,
+      "learning_rate": 0.00021211538461538462,
+      "loss": 0.274,
+      "step": 6491
+    },
+    {
+      "epoch": 59.56,
+      "learning_rate": 0.0002120673076923077,
+      "loss": 0.3949,
+      "step": 6492
+    },
+    {
+      "epoch": 59.57,
+      "learning_rate": 0.00021201923076923077,
+      "loss": 0.2744,
+      "step": 6493
+    },
+    {
+      "epoch": 59.58,
+      "learning_rate": 0.00021197115384615384,
+      "loss": 0.3114,
+      "step": 6494
+    },
+    {
+      "epoch": 59.59,
+      "learning_rate": 0.00021192307692307694,
+      "loss": 0.2673,
+      "step": 6495
+    },
+    {
+      "epoch": 59.6,
+      "learning_rate": 0.00021187500000000001,
+      "loss": 0.3037,
+      "step": 6496
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 0.00021182692307692309,
+      "loss": 0.2906,
+      "step": 6497
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 0.00021177884615384616,
+      "loss": 0.2893,
+      "step": 6498
+    },
+    {
+      "epoch": 59.62,
+      "learning_rate": 0.00021173076923076923,
+      "loss": 0.3025,
+      "step": 6499
+    },
+    {
+      "epoch": 59.63,
+      "learning_rate": 0.0002116826923076923,
+      "loss": 0.2467,
+      "step": 6500
+    },
+    {
+      "epoch": 59.63,
+      "eval_cer": 0.15632918034753146,
+      "eval_loss": 0.6656838655471802,
+      "eval_runtime": 88.798,
+      "eval_samples_per_second": 18.548,
+      "eval_steps_per_second": 2.32,
+      "eval_wer": 0.5374323358186089,
+      "step": 6500
+    },
+    {
+      "epoch": 59.64,
+      "learning_rate": 0.0002116346153846154,
+      "loss": 0.3859,
+      "step": 6501
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 0.00021158653846153848,
+      "loss": 0.231,
+      "step": 6502
+    },
+    {
+      "epoch": 59.66,
+      "learning_rate": 0.00021153846153846155,
+      "loss": 0.3573,
+      "step": 6503
+    },
+    {
+      "epoch": 59.67,
+      "learning_rate": 0.00021149038461538462,
+      "loss": 0.2825,
+      "step": 6504
+    },
+    {
+      "epoch": 59.68,
+      "learning_rate": 0.0002114423076923077,
+      "loss": 0.2362,
+      "step": 6505
+    },
+    {
+      "epoch": 59.69,
+      "learning_rate": 0.0002113942307692308,
+      "loss": 0.2616,
+      "step": 6506
+    },
+    {
+      "epoch": 59.7,
+      "learning_rate": 0.00021134615384615387,
+      "loss": 0.3015,
+      "step": 6507
+    },
+    {
+      "epoch": 59.71,
+      "learning_rate": 0.00021129807692307694,
+      "loss": 0.2927,
+      "step": 6508
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 0.00021124999999999998,
+      "loss": 0.3021,
+      "step": 6509
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 0.00021120192307692306,
+      "loss": 0.3698,
+      "step": 6510
+    },
+    {
+      "epoch": 59.73,
+      "learning_rate": 0.00021115384615384616,
+      "loss": 0.2984,
+      "step": 6511
+    },
+    {
+      "epoch": 59.74,
+      "learning_rate": 0.00021110576923076923,
+      "loss": 0.1508,
+      "step": 6512
+    },
+    {
+      "epoch": 59.75,
+      "learning_rate": 0.0002110576923076923,
+      "loss": 0.398,
+      "step": 6513
+    },
+    {
+      "epoch": 59.76,
+      "learning_rate": 0.00021100961538461538,
+      "loss": 0.3579,
+      "step": 6514
+    },
+    {
+      "epoch": 59.77,
+      "learning_rate": 0.00021096153846153845,
+      "loss": 0.3479,
+      "step": 6515
+    },
+    {
+      "epoch": 59.78,
+      "learning_rate": 0.00021091346153846155,
+      "loss": 0.3379,
+      "step": 6516
+    },
+    {
+      "epoch": 59.79,
+      "learning_rate": 0.00021086538461538462,
+      "loss": 0.3011,
+      "step": 6517
+    },
+    {
+      "epoch": 59.8,
+      "learning_rate": 0.0002108173076923077,
+      "loss": 0.3317,
+      "step": 6518
+    },
+    {
+      "epoch": 59.81,
+      "learning_rate": 0.00021076923076923077,
+      "loss": 0.2728,
+      "step": 6519
+    },
+    {
+      "epoch": 59.82,
+      "learning_rate": 0.00021072115384615384,
+      "loss": 0.2521,
+      "step": 6520
+    },
+    {
+      "epoch": 59.83,
+      "learning_rate": 0.00021067307692307694,
+      "loss": 0.3455,
+      "step": 6521
+    },
+    {
+      "epoch": 59.83,
+      "learning_rate": 0.000210625,
+      "loss": 0.3047,
+      "step": 6522
+    },
+    {
+      "epoch": 59.84,
+      "learning_rate": 0.00021057692307692308,
+      "loss": 0.3397,
+      "step": 6523
+    },
+    {
+      "epoch": 59.85,
+      "learning_rate": 0.00021052884615384616,
+      "loss": 0.3067,
+      "step": 6524
+    },
+    {
+      "epoch": 59.86,
+      "learning_rate": 0.00021048076923076923,
+      "loss": 0.2896,
+      "step": 6525
+    },
+    {
+      "epoch": 59.87,
+      "learning_rate": 0.00021043269230769233,
+      "loss": 0.2832,
+      "step": 6526
+    },
+    {
+      "epoch": 59.88,
+      "learning_rate": 0.0002103846153846154,
+      "loss": 0.1818,
+      "step": 6527
+    },
+    {
+      "epoch": 59.89,
+      "learning_rate": 0.00021033653846153847,
+      "loss": 0.2513,
+      "step": 6528
+    },
+    {
+      "epoch": 59.9,
+      "learning_rate": 0.00021028846153846155,
+      "loss": 0.2472,
+      "step": 6529
+    },
+    {
+      "epoch": 59.91,
+      "learning_rate": 0.00021024038461538462,
+      "loss": 0.36,
+      "step": 6530
+    },
+    {
+      "epoch": 59.92,
+      "learning_rate": 0.00021019230769230772,
+      "loss": 0.3625,
+      "step": 6531
+    },
+    {
+      "epoch": 59.93,
+      "learning_rate": 0.0002101442307692308,
+      "loss": 0.2997,
+      "step": 6532
+    },
+    {
+      "epoch": 59.94,
+      "learning_rate": 0.00021009615384615386,
+      "loss": 0.2765,
+      "step": 6533
+    },
+    {
+      "epoch": 59.94,
+      "learning_rate": 0.00021004807692307694,
+      "loss": 0.3132,
+      "step": 6534
+    },
+    {
+      "epoch": 59.95,
+      "learning_rate": 0.00021,
+      "loss": 0.2559,
+      "step": 6535
+    },
+    {
+      "epoch": 59.96,
+      "learning_rate": 0.00020995192307692308,
+      "loss": 0.207,
+      "step": 6536
+    },
+    {
+      "epoch": 59.97,
+      "learning_rate": 0.00020990384615384615,
+      "loss": 0.2154,
+      "step": 6537
+    },
+    {
+      "epoch": 59.98,
+      "learning_rate": 0.00020985576923076923,
+      "loss": 0.2767,
+      "step": 6538
+    },
+    {
+      "epoch": 59.99,
+      "learning_rate": 0.0002098076923076923,
+      "loss": 0.2212,
+      "step": 6539
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 0.00020975961538461537,
+      "loss": 0.3383,
+      "step": 6540
+    },
+    {
+      "epoch": 60.01,
+      "learning_rate": 0.00020971153846153844,
+      "loss": 0.333,
+      "step": 6541
+    },
+    {
+      "epoch": 60.02,
+      "learning_rate": 0.00020966346153846154,
+      "loss": 0.3896,
+      "step": 6542
+    },
+    {
+      "epoch": 60.03,
+      "learning_rate": 0.00020961538461538462,
+      "loss": 0.2735,
+      "step": 6543
+    },
+    {
+      "epoch": 60.04,
+      "learning_rate": 0.0002095673076923077,
+      "loss": 0.2322,
+      "step": 6544
+    },
+    {
+      "epoch": 60.05,
+      "learning_rate": 0.00020951923076923076,
+      "loss": 0.3251,
+      "step": 6545
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 0.00020947115384615383,
+      "loss": 0.3204,
+      "step": 6546
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 0.00020942307692307693,
+      "loss": 0.2922,
+      "step": 6547
+    },
+    {
+      "epoch": 60.07,
+      "learning_rate": 0.000209375,
+      "loss": 0.3091,
+      "step": 6548
+    },
+    {
+      "epoch": 60.08,
+      "learning_rate": 0.00020932692307692308,
+      "loss": 0.2738,
+      "step": 6549
+    },
+    {
+      "epoch": 60.09,
+      "learning_rate": 0.00020927884615384615,
+      "loss": 0.2403,
+      "step": 6550
+    },
+    {
+      "epoch": 60.1,
+      "learning_rate": 0.00020923076923076922,
+      "loss": 0.2331,
+      "step": 6551
+    },
+    {
+      "epoch": 60.11,
+      "learning_rate": 0.00020918269230769232,
+      "loss": 0.2441,
+      "step": 6552
+    },
+    {
+      "epoch": 60.12,
+      "learning_rate": 0.0002091346153846154,
+      "loss": 0.2209,
+      "step": 6553
+    },
+    {
+      "epoch": 60.13,
+      "learning_rate": 0.00020908653846153847,
+      "loss": 0.3573,
+      "step": 6554
+    },
+    {
+      "epoch": 60.14,
+      "learning_rate": 0.00020903846153846154,
+      "loss": 0.2468,
+      "step": 6555
+    },
+    {
+      "epoch": 60.15,
+      "learning_rate": 0.00020899038461538462,
+      "loss": 0.1947,
+      "step": 6556
+    },
+    {
+      "epoch": 60.16,
+      "learning_rate": 0.00020894230769230771,
+      "loss": 0.2097,
+      "step": 6557
+    },
+    {
+      "epoch": 60.17,
+      "learning_rate": 0.0002088942307692308,
+      "loss": 0.3332,
+      "step": 6558
+    },
+    {
+      "epoch": 60.17,
+      "learning_rate": 0.00020884615384615386,
+      "loss": 0.2853,
+      "step": 6559
+    },
+    {
+      "epoch": 60.18,
+      "learning_rate": 0.00020879807692307693,
+      "loss": 0.2507,
+      "step": 6560
+    },
+    {
+      "epoch": 60.19,
+      "learning_rate": 0.00020875,
+      "loss": 0.2701,
+      "step": 6561
+    },
+    {
+      "epoch": 60.2,
+      "learning_rate": 0.0002087019230769231,
+      "loss": 0.1951,
+      "step": 6562
+    },
+    {
+      "epoch": 60.21,
+      "learning_rate": 0.00020865384615384618,
+      "loss": 0.2998,
+      "step": 6563
+    },
+    {
+      "epoch": 60.22,
+      "learning_rate": 0.00020860576923076925,
+      "loss": 0.2907,
+      "step": 6564
+    },
+    {
+      "epoch": 60.23,
+      "learning_rate": 0.0002085576923076923,
+      "loss": 0.2182,
+      "step": 6565
+    },
+    {
+      "epoch": 60.24,
+      "learning_rate": 0.00020850961538461537,
+      "loss": 0.2561,
+      "step": 6566
+    },
+    {
+      "epoch": 60.25,
+      "learning_rate": 0.00020846153846153844,
+      "loss": 0.2662,
+      "step": 6567
+    },
+    {
+      "epoch": 60.26,
+      "learning_rate": 0.00020841346153846154,
+      "loss": 0.4076,
+      "step": 6568
+    },
+    {
+      "epoch": 60.27,
+      "learning_rate": 0.0002083653846153846,
+      "loss": 0.3094,
+      "step": 6569
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 0.00020831730769230769,
+      "loss": 0.3396,
+      "step": 6570
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 0.00020826923076923076,
+      "loss": 0.3105,
+      "step": 6571
+    },
+    {
+      "epoch": 60.29,
+      "learning_rate": 0.00020822115384615383,
+      "loss": 0.2609,
+      "step": 6572
+    },
+    {
+      "epoch": 60.3,
+      "learning_rate": 0.00020817307692307693,
+      "loss": 0.2341,
+      "step": 6573
+    },
+    {
+      "epoch": 60.31,
+      "learning_rate": 0.000208125,
+      "loss": 0.304,
+      "step": 6574
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 0.00020807692307692308,
+      "loss": 0.2375,
+      "step": 6575
+    },
+    {
+      "epoch": 60.33,
+      "learning_rate": 0.00020802884615384615,
+      "loss": 0.2371,
+      "step": 6576
+    },
+    {
+      "epoch": 60.34,
+      "learning_rate": 0.00020798076923076922,
+      "loss": 0.2844,
+      "step": 6577
+    },
+    {
+      "epoch": 60.35,
+      "learning_rate": 0.00020793269230769232,
+      "loss": 0.4385,
+      "step": 6578
+    },
+    {
+      "epoch": 60.36,
+      "learning_rate": 0.0002078846153846154,
+      "loss": 0.221,
+      "step": 6579
+    },
+    {
+      "epoch": 60.37,
+      "learning_rate": 0.00020783653846153847,
+      "loss": 0.2605,
+      "step": 6580
+    },
+    {
+      "epoch": 60.38,
+      "learning_rate": 0.00020778846153846154,
+      "loss": 0.2611,
+      "step": 6581
+    },
+    {
+      "epoch": 60.39,
+      "learning_rate": 0.0002077403846153846,
+      "loss": 0.3336,
+      "step": 6582
+    },
+    {
+      "epoch": 60.39,
+      "learning_rate": 0.0002076923076923077,
+      "loss": 0.3497,
+      "step": 6583
+    },
+    {
+      "epoch": 60.4,
+      "learning_rate": 0.00020764423076923078,
+      "loss": 0.2758,
+      "step": 6584
+    },
+    {
+      "epoch": 60.41,
+      "learning_rate": 0.00020759615384615386,
+      "loss": 0.3565,
+      "step": 6585
+    },
+    {
+      "epoch": 60.42,
+      "learning_rate": 0.00020754807692307693,
+      "loss": 0.2864,
+      "step": 6586
+    },
+    {
+      "epoch": 60.43,
+      "learning_rate": 0.0002075,
+      "loss": 0.2879,
+      "step": 6587
+    },
+    {
+      "epoch": 60.44,
+      "learning_rate": 0.0002074519230769231,
+      "loss": 0.2576,
+      "step": 6588
+    },
+    {
+      "epoch": 60.45,
+      "learning_rate": 0.00020740384615384617,
+      "loss": 0.3331,
+      "step": 6589
+    },
+    {
+      "epoch": 60.46,
+      "learning_rate": 0.00020735576923076925,
+      "loss": 0.327,
+      "step": 6590
+    },
+    {
+      "epoch": 60.47,
+      "learning_rate": 0.00020730769230769232,
+      "loss": 0.2721,
+      "step": 6591
+    },
+    {
+      "epoch": 60.48,
+      "learning_rate": 0.0002072596153846154,
+      "loss": 0.3235,
+      "step": 6592
+    },
+    {
+      "epoch": 60.49,
+      "learning_rate": 0.0002072115384615385,
+      "loss": 0.2378,
+      "step": 6593
+    },
+    {
+      "epoch": 60.5,
+      "learning_rate": 0.00020716346153846154,
+      "loss": 0.2217,
+      "step": 6594
+    },
+    {
+      "epoch": 60.5,
+      "learning_rate": 0.0002071153846153846,
+      "loss": 0.3758,
+      "step": 6595
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 0.00020706730769230768,
+      "loss": 0.3205,
+      "step": 6596
+    },
+    {
+      "epoch": 60.52,
+      "learning_rate": 0.00020701923076923076,
+      "loss": 0.344,
+      "step": 6597
+    },
+    {
+      "epoch": 60.53,
+      "learning_rate": 0.00020697115384615383,
+      "loss": 0.356,
+      "step": 6598
+    },
+    {
+      "epoch": 60.54,
+      "learning_rate": 0.00020692307692307693,
+      "loss": 0.2742,
+      "step": 6599
+    },
+    {
+      "epoch": 60.55,
+      "learning_rate": 0.000206875,
+      "loss": 0.2921,
+      "step": 6600
+    },
+    {
+      "epoch": 60.56,
+      "learning_rate": 0.00020682692307692307,
+      "loss": 0.3124,
+      "step": 6601
+    },
+    {
+      "epoch": 60.57,
+      "learning_rate": 0.00020677884615384615,
+      "loss": 0.2198,
+      "step": 6602
+    },
+    {
+      "epoch": 60.58,
+      "learning_rate": 0.00020673076923076922,
+      "loss": 0.2565,
+      "step": 6603
+    },
+    {
+      "epoch": 60.59,
+      "learning_rate": 0.00020668269230769232,
+      "loss": 0.2482,
+      "step": 6604
+    },
+    {
+      "epoch": 60.6,
+      "learning_rate": 0.0002066346153846154,
+      "loss": 0.2475,
+      "step": 6605
+    },
+    {
+      "epoch": 60.61,
+      "learning_rate": 0.00020658653846153846,
+      "loss": 0.1855,
+      "step": 6606
+    },
+    {
+      "epoch": 60.61,
+      "learning_rate": 0.00020653846153846154,
+      "loss": 0.2694,
+      "step": 6607
+    },
+    {
+      "epoch": 60.62,
+      "learning_rate": 0.0002064903846153846,
+      "loss": 0.3353,
+      "step": 6608
+    },
+    {
+      "epoch": 60.63,
+      "learning_rate": 0.0002064423076923077,
+      "loss": 0.2921,
+      "step": 6609
+    },
+    {
+      "epoch": 60.64,
+      "learning_rate": 0.00020639423076923078,
+      "loss": 0.2159,
+      "step": 6610
+    },
+    {
+      "epoch": 60.65,
+      "learning_rate": 0.00020634615384615385,
+      "loss": 0.2871,
+      "step": 6611
+    },
+    {
+      "epoch": 60.66,
+      "learning_rate": 0.00020629807692307693,
+      "loss": 0.3447,
+      "step": 6612
+    },
+    {
+      "epoch": 60.67,
+      "learning_rate": 0.00020625,
+      "loss": 0.2526,
+      "step": 6613
+    },
+    {
+      "epoch": 60.68,
+      "learning_rate": 0.0002062019230769231,
+      "loss": 0.3029,
+      "step": 6614
+    },
+    {
+      "epoch": 60.69,
+      "learning_rate": 0.00020615384615384617,
+      "loss": 0.2546,
+      "step": 6615
+    },
+    {
+      "epoch": 60.7,
+      "learning_rate": 0.00020610576923076924,
+      "loss": 0.1817,
+      "step": 6616
+    },
+    {
+      "epoch": 60.71,
+      "learning_rate": 0.00020605769230769232,
+      "loss": 0.3545,
+      "step": 6617
+    },
+    {
+      "epoch": 60.72,
+      "learning_rate": 0.0002060096153846154,
+      "loss": 0.2698,
+      "step": 6618
+    },
+    {
+      "epoch": 60.72,
+      "learning_rate": 0.0002059615384615385,
+      "loss": 0.3094,
+      "step": 6619
+    },
+    {
+      "epoch": 60.73,
+      "learning_rate": 0.00020591346153846156,
+      "loss": 0.269,
+      "step": 6620
+    },
+    {
+      "epoch": 60.74,
+      "learning_rate": 0.00020586538461538463,
+      "loss": 0.1807,
+      "step": 6621
+    },
+    {
+      "epoch": 60.75,
+      "learning_rate": 0.00020581730769230768,
+      "loss": 0.4032,
+      "step": 6622
+    },
+    {
+      "epoch": 60.76,
+      "learning_rate": 0.00020576923076923075,
+      "loss": 0.3276,
+      "step": 6623
+    },
+    {
+      "epoch": 60.77,
+      "learning_rate": 0.00020572115384615385,
+      "loss": 0.3535,
+      "step": 6624
+    },
+    {
+      "epoch": 60.78,
+      "learning_rate": 0.00020567307692307692,
+      "loss": 0.2766,
+      "step": 6625
+    },
+    {
+      "epoch": 60.79,
+      "learning_rate": 0.000205625,
+      "loss": 0.3081,
+      "step": 6626
+    },
+    {
+      "epoch": 60.8,
+      "learning_rate": 0.00020557692307692307,
+      "loss": 0.309,
+      "step": 6627
+    },
+    {
+      "epoch": 60.81,
+      "learning_rate": 0.00020552884615384614,
+      "loss": 0.2738,
+      "step": 6628
+    },
+    {
+      "epoch": 60.82,
+      "learning_rate": 0.00020548076923076924,
+      "loss": 0.3347,
+      "step": 6629
+    },
+    {
+      "epoch": 60.83,
+      "learning_rate": 0.00020543269230769231,
+      "loss": 0.306,
+      "step": 6630
+    },
+    {
+      "epoch": 60.83,
+      "learning_rate": 0.0002053846153846154,
+      "loss": 0.2197,
+      "step": 6631
+    },
+    {
+      "epoch": 60.84,
+      "learning_rate": 0.00020533653846153846,
+      "loss": 0.2277,
+      "step": 6632
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 0.00020528846153846153,
+      "loss": 0.2012,
+      "step": 6633
+    },
+    {
+      "epoch": 60.86,
+      "learning_rate": 0.0002052403846153846,
+      "loss": 0.2803,
+      "step": 6634
+    },
+    {
+      "epoch": 60.87,
+      "learning_rate": 0.0002051923076923077,
+      "loss": 0.2239,
+      "step": 6635
+    },
+    {
+      "epoch": 60.88,
+      "learning_rate": 0.00020514423076923078,
+      "loss": 0.3342,
+      "step": 6636
+    },
+    {
+      "epoch": 60.89,
+      "learning_rate": 0.00020509615384615385,
+      "loss": 0.2958,
+      "step": 6637
+    },
+    {
+      "epoch": 60.9,
+      "learning_rate": 0.00020504807692307692,
+      "loss": 0.3989,
+      "step": 6638
+    },
+    {
+      "epoch": 60.91,
+      "learning_rate": 0.000205,
+      "loss": 0.3951,
+      "step": 6639
+    },
+    {
+      "epoch": 60.92,
+      "learning_rate": 0.0002049519230769231,
+      "loss": 0.3281,
+      "step": 6640
+    },
+    {
+      "epoch": 60.93,
+      "learning_rate": 0.00020490384615384617,
+      "loss": 0.2801,
+      "step": 6641
+    },
+    {
+      "epoch": 60.94,
+      "learning_rate": 0.00020485576923076924,
+      "loss": 0.3021,
+      "step": 6642
+    },
+    {
+      "epoch": 60.94,
+      "learning_rate": 0.0002048076923076923,
+      "loss": 0.2391,
+      "step": 6643
+    },
+    {
+      "epoch": 60.95,
+      "learning_rate": 0.00020475961538461539,
+      "loss": 0.1955,
+      "step": 6644
+    },
+    {
+      "epoch": 60.96,
+      "learning_rate": 0.00020471153846153849,
+      "loss": 0.3721,
+      "step": 6645
+    },
+    {
+      "epoch": 60.97,
+      "learning_rate": 0.00020466346153846156,
+      "loss": 0.3596,
+      "step": 6646
+    },
+    {
+      "epoch": 60.98,
+      "learning_rate": 0.00020461538461538463,
+      "loss": 0.209,
+      "step": 6647
+    },
+    {
+      "epoch": 60.99,
+      "learning_rate": 0.0002045673076923077,
+      "loss": 0.2204,
+      "step": 6648
+    },
+    {
+      "epoch": 61.0,
+      "learning_rate": 0.00020451923076923078,
+      "loss": 0.277,
+      "step": 6649
+    },
+    {
+      "epoch": 61.01,
+      "learning_rate": 0.00020447115384615388,
+      "loss": 0.2893,
+      "step": 6650
+    },
+    {
+      "epoch": 61.02,
+      "learning_rate": 0.00020442307692307692,
+      "loss": 0.3013,
+      "step": 6651
+    },
+    {
+      "epoch": 61.03,
+      "learning_rate": 0.000204375,
+      "loss": 0.298,
+      "step": 6652
+    },
+    {
+      "epoch": 61.04,
+      "learning_rate": 0.00020432692307692307,
+      "loss": 0.2997,
+      "step": 6653
+    },
+    {
+      "epoch": 61.05,
+      "learning_rate": 0.00020427884615384614,
+      "loss": 0.2744,
+      "step": 6654
+    },
+    {
+      "epoch": 61.06,
+      "learning_rate": 0.00020423076923076924,
+      "loss": 0.3168,
+      "step": 6655
+    },
+    {
+      "epoch": 61.06,
+      "learning_rate": 0.0002041826923076923,
+      "loss": 0.2644,
+      "step": 6656
+    },
+    {
+      "epoch": 61.07,
+      "learning_rate": 0.00020413461538461538,
+      "loss": 0.3172,
+      "step": 6657
+    },
+    {
+      "epoch": 61.08,
+      "learning_rate": 0.00020408653846153846,
+      "loss": 0.223,
+      "step": 6658
+    },
+    {
+      "epoch": 61.09,
+      "learning_rate": 0.00020403846153846153,
+      "loss": 0.2539,
+      "step": 6659
+    },
+    {
+      "epoch": 61.1,
+      "learning_rate": 0.00020399038461538463,
+      "loss": 0.2094,
+      "step": 6660
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 0.0002039423076923077,
+      "loss": 0.3013,
+      "step": 6661
+    },
+    {
+      "epoch": 61.12,
+      "learning_rate": 0.00020389423076923077,
+      "loss": 0.1945,
+      "step": 6662
+    },
+    {
+      "epoch": 61.13,
+      "learning_rate": 0.00020384615384615385,
+      "loss": 0.2758,
+      "step": 6663
+    },
+    {
+      "epoch": 61.14,
+      "learning_rate": 0.00020379807692307692,
+      "loss": 0.1855,
+      "step": 6664
+    },
+    {
+      "epoch": 61.15,
+      "learning_rate": 0.00020375,
+      "loss": 0.2362,
+      "step": 6665
+    },
+    {
+      "epoch": 61.16,
+      "learning_rate": 0.0002037019230769231,
+      "loss": 0.2466,
+      "step": 6666
+    },
+    {
+      "epoch": 61.17,
+      "learning_rate": 0.00020365384615384616,
+      "loss": 0.2948,
+      "step": 6667
+    },
+    {
+      "epoch": 61.17,
+      "learning_rate": 0.00020360576923076924,
+      "loss": 0.2351,
+      "step": 6668
+    },
+    {
+      "epoch": 61.18,
+      "learning_rate": 0.0002035576923076923,
+      "loss": 0.3965,
+      "step": 6669
+    },
+    {
+      "epoch": 61.19,
+      "learning_rate": 0.00020350961538461538,
+      "loss": 0.2389,
+      "step": 6670
+    },
+    {
+      "epoch": 61.2,
+      "learning_rate": 0.00020346153846153848,
+      "loss": 0.2726,
+      "step": 6671
+    },
+    {
+      "epoch": 61.21,
+      "learning_rate": 0.00020341346153846155,
+      "loss": 0.2069,
+      "step": 6672
+    },
+    {
+      "epoch": 61.22,
+      "learning_rate": 0.00020336538461538463,
+      "loss": 0.2885,
+      "step": 6673
+    },
+    {
+      "epoch": 61.23,
+      "learning_rate": 0.0002033173076923077,
+      "loss": 0.2062,
+      "step": 6674
+    },
+    {
+      "epoch": 61.24,
+      "learning_rate": 0.00020326923076923077,
+      "loss": 0.2673,
+      "step": 6675
+    },
+    {
+      "epoch": 61.25,
+      "learning_rate": 0.00020322115384615387,
+      "loss": 0.1734,
+      "step": 6676
+    },
+    {
+      "epoch": 61.26,
+      "learning_rate": 0.00020317307692307695,
+      "loss": 0.2918,
+      "step": 6677
+    },
+    {
+      "epoch": 61.27,
+      "learning_rate": 0.00020312500000000002,
+      "loss": 0.2632,
+      "step": 6678
+    },
+    {
+      "epoch": 61.28,
+      "learning_rate": 0.00020307692307692306,
+      "loss": 0.3554,
+      "step": 6679
+    },
+    {
+      "epoch": 61.28,
+      "learning_rate": 0.00020302884615384614,
+      "loss": 0.3044,
+      "step": 6680
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 0.00020298076923076924,
+      "loss": 0.2707,
+      "step": 6681
+    },
+    {
+      "epoch": 61.3,
+      "learning_rate": 0.0002029326923076923,
+      "loss": 0.3136,
+      "step": 6682
+    },
+    {
+      "epoch": 61.31,
+      "learning_rate": 0.00020288461538461538,
+      "loss": 0.329,
+      "step": 6683
+    },
+    {
+      "epoch": 61.32,
+      "learning_rate": 0.00020283653846153845,
+      "loss": 0.2586,
+      "step": 6684
+    },
+    {
+      "epoch": 61.33,
+      "learning_rate": 0.00020278846153846153,
+      "loss": 0.2866,
+      "step": 6685
+    },
+    {
+      "epoch": 61.34,
+      "learning_rate": 0.00020274038461538463,
+      "loss": 0.314,
+      "step": 6686
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 0.0002026923076923077,
+      "loss": 0.232,
+      "step": 6687
+    },
+    {
+      "epoch": 61.36,
+      "learning_rate": 0.00020264423076923077,
+      "loss": 0.2937,
+      "step": 6688
+    },
+    {
+      "epoch": 61.37,
+      "learning_rate": 0.00020259615384615384,
+      "loss": 0.2868,
+      "step": 6689
+    },
+    {
+      "epoch": 61.38,
+      "learning_rate": 0.00020254807692307692,
+      "loss": 0.1716,
+      "step": 6690
+    },
+    {
+      "epoch": 61.39,
+      "learning_rate": 0.00020250000000000002,
+      "loss": 0.2225,
+      "step": 6691
+    },
+    {
+      "epoch": 61.39,
+      "learning_rate": 0.0002024519230769231,
+      "loss": 0.2275,
+      "step": 6692
+    },
+    {
+      "epoch": 61.4,
+      "learning_rate": 0.00020240384615384616,
+      "loss": 0.2988,
+      "step": 6693
+    },
+    {
+      "epoch": 61.41,
+      "learning_rate": 0.00020235576923076923,
+      "loss": 0.3608,
+      "step": 6694
+    },
+    {
+      "epoch": 61.42,
+      "learning_rate": 0.0002023076923076923,
+      "loss": 0.3024,
+      "step": 6695
+    },
+    {
+      "epoch": 61.43,
+      "learning_rate": 0.00020225961538461538,
+      "loss": 0.202,
+      "step": 6696
+    },
+    {
+      "epoch": 61.44,
+      "learning_rate": 0.00020221153846153848,
+      "loss": 0.1659,
+      "step": 6697
+    },
+    {
+      "epoch": 61.45,
+      "learning_rate": 0.00020216346153846155,
+      "loss": 0.2502,
+      "step": 6698
+    },
+    {
+      "epoch": 61.46,
+      "learning_rate": 0.00020211538461538462,
+      "loss": 0.2382,
+      "step": 6699
+    },
+    {
+      "epoch": 61.47,
+      "learning_rate": 0.0002020673076923077,
+      "loss": 0.2965,
+      "step": 6700
+    },
+    {
+      "epoch": 61.48,
+      "learning_rate": 0.00020201923076923077,
+      "loss": 0.299,
+      "step": 6701
+    },
+    {
+      "epoch": 61.49,
+      "learning_rate": 0.00020197115384615387,
+      "loss": 0.2734,
+      "step": 6702
+    },
+    {
+      "epoch": 61.5,
+      "learning_rate": 0.00020192307692307694,
+      "loss": 0.1326,
+      "step": 6703
+    },
+    {
+      "epoch": 61.5,
+      "learning_rate": 0.00020187500000000001,
+      "loss": 0.4088,
+      "step": 6704
+    },
+    {
+      "epoch": 61.51,
+      "learning_rate": 0.0002018269230769231,
+      "loss": 0.3575,
+      "step": 6705
+    },
+    {
+      "epoch": 61.52,
+      "learning_rate": 0.00020177884615384613,
+      "loss": 0.3698,
+      "step": 6706
+    },
+    {
+      "epoch": 61.53,
+      "learning_rate": 0.00020173076923076923,
+      "loss": 0.3438,
+      "step": 6707
+    },
+    {
+      "epoch": 61.54,
+      "learning_rate": 0.0002016826923076923,
+      "loss": 0.3168,
+      "step": 6708
+    },
+    {
+      "epoch": 61.55,
+      "learning_rate": 0.00020163461538461538,
+      "loss": 0.3273,
+      "step": 6709
+    },
+    {
+      "epoch": 61.56,
+      "learning_rate": 0.00020158653846153845,
+      "loss": 0.1909,
+      "step": 6710
+    },
+    {
+      "epoch": 61.57,
+      "learning_rate": 0.00020153846153846152,
+      "loss": 0.2242,
+      "step": 6711
+    },
+    {
+      "epoch": 61.58,
+      "learning_rate": 0.00020149038461538462,
+      "loss": 0.1949,
+      "step": 6712
+    },
+    {
+      "epoch": 61.59,
+      "learning_rate": 0.0002014423076923077,
+      "loss": 0.2747,
+      "step": 6713
+    },
+    {
+      "epoch": 61.6,
+      "learning_rate": 0.00020139423076923077,
+      "loss": 0.3065,
+      "step": 6714
+    },
+    {
+      "epoch": 61.61,
+      "learning_rate": 0.00020134615384615384,
+      "loss": 0.2327,
+      "step": 6715
+    },
+    {
+      "epoch": 61.61,
+      "learning_rate": 0.0002012980769230769,
+      "loss": 0.3018,
+      "step": 6716
+    },
+    {
+      "epoch": 61.62,
+      "learning_rate": 0.00020125,
+      "loss": 0.3319,
+      "step": 6717
+    },
+    {
+      "epoch": 61.63,
+      "learning_rate": 0.00020120192307692309,
+      "loss": 0.1995,
+      "step": 6718
+    },
+    {
+      "epoch": 61.64,
+      "learning_rate": 0.00020115384615384616,
+      "loss": 0.3088,
+      "step": 6719
+    },
+    {
+      "epoch": 61.65,
+      "learning_rate": 0.00020110576923076923,
+      "loss": 0.244,
+      "step": 6720
+    },
+    {
+      "epoch": 61.66,
+      "learning_rate": 0.0002010576923076923,
+      "loss": 0.2441,
+      "step": 6721
+    },
+    {
+      "epoch": 61.67,
+      "learning_rate": 0.0002010096153846154,
+      "loss": 0.2595,
+      "step": 6722
+    },
+    {
+      "epoch": 61.68,
+      "learning_rate": 0.00020096153846153848,
+      "loss": 0.2703,
+      "step": 6723
+    },
+    {
+      "epoch": 61.69,
+      "learning_rate": 0.00020091346153846155,
+      "loss": 0.2174,
+      "step": 6724
+    },
+    {
+      "epoch": 61.7,
+      "learning_rate": 0.00020086538461538462,
+      "loss": 0.2394,
+      "step": 6725
+    },
+    {
+      "epoch": 61.71,
+      "learning_rate": 0.0002008173076923077,
+      "loss": 0.4072,
+      "step": 6726
+    },
+    {
+      "epoch": 61.72,
+      "learning_rate": 0.00020076923076923077,
+      "loss": 0.2846,
+      "step": 6727
+    },
+    {
+      "epoch": 61.72,
+      "learning_rate": 0.00020072115384615387,
+      "loss": 0.2389,
+      "step": 6728
+    },
+    {
+      "epoch": 61.73,
+      "learning_rate": 0.00020067307692307694,
+      "loss": 0.2762,
+      "step": 6729
+    },
+    {
+      "epoch": 61.74,
+      "learning_rate": 0.000200625,
+      "loss": 0.18,
+      "step": 6730
+    },
+    {
+      "epoch": 61.75,
+      "learning_rate": 0.00020057692307692308,
+      "loss": 0.4099,
+      "step": 6731
+    },
+    {
+      "epoch": 61.76,
+      "learning_rate": 0.00020052884615384616,
+      "loss": 0.3185,
+      "step": 6732
+    },
+    {
+      "epoch": 61.77,
+      "learning_rate": 0.00020048076923076926,
+      "loss": 0.2828,
+      "step": 6733
+    },
+    {
+      "epoch": 61.78,
+      "learning_rate": 0.00020043269230769233,
+      "loss": 0.3318,
+      "step": 6734
+    },
+    {
+      "epoch": 61.79,
+      "learning_rate": 0.00020038461538461537,
+      "loss": 0.2622,
+      "step": 6735
+    },
+    {
+      "epoch": 61.8,
+      "learning_rate": 0.00020033653846153845,
+      "loss": 0.3526,
+      "step": 6736
+    },
+    {
+      "epoch": 61.81,
+      "learning_rate": 0.00020028846153846152,
+      "loss": 0.3267,
+      "step": 6737
+    },
+    {
+      "epoch": 61.82,
+      "learning_rate": 0.00020024038461538462,
+      "loss": 0.2933,
+      "step": 6738
+    },
+    {
+      "epoch": 61.83,
+      "learning_rate": 0.0002001923076923077,
+      "loss": 0.3497,
+      "step": 6739
+    },
+    {
+      "epoch": 61.83,
+      "learning_rate": 0.00020014423076923076,
+      "loss": 0.2494,
+      "step": 6740
+    },
+    {
+      "epoch": 61.84,
+      "learning_rate": 0.00020009615384615384,
+      "loss": 0.2733,
+      "step": 6741
+    },
+    {
+      "epoch": 61.85,
+      "learning_rate": 0.0002000480769230769,
+      "loss": 0.2581,
+      "step": 6742
+    },
+    {
+      "epoch": 61.86,
+      "learning_rate": 0.0002,
+      "loss": 0.185,
+      "step": 6743
+    },
+    {
+      "epoch": 61.87,
+      "learning_rate": 0.00019995192307692308,
+      "loss": 0.3522,
+      "step": 6744
+    },
+    {
+      "epoch": 61.88,
+      "learning_rate": 0.00019990384615384615,
+      "loss": 0.2604,
+      "step": 6745
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 0.00019985576923076923,
+      "loss": 0.2958,
+      "step": 6746
+    },
+    {
+      "epoch": 61.9,
+      "learning_rate": 0.0001998076923076923,
+      "loss": 0.2415,
+      "step": 6747
+    },
+    {
+      "epoch": 61.91,
+      "learning_rate": 0.0001997596153846154,
+      "loss": 0.311,
+      "step": 6748
+    },
+    {
+      "epoch": 61.92,
+      "learning_rate": 0.00019971153846153847,
+      "loss": 0.3056,
+      "step": 6749
+    },
+    {
+      "epoch": 61.93,
+      "learning_rate": 0.00019966346153846155,
+      "loss": 0.2333,
+      "step": 6750
+    },
+    {
+      "epoch": 61.94,
+      "learning_rate": 0.00019961538461538462,
+      "loss": 0.2297,
+      "step": 6751
+    },
+    {
+      "epoch": 61.94,
+      "learning_rate": 0.0001995673076923077,
+      "loss": 0.1936,
+      "step": 6752
+    },
+    {
+      "epoch": 61.95,
+      "learning_rate": 0.0001995192307692308,
+      "loss": 0.2954,
+      "step": 6753
+    },
+    {
+      "epoch": 61.96,
+      "learning_rate": 0.00019947115384615386,
+      "loss": 0.243,
+      "step": 6754
+    },
+    {
+      "epoch": 61.97,
+      "learning_rate": 0.00019942307692307694,
+      "loss": 0.2435,
+      "step": 6755
+    },
+    {
+      "epoch": 61.98,
+      "learning_rate": 0.000199375,
+      "loss": 0.3118,
+      "step": 6756
+    },
+    {
+      "epoch": 61.99,
+      "learning_rate": 0.00019932692307692308,
+      "loss": 0.1288,
+      "step": 6757
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 0.00019927884615384618,
+      "loss": 0.2862,
+      "step": 6758
+    },
+    {
+      "epoch": 62.01,
+      "learning_rate": 0.00019923076923076925,
+      "loss": 0.3639,
+      "step": 6759
+    },
+    {
+      "epoch": 62.02,
+      "learning_rate": 0.00019918269230769233,
+      "loss": 0.3221,
+      "step": 6760
+    },
+    {
+      "epoch": 62.03,
+      "learning_rate": 0.0001991346153846154,
+      "loss": 0.3386,
+      "step": 6761
+    },
+    {
+      "epoch": 62.04,
+      "learning_rate": 0.00019908653846153847,
+      "loss": 0.2479,
+      "step": 6762
+    },
+    {
+      "epoch": 62.05,
+      "learning_rate": 0.00019903846153846152,
+      "loss": 0.2469,
+      "step": 6763
+    },
+    {
+      "epoch": 62.06,
+      "learning_rate": 0.00019899038461538462,
+      "loss": 0.2822,
+      "step": 6764
+    },
+    {
+      "epoch": 62.06,
+      "learning_rate": 0.0001989423076923077,
+      "loss": 0.1978,
+      "step": 6765
+    },
+    {
+      "epoch": 62.07,
+      "learning_rate": 0.00019889423076923076,
+      "loss": 0.2872,
+      "step": 6766
+    },
+    {
+      "epoch": 62.08,
+      "learning_rate": 0.00019884615384615383,
+      "loss": 0.373,
+      "step": 6767
+    },
+    {
+      "epoch": 62.09,
+      "learning_rate": 0.0001987980769230769,
+      "loss": 0.2085,
+      "step": 6768
+    },
+    {
+      "epoch": 62.1,
+      "learning_rate": 0.00019875,
+      "loss": 0.2737,
+      "step": 6769
+    },
+    {
+      "epoch": 62.11,
+      "learning_rate": 0.00019870192307692308,
+      "loss": 0.2759,
+      "step": 6770
+    },
+    {
+      "epoch": 62.12,
+      "learning_rate": 0.00019865384615384615,
+      "loss": 0.2235,
+      "step": 6771
+    },
+    {
+      "epoch": 62.13,
+      "learning_rate": 0.00019860576923076922,
+      "loss": 0.2034,
+      "step": 6772
+    },
+    {
+      "epoch": 62.14,
+      "learning_rate": 0.0001985576923076923,
+      "loss": 0.2979,
+      "step": 6773
+    },
+    {
+      "epoch": 62.15,
+      "learning_rate": 0.0001985096153846154,
+      "loss": 0.2851,
+      "step": 6774
+    },
+    {
+      "epoch": 62.16,
+      "learning_rate": 0.00019846153846153847,
+      "loss": 0.2748,
+      "step": 6775
+    },
+    {
+      "epoch": 62.17,
+      "learning_rate": 0.00019841346153846154,
+      "loss": 0.249,
+      "step": 6776
+    },
+    {
+      "epoch": 62.17,
+      "learning_rate": 0.00019836538461538461,
+      "loss": 0.2579,
+      "step": 6777
+    },
+    {
+      "epoch": 62.18,
+      "learning_rate": 0.0001983173076923077,
+      "loss": 0.2374,
+      "step": 6778
+    },
+    {
+      "epoch": 62.19,
+      "learning_rate": 0.0001982692307692308,
+      "loss": 0.2958,
+      "step": 6779
+    },
+    {
+      "epoch": 62.2,
+      "learning_rate": 0.00019822115384615386,
+      "loss": 0.2957,
+      "step": 6780
+    },
+    {
+      "epoch": 62.21,
+      "learning_rate": 0.00019817307692307693,
+      "loss": 0.2522,
+      "step": 6781
+    },
+    {
+      "epoch": 62.22,
+      "learning_rate": 0.000198125,
+      "loss": 0.1507,
+      "step": 6782
+    },
+    {
+      "epoch": 62.23,
+      "learning_rate": 0.00019807692307692308,
+      "loss": 0.2421,
+      "step": 6783
+    },
+    {
+      "epoch": 62.24,
+      "learning_rate": 0.00019802884615384618,
+      "loss": 0.3292,
+      "step": 6784
+    },
+    {
+      "epoch": 62.25,
+      "learning_rate": 0.00019798076923076925,
+      "loss": 0.2853,
+      "step": 6785
+    },
+    {
+      "epoch": 62.26,
+      "learning_rate": 0.00019793269230769232,
+      "loss": 0.3148,
+      "step": 6786
+    },
+    {
+      "epoch": 62.27,
+      "learning_rate": 0.0001978846153846154,
+      "loss": 0.2089,
+      "step": 6787
+    },
+    {
+      "epoch": 62.28,
+      "learning_rate": 0.00019783653846153847,
+      "loss": 0.2676,
+      "step": 6788
+    },
+    {
+      "epoch": 62.28,
+      "learning_rate": 0.00019778846153846157,
+      "loss": 0.2951,
+      "step": 6789
+    },
+    {
+      "epoch": 62.29,
+      "learning_rate": 0.00019774038461538464,
+      "loss": 0.2487,
+      "step": 6790
+    },
+    {
+      "epoch": 62.3,
+      "learning_rate": 0.0001976923076923077,
+      "loss": 0.3169,
+      "step": 6791
+    },
+    {
+      "epoch": 62.31,
+      "learning_rate": 0.00019764423076923076,
+      "loss": 0.3147,
+      "step": 6792
+    },
+    {
+      "epoch": 62.32,
+      "learning_rate": 0.00019759615384615383,
+      "loss": 0.3251,
+      "step": 6793
+    },
+    {
+      "epoch": 62.33,
+      "learning_rate": 0.0001975480769230769,
+      "loss": 0.2987,
+      "step": 6794
+    },
+    {
+      "epoch": 62.34,
+      "learning_rate": 0.0001975,
+      "loss": 0.2928,
+      "step": 6795
+    },
+    {
+      "epoch": 62.35,
+      "learning_rate": 0.00019745192307692308,
+      "loss": 0.215,
+      "step": 6796
+    },
+    {
+      "epoch": 62.36,
+      "learning_rate": 0.00019740384615384615,
+      "loss": 0.2983,
+      "step": 6797
+    },
+    {
+      "epoch": 62.37,
+      "learning_rate": 0.00019735576923076922,
+      "loss": 0.2517,
+      "step": 6798
+    },
+    {
+      "epoch": 62.38,
+      "learning_rate": 0.0001973076923076923,
+      "loss": 0.3459,
+      "step": 6799
+    },
+    {
+      "epoch": 62.39,
+      "learning_rate": 0.0001972596153846154,
+      "loss": 0.3149,
+      "step": 6800
+    },
+    {
+      "epoch": 62.39,
+      "learning_rate": 0.00019721153846153847,
+      "loss": 0.2777,
+      "step": 6801
+    },
+    {
+      "epoch": 62.4,
+      "learning_rate": 0.00019716346153846154,
+      "loss": 0.1857,
+      "step": 6802
+    },
+    {
+      "epoch": 62.41,
+      "learning_rate": 0.0001971153846153846,
+      "loss": 0.233,
+      "step": 6803
+    },
+    {
+      "epoch": 62.42,
+      "learning_rate": 0.00019706730769230768,
+      "loss": 0.2386,
+      "step": 6804
+    },
+    {
+      "epoch": 62.43,
+      "learning_rate": 0.00019701923076923078,
+      "loss": 0.2436,
+      "step": 6805
+    },
+    {
+      "epoch": 62.44,
+      "learning_rate": 0.00019697115384615386,
+      "loss": 0.2698,
+      "step": 6806
+    },
+    {
+      "epoch": 62.45,
+      "learning_rate": 0.00019692307692307693,
+      "loss": 0.2628,
+      "step": 6807
+    },
+    {
+      "epoch": 62.46,
+      "learning_rate": 0.000196875,
+      "loss": 0.1805,
+      "step": 6808
+    },
+    {
+      "epoch": 62.47,
+      "learning_rate": 0.00019682692307692307,
+      "loss": 0.2009,
+      "step": 6809
+    },
+    {
+      "epoch": 62.48,
+      "learning_rate": 0.00019677884615384617,
+      "loss": 0.2823,
+      "step": 6810
+    },
+    {
+      "epoch": 62.49,
+      "learning_rate": 0.00019673076923076925,
+      "loss": 0.1688,
+      "step": 6811
+    },
+    {
+      "epoch": 62.5,
+      "learning_rate": 0.00019668269230769232,
+      "loss": 0.2787,
+      "step": 6812
+    },
+    {
+      "epoch": 62.5,
+      "learning_rate": 0.0001966346153846154,
+      "loss": 0.4017,
+      "step": 6813
+    },
+    {
+      "epoch": 62.51,
+      "learning_rate": 0.00019658653846153846,
+      "loss": 0.2617,
+      "step": 6814
+    },
+    {
+      "epoch": 62.52,
+      "learning_rate": 0.00019653846153846156,
+      "loss": 0.3128,
+      "step": 6815
+    },
+    {
+      "epoch": 62.53,
+      "learning_rate": 0.00019649038461538464,
+      "loss": 0.2937,
+      "step": 6816
+    },
+    {
+      "epoch": 62.54,
+      "learning_rate": 0.0001964423076923077,
+      "loss": 0.3613,
+      "step": 6817
+    },
+    {
+      "epoch": 62.55,
+      "learning_rate": 0.00019639423076923078,
+      "loss": 0.2474,
+      "step": 6818
+    },
+    {
+      "epoch": 62.56,
+      "learning_rate": 0.00019634615384615385,
+      "loss": 0.2478,
+      "step": 6819
+    },
+    {
+      "epoch": 62.57,
+      "learning_rate": 0.00019629807692307693,
+      "loss": 0.2291,
+      "step": 6820
+    },
+    {
+      "epoch": 62.58,
+      "learning_rate": 0.00019625,
+      "loss": 0.2582,
+      "step": 6821
+    },
+    {
+      "epoch": 62.59,
+      "learning_rate": 0.00019620192307692307,
+      "loss": 0.2705,
+      "step": 6822
+    },
+    {
+      "epoch": 62.6,
+      "learning_rate": 0.00019615384615384615,
+      "loss": 0.2791,
+      "step": 6823
+    },
+    {
+      "epoch": 62.61,
+      "learning_rate": 0.00019610576923076922,
+      "loss": 0.216,
+      "step": 6824
+    },
+    {
+      "epoch": 62.61,
+      "learning_rate": 0.0001960576923076923,
+      "loss": 0.3359,
+      "step": 6825
+    },
+    {
+      "epoch": 62.62,
+      "learning_rate": 0.0001960096153846154,
+      "loss": 0.3118,
+      "step": 6826
+    },
+    {
+      "epoch": 62.63,
+      "learning_rate": 0.00019596153846153846,
+      "loss": 0.3099,
+      "step": 6827
+    },
+    {
+      "epoch": 62.64,
+      "learning_rate": 0.00019591346153846154,
+      "loss": 0.2522,
+      "step": 6828
+    },
+    {
+      "epoch": 62.65,
+      "learning_rate": 0.0001958653846153846,
+      "loss": 0.2585,
+      "step": 6829
+    },
+    {
+      "epoch": 62.66,
+      "learning_rate": 0.00019581730769230768,
+      "loss": 0.2152,
+      "step": 6830
+    },
+    {
+      "epoch": 62.67,
+      "learning_rate": 0.00019576923076923078,
+      "loss": 0.1951,
+      "step": 6831
+    },
+    {
+      "epoch": 62.68,
+      "learning_rate": 0.00019572115384615385,
+      "loss": 0.2873,
+      "step": 6832
+    },
+    {
+      "epoch": 62.69,
+      "learning_rate": 0.00019567307692307693,
+      "loss": 0.2315,
+      "step": 6833
+    },
+    {
+      "epoch": 62.7,
+      "learning_rate": 0.000195625,
+      "loss": 0.2941,
+      "step": 6834
+    },
+    {
+      "epoch": 62.71,
+      "learning_rate": 0.00019557692307692307,
+      "loss": 0.3404,
+      "step": 6835
+    },
+    {
+      "epoch": 62.72,
+      "learning_rate": 0.00019552884615384617,
+      "loss": 0.1777,
+      "step": 6836
+    },
+    {
+      "epoch": 62.72,
+      "learning_rate": 0.00019548076923076924,
+      "loss": 0.2905,
+      "step": 6837
+    },
+    {
+      "epoch": 62.73,
+      "learning_rate": 0.00019543269230769232,
+      "loss": 0.209,
+      "step": 6838
+    },
+    {
+      "epoch": 62.74,
+      "learning_rate": 0.0001953846153846154,
+      "loss": 0.2039,
+      "step": 6839
+    },
+    {
+      "epoch": 62.75,
+      "learning_rate": 0.00019533653846153846,
+      "loss": 0.352,
+      "step": 6840
+    },
+    {
+      "epoch": 62.76,
+      "learning_rate": 0.00019528846153846156,
+      "loss": 0.4394,
+      "step": 6841
+    },
+    {
+      "epoch": 62.77,
+      "learning_rate": 0.00019524038461538463,
+      "loss": 0.2976,
+      "step": 6842
+    },
+    {
+      "epoch": 62.78,
+      "learning_rate": 0.0001951923076923077,
+      "loss": 0.2655,
+      "step": 6843
+    },
+    {
+      "epoch": 62.79,
+      "learning_rate": 0.00019514423076923078,
+      "loss": 0.2355,
+      "step": 6844
+    },
+    {
+      "epoch": 62.8,
+      "learning_rate": 0.00019509615384615385,
+      "loss": 0.2575,
+      "step": 6845
+    },
+    {
+      "epoch": 62.81,
+      "learning_rate": 0.00019504807692307695,
+      "loss": 0.2784,
+      "step": 6846
+    },
+    {
+      "epoch": 62.82,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 0.2569,
+      "step": 6847
+    },
+    {
+      "epoch": 62.83,
+      "learning_rate": 0.00019495192307692307,
+      "loss": 0.2478,
+      "step": 6848
+    },
+    {
+      "epoch": 62.83,
+      "learning_rate": 0.00019490384615384614,
+      "loss": 0.2611,
+      "step": 6849
+    },
+    {
+      "epoch": 62.84,
+      "learning_rate": 0.00019485576923076921,
+      "loss": 0.226,
+      "step": 6850
+    },
+    {
+      "epoch": 62.85,
+      "learning_rate": 0.00019480769230769231,
+      "loss": 0.2379,
+      "step": 6851
+    },
+    {
+      "epoch": 62.86,
+      "learning_rate": 0.0001947596153846154,
+      "loss": 0.2194,
+      "step": 6852
+    },
+    {
+      "epoch": 62.87,
+      "learning_rate": 0.00019471153846153846,
+      "loss": 0.3419,
+      "step": 6853
+    },
+    {
+      "epoch": 62.88,
+      "learning_rate": 0.00019466346153846153,
+      "loss": 0.3775,
+      "step": 6854
+    },
+    {
+      "epoch": 62.89,
+      "learning_rate": 0.0001946153846153846,
+      "loss": 0.3154,
+      "step": 6855
+    },
+    {
+      "epoch": 62.9,
+      "learning_rate": 0.0001945673076923077,
+      "loss": 0.2757,
+      "step": 6856
+    },
+    {
+      "epoch": 62.91,
+      "learning_rate": 0.00019451923076923078,
+      "loss": 0.2214,
+      "step": 6857
+    },
+    {
+      "epoch": 62.92,
+      "learning_rate": 0.00019447115384615385,
+      "loss": 0.2263,
+      "step": 6858
+    },
+    {
+      "epoch": 62.93,
+      "learning_rate": 0.00019442307692307692,
+      "loss": 0.2477,
+      "step": 6859
+    },
+    {
+      "epoch": 62.94,
+      "learning_rate": 0.000194375,
+      "loss": 0.2977,
+      "step": 6860
+    },
+    {
+      "epoch": 62.94,
+      "learning_rate": 0.00019432692307692307,
+      "loss": 0.1514,
+      "step": 6861
+    },
+    {
+      "epoch": 62.95,
+      "learning_rate": 0.00019427884615384617,
+      "loss": 0.1783,
+      "step": 6862
+    },
+    {
+      "epoch": 62.96,
+      "learning_rate": 0.00019423076923076924,
+      "loss": 0.3376,
+      "step": 6863
+    },
+    {
+      "epoch": 62.97,
+      "learning_rate": 0.0001941826923076923,
+      "loss": 0.2147,
+      "step": 6864
+    },
+    {
+      "epoch": 62.98,
+      "learning_rate": 0.00019413461538461539,
+      "loss": 0.3093,
+      "step": 6865
+    },
+    {
+      "epoch": 62.99,
+      "learning_rate": 0.00019408653846153846,
+      "loss": 0.2095,
+      "step": 6866
+    },
+    {
+      "epoch": 63.0,
+      "learning_rate": 0.00019403846153846156,
+      "loss": 0.2412,
+      "step": 6867
+    },
+    {
+      "epoch": 63.01,
+      "learning_rate": 0.00019399038461538463,
+      "loss": 0.327,
+      "step": 6868
+    },
+    {
+      "epoch": 63.02,
+      "learning_rate": 0.0001939423076923077,
+      "loss": 0.2479,
+      "step": 6869
+    },
+    {
+      "epoch": 63.03,
+      "learning_rate": 0.00019389423076923078,
+      "loss": 0.2711,
+      "step": 6870
+    },
+    {
+      "epoch": 63.04,
+      "learning_rate": 0.00019384615384615385,
+      "loss": 0.3019,
+      "step": 6871
+    },
+    {
+      "epoch": 63.05,
+      "learning_rate": 0.00019379807692307695,
+      "loss": 0.2894,
+      "step": 6872
+    },
+    {
+      "epoch": 63.06,
+      "learning_rate": 0.00019375000000000002,
+      "loss": 0.3228,
+      "step": 6873
+    },
+    {
+      "epoch": 63.06,
+      "learning_rate": 0.0001937019230769231,
+      "loss": 0.3391,
+      "step": 6874
+    },
+    {
+      "epoch": 63.07,
+      "learning_rate": 0.00019365384615384617,
+      "loss": 0.2472,
+      "step": 6875
+    },
+    {
+      "epoch": 63.08,
+      "learning_rate": 0.0001936057692307692,
+      "loss": 0.2725,
+      "step": 6876
+    },
+    {
+      "epoch": 63.09,
+      "learning_rate": 0.0001935576923076923,
+      "loss": 0.2313,
+      "step": 6877
+    },
+    {
+      "epoch": 63.1,
+      "learning_rate": 0.00019350961538461538,
+      "loss": 0.2704,
+      "step": 6878
+    },
+    {
+      "epoch": 63.11,
+      "learning_rate": 0.00019346153846153846,
+      "loss": 0.3133,
+      "step": 6879
+    },
+    {
+      "epoch": 63.12,
+      "learning_rate": 0.00019341346153846153,
+      "loss": 0.1714,
+      "step": 6880
+    },
+    {
+      "epoch": 63.13,
+      "learning_rate": 0.0001933653846153846,
+      "loss": 0.2571,
+      "step": 6881
+    },
+    {
+      "epoch": 63.14,
+      "learning_rate": 0.0001933173076923077,
+      "loss": 0.2714,
+      "step": 6882
+    },
+    {
+      "epoch": 63.15,
+      "learning_rate": 0.00019326923076923077,
+      "loss": 0.1969,
+      "step": 6883
+    },
+    {
+      "epoch": 63.16,
+      "learning_rate": 0.00019322115384615385,
+      "loss": 0.1568,
+      "step": 6884
+    },
+    {
+      "epoch": 63.17,
+      "learning_rate": 0.00019317307692307692,
+      "loss": 0.1948,
+      "step": 6885
+    },
+    {
+      "epoch": 63.17,
+      "learning_rate": 0.000193125,
+      "loss": 0.2601,
+      "step": 6886
+    },
+    {
+      "epoch": 63.18,
+      "learning_rate": 0.0001930769230769231,
+      "loss": 0.2982,
+      "step": 6887
+    },
+    {
+      "epoch": 63.19,
+      "learning_rate": 0.00019302884615384616,
+      "loss": 0.1641,
+      "step": 6888
+    },
+    {
+      "epoch": 63.2,
+      "learning_rate": 0.00019298076923076924,
+      "loss": 0.2862,
+      "step": 6889
+    },
+    {
+      "epoch": 63.21,
+      "learning_rate": 0.0001929326923076923,
+      "loss": 0.2074,
+      "step": 6890
+    },
+    {
+      "epoch": 63.22,
+      "learning_rate": 0.00019288461538461538,
+      "loss": 0.1956,
+      "step": 6891
+    },
+    {
+      "epoch": 63.23,
+      "learning_rate": 0.00019283653846153845,
+      "loss": 0.2623,
+      "step": 6892
+    },
+    {
+      "epoch": 63.24,
+      "learning_rate": 0.00019278846153846155,
+      "loss": 0.3083,
+      "step": 6893
+    },
+    {
+      "epoch": 63.25,
+      "learning_rate": 0.00019274038461538463,
+      "loss": 0.1817,
+      "step": 6894
+    },
+    {
+      "epoch": 63.26,
+      "learning_rate": 0.0001926923076923077,
+      "loss": 0.319,
+      "step": 6895
+    },
+    {
+      "epoch": 63.27,
+      "learning_rate": 0.00019264423076923077,
+      "loss": 0.3223,
+      "step": 6896
+    },
+    {
+      "epoch": 63.28,
+      "learning_rate": 0.00019259615384615384,
+      "loss": 0.3266,
+      "step": 6897
+    },
+    {
+      "epoch": 63.28,
+      "learning_rate": 0.00019254807692307694,
+      "loss": 0.2232,
+      "step": 6898
+    },
+    {
+      "epoch": 63.29,
+      "learning_rate": 0.00019250000000000002,
+      "loss": 0.3382,
+      "step": 6899
+    },
+    {
+      "epoch": 63.3,
+      "learning_rate": 0.0001924519230769231,
+      "loss": 0.2518,
+      "step": 6900
+    },
+    {
+      "epoch": 63.31,
+      "learning_rate": 0.00019240384615384616,
+      "loss": 0.2937,
+      "step": 6901
+    },
+    {
+      "epoch": 63.32,
+      "learning_rate": 0.00019235576923076923,
+      "loss": 0.1955,
+      "step": 6902
+    },
+    {
+      "epoch": 63.33,
+      "learning_rate": 0.00019230769230769233,
+      "loss": 0.1628,
+      "step": 6903
+    },
+    {
+      "epoch": 63.34,
+      "learning_rate": 0.0001922596153846154,
+      "loss": 0.2575,
+      "step": 6904
+    },
+    {
+      "epoch": 63.35,
+      "learning_rate": 0.00019221153846153845,
+      "loss": 0.2593,
+      "step": 6905
+    },
+    {
+      "epoch": 63.36,
+      "learning_rate": 0.00019216346153846153,
+      "loss": 0.2762,
+      "step": 6906
+    },
+    {
+      "epoch": 63.37,
+      "learning_rate": 0.0001921153846153846,
+      "loss": 0.2003,
+      "step": 6907
+    },
+    {
+      "epoch": 63.38,
+      "learning_rate": 0.0001920673076923077,
+      "loss": 0.3577,
+      "step": 6908
+    },
+    {
+      "epoch": 63.39,
+      "learning_rate": 0.00019201923076923077,
+      "loss": 0.2349,
+      "step": 6909
+    },
+    {
+      "epoch": 63.39,
+      "learning_rate": 0.00019197115384615384,
+      "loss": 0.2931,
+      "step": 6910
+    },
+    {
+      "epoch": 63.4,
+      "learning_rate": 0.00019192307692307692,
+      "loss": 0.2539,
+      "step": 6911
+    },
+    {
+      "epoch": 63.41,
+      "learning_rate": 0.000191875,
+      "loss": 0.1299,
+      "step": 6912
+    },
+    {
+      "epoch": 63.42,
+      "learning_rate": 0.0001918269230769231,
+      "loss": 0.2778,
+      "step": 6913
+    },
+    {
+      "epoch": 63.43,
+      "learning_rate": 0.00019177884615384616,
+      "loss": 0.2709,
+      "step": 6914
+    },
+    {
+      "epoch": 63.44,
+      "learning_rate": 0.00019173076923076923,
+      "loss": 0.2071,
+      "step": 6915
+    },
+    {
+      "epoch": 63.45,
+      "learning_rate": 0.0001916826923076923,
+      "loss": 0.3227,
+      "step": 6916
+    },
+    {
+      "epoch": 63.46,
+      "learning_rate": 0.00019163461538461538,
+      "loss": 0.2632,
+      "step": 6917
+    },
+    {
+      "epoch": 63.47,
+      "learning_rate": 0.00019158653846153848,
+      "loss": 0.1553,
+      "step": 6918
+    },
+    {
+      "epoch": 63.48,
+      "learning_rate": 0.00019153846153846155,
+      "loss": 0.3222,
+      "step": 6919
+    },
+    {
+      "epoch": 63.49,
+      "learning_rate": 0.00019149038461538462,
+      "loss": 0.1846,
+      "step": 6920
+    },
+    {
+      "epoch": 63.5,
+      "learning_rate": 0.0001914423076923077,
+      "loss": 0.2319,
+      "step": 6921
+    },
+    {
+      "epoch": 63.5,
+      "learning_rate": 0.00019139423076923077,
+      "loss": 0.3695,
+      "step": 6922
+    },
+    {
+      "epoch": 63.51,
+      "learning_rate": 0.00019134615384615384,
+      "loss": 0.3768,
+      "step": 6923
+    },
+    {
+      "epoch": 63.52,
+      "learning_rate": 0.00019129807692307694,
+      "loss": 0.3946,
+      "step": 6924
+    },
+    {
+      "epoch": 63.53,
+      "learning_rate": 0.00019125000000000001,
+      "loss": 0.3375,
+      "step": 6925
+    },
+    {
+      "epoch": 63.54,
+      "learning_rate": 0.00019120192307692309,
+      "loss": 0.2511,
+      "step": 6926
+    },
+    {
+      "epoch": 63.55,
+      "learning_rate": 0.00019115384615384616,
+      "loss": 0.301,
+      "step": 6927
+    },
+    {
+      "epoch": 63.56,
+      "learning_rate": 0.00019110576923076923,
+      "loss": 0.2976,
+      "step": 6928
+    },
+    {
+      "epoch": 63.57,
+      "learning_rate": 0.00019105769230769233,
+      "loss": 0.2769,
+      "step": 6929
+    },
+    {
+      "epoch": 63.58,
+      "learning_rate": 0.0001910096153846154,
+      "loss": 0.2718,
+      "step": 6930
+    },
+    {
+      "epoch": 63.59,
+      "learning_rate": 0.00019096153846153848,
+      "loss": 0.2656,
+      "step": 6931
+    },
+    {
+      "epoch": 63.6,
+      "learning_rate": 0.00019091346153846155,
+      "loss": 0.3542,
+      "step": 6932
+    },
+    {
+      "epoch": 63.61,
+      "learning_rate": 0.0001908653846153846,
+      "loss": 0.225,
+      "step": 6933
+    },
+    {
+      "epoch": 63.61,
+      "learning_rate": 0.0001908173076923077,
+      "loss": 0.3035,
+      "step": 6934
+    },
+    {
+      "epoch": 63.62,
+      "learning_rate": 0.00019076923076923077,
+      "loss": 0.2171,
+      "step": 6935
+    },
+    {
+      "epoch": 63.63,
+      "learning_rate": 0.00019072115384615384,
+      "loss": 0.2889,
+      "step": 6936
+    },
+    {
+      "epoch": 63.64,
+      "learning_rate": 0.0001906730769230769,
+      "loss": 0.3202,
+      "step": 6937
+    },
+    {
+      "epoch": 63.65,
+      "learning_rate": 0.00019062499999999998,
+      "loss": 0.2307,
+      "step": 6938
+    },
+    {
+      "epoch": 63.66,
+      "learning_rate": 0.00019057692307692308,
+      "loss": 0.288,
+      "step": 6939
+    },
+    {
+      "epoch": 63.67,
+      "learning_rate": 0.00019052884615384616,
+      "loss": 0.2384,
+      "step": 6940
+    },
+    {
+      "epoch": 63.68,
+      "learning_rate": 0.00019048076923076923,
+      "loss": 0.296,
+      "step": 6941
+    },
+    {
+      "epoch": 63.69,
+      "learning_rate": 0.0001904326923076923,
+      "loss": 0.3529,
+      "step": 6942
+    },
+    {
+      "epoch": 63.7,
+      "learning_rate": 0.00019038461538461538,
+      "loss": 0.2935,
+      "step": 6943
+    },
+    {
+      "epoch": 63.71,
+      "learning_rate": 0.00019033653846153847,
+      "loss": 0.2458,
+      "step": 6944
+    },
+    {
+      "epoch": 63.72,
+      "learning_rate": 0.00019028846153846155,
+      "loss": 0.2616,
+      "step": 6945
+    },
+    {
+      "epoch": 63.72,
+      "learning_rate": 0.00019024038461538462,
+      "loss": 0.2577,
+      "step": 6946
+    },
+    {
+      "epoch": 63.73,
+      "learning_rate": 0.0001901923076923077,
+      "loss": 0.2506,
+      "step": 6947
+    },
+    {
+      "epoch": 63.74,
+      "learning_rate": 0.00019014423076923077,
+      "loss": 0.2497,
+      "step": 6948
+    },
+    {
+      "epoch": 63.75,
+      "learning_rate": 0.00019009615384615387,
+      "loss": 0.32,
+      "step": 6949
+    },
+    {
+      "epoch": 63.76,
+      "learning_rate": 0.00019004807692307694,
+      "loss": 0.3556,
+      "step": 6950
+    },
+    {
+      "epoch": 63.77,
+      "learning_rate": 0.00019,
+      "loss": 0.2752,
+      "step": 6951
+    },
+    {
+      "epoch": 63.78,
+      "learning_rate": 0.00018995192307692308,
+      "loss": 0.287,
+      "step": 6952
+    },
+    {
+      "epoch": 63.79,
+      "learning_rate": 0.00018990384615384616,
+      "loss": 0.3224,
+      "step": 6953
+    },
+    {
+      "epoch": 63.8,
+      "learning_rate": 0.00018985576923076926,
+      "loss": 0.2638,
+      "step": 6954
+    },
+    {
+      "epoch": 63.81,
+      "learning_rate": 0.00018980769230769233,
+      "loss": 0.3634,
+      "step": 6955
+    },
+    {
+      "epoch": 63.82,
+      "learning_rate": 0.0001897596153846154,
+      "loss": 0.3224,
+      "step": 6956
+    },
+    {
+      "epoch": 63.83,
+      "learning_rate": 0.00018971153846153847,
+      "loss": 0.2287,
+      "step": 6957
+    },
+    {
+      "epoch": 63.83,
+      "learning_rate": 0.00018966346153846155,
+      "loss": 0.2452,
+      "step": 6958
+    },
+    {
+      "epoch": 63.84,
+      "learning_rate": 0.00018961538461538462,
+      "loss": 0.2894,
+      "step": 6959
+    },
+    {
+      "epoch": 63.85,
+      "learning_rate": 0.00018956730769230772,
+      "loss": 0.2607,
+      "step": 6960
+    },
+    {
+      "epoch": 63.86,
+      "learning_rate": 0.00018951923076923076,
+      "loss": 0.2373,
+      "step": 6961
+    },
+    {
+      "epoch": 63.87,
+      "learning_rate": 0.00018947115384615384,
+      "loss": 0.2072,
+      "step": 6962
+    },
+    {
+      "epoch": 63.88,
+      "learning_rate": 0.0001894230769230769,
+      "loss": 0.2454,
+      "step": 6963
+    },
+    {
+      "epoch": 63.89,
+      "learning_rate": 0.00018937499999999998,
+      "loss": 0.2414,
+      "step": 6964
+    },
+    {
+      "epoch": 63.9,
+      "learning_rate": 0.00018932692307692308,
+      "loss": 0.2471,
+      "step": 6965
+    },
+    {
+      "epoch": 63.91,
+      "learning_rate": 0.00018927884615384615,
+      "loss": 0.2896,
+      "step": 6966
+    },
+    {
+      "epoch": 63.92,
+      "learning_rate": 0.00018923076923076923,
+      "loss": 0.2806,
+      "step": 6967
+    },
+    {
+      "epoch": 63.93,
+      "learning_rate": 0.0001891826923076923,
+      "loss": 0.2661,
+      "step": 6968
+    },
+    {
+      "epoch": 63.94,
+      "learning_rate": 0.00018913461538461537,
+      "loss": 0.2691,
+      "step": 6969
+    },
+    {
+      "epoch": 63.94,
+      "learning_rate": 0.00018908653846153847,
+      "loss": 0.2725,
+      "step": 6970
+    },
+    {
+      "epoch": 63.95,
+      "learning_rate": 0.00018903846153846154,
+      "loss": 0.3382,
+      "step": 6971
+    },
+    {
+      "epoch": 63.96,
+      "learning_rate": 0.00018899038461538462,
+      "loss": 0.3167,
+      "step": 6972
+    },
+    {
+      "epoch": 63.97,
+      "learning_rate": 0.0001889423076923077,
+      "loss": 0.2776,
+      "step": 6973
+    },
+    {
+      "epoch": 63.98,
+      "learning_rate": 0.00018889423076923076,
+      "loss": 0.27,
+      "step": 6974
+    },
+    {
+      "epoch": 63.99,
+      "learning_rate": 0.00018884615384615386,
+      "loss": 0.2692,
+      "step": 6975
+    },
+    {
+      "epoch": 64.0,
+      "learning_rate": 0.00018879807692307693,
+      "loss": 0.285,
+      "step": 6976
+    },
+    {
+      "epoch": 64.01,
+      "learning_rate": 0.00018875,
+      "loss": 0.3708,
+      "step": 6977
+    },
+    {
+      "epoch": 64.02,
+      "learning_rate": 0.00018870192307692308,
+      "loss": 0.3545,
+      "step": 6978
+    },
+    {
+      "epoch": 64.03,
+      "learning_rate": 0.00018865384615384615,
+      "loss": 0.3102,
+      "step": 6979
+    },
+    {
+      "epoch": 64.04,
+      "learning_rate": 0.00018860576923076925,
+      "loss": 0.3124,
+      "step": 6980
+    },
+    {
+      "epoch": 64.05,
+      "learning_rate": 0.00018855769230769232,
+      "loss": 0.2895,
+      "step": 6981
+    },
+    {
+      "epoch": 64.06,
+      "learning_rate": 0.0001885096153846154,
+      "loss": 0.2343,
+      "step": 6982
+    },
+    {
+      "epoch": 64.06,
+      "learning_rate": 0.00018846153846153847,
+      "loss": 0.3431,
+      "step": 6983
+    },
+    {
+      "epoch": 64.07,
+      "learning_rate": 0.00018841346153846154,
+      "loss": 0.2434,
+      "step": 6984
+    },
+    {
+      "epoch": 64.08,
+      "learning_rate": 0.00018836538461538464,
+      "loss": 0.2918,
+      "step": 6985
+    },
+    {
+      "epoch": 64.09,
+      "learning_rate": 0.00018831730769230772,
+      "loss": 0.2783,
+      "step": 6986
+    },
+    {
+      "epoch": 64.1,
+      "learning_rate": 0.0001882692307692308,
+      "loss": 0.2871,
+      "step": 6987
+    },
+    {
+      "epoch": 64.11,
+      "learning_rate": 0.00018822115384615386,
+      "loss": 0.2966,
+      "step": 6988
+    },
+    {
+      "epoch": 64.12,
+      "learning_rate": 0.0001881730769230769,
+      "loss": 0.2792,
+      "step": 6989
+    },
+    {
+      "epoch": 64.13,
+      "learning_rate": 0.00018812499999999998,
+      "loss": 0.2447,
+      "step": 6990
+    },
+    {
+      "epoch": 64.14,
+      "learning_rate": 0.00018807692307692308,
+      "loss": 0.2797,
+      "step": 6991
+    },
+    {
+      "epoch": 64.15,
+      "learning_rate": 0.00018802884615384615,
+      "loss": 0.3314,
+      "step": 6992
+    },
+    {
+      "epoch": 64.16,
+      "learning_rate": 0.00018798076923076922,
+      "loss": 0.2438,
+      "step": 6993
+    },
+    {
+      "epoch": 64.17,
+      "learning_rate": 0.0001879326923076923,
+      "loss": 0.2671,
+      "step": 6994
+    },
+    {
+      "epoch": 64.17,
+      "learning_rate": 0.00018788461538461537,
+      "loss": 0.2731,
+      "step": 6995
+    },
+    {
+      "epoch": 64.18,
+      "learning_rate": 0.00018783653846153847,
+      "loss": 0.2328,
+      "step": 6996
+    },
+    {
+      "epoch": 64.19,
+      "learning_rate": 0.00018778846153846154,
+      "loss": 0.289,
+      "step": 6997
+    },
+    {
+      "epoch": 64.2,
+      "learning_rate": 0.00018774038461538461,
+      "loss": 0.2009,
+      "step": 6998
+    },
+    {
+      "epoch": 64.21,
+      "learning_rate": 0.00018769230769230769,
+      "loss": 0.2181,
+      "step": 6999
+    },
+    {
+      "epoch": 64.22,
+      "learning_rate": 0.00018764423076923076,
+      "loss": 0.2576,
+      "step": 7000
+    },
+    {
+      "epoch": 64.22,
+      "eval_cer": 0.15538463845035158,
+      "eval_loss": 0.6919934153556824,
+      "eval_runtime": 87.0228,
+      "eval_samples_per_second": 18.926,
+      "eval_steps_per_second": 2.367,
+      "eval_wer": 0.5316106628536411,
+      "step": 7000
+    },
+    {
+      "epoch": 64.23,
+      "learning_rate": 0.00018759615384615386,
+      "loss": 0.3066,
+      "step": 7001
+    },
+    {
+      "epoch": 64.24,
+      "learning_rate": 0.00018754807692307693,
+      "loss": 0.3449,
+      "step": 7002
+    },
+    {
+      "epoch": 64.25,
+      "learning_rate": 0.0001875,
+      "loss": 0.1924,
+      "step": 7003
+    },
+    {
+      "epoch": 64.26,
+      "learning_rate": 0.00018745192307692308,
+      "loss": 0.2826,
+      "step": 7004
+    },
+    {
+      "epoch": 64.27,
+      "learning_rate": 0.00018740384615384615,
+      "loss": 0.3369,
+      "step": 7005
+    },
+    {
+      "epoch": 64.28,
+      "learning_rate": 0.00018735576923076925,
+      "loss": 0.2482,
+      "step": 7006
+    },
+    {
+      "epoch": 64.28,
+      "learning_rate": 0.00018730769230769232,
+      "loss": 0.3332,
+      "step": 7007
+    },
+    {
+      "epoch": 64.29,
+      "learning_rate": 0.0001872596153846154,
+      "loss": 0.2914,
+      "step": 7008
+    },
+    {
+      "epoch": 64.3,
+      "learning_rate": 0.00018721153846153847,
+      "loss": 0.3366,
+      "step": 7009
+    },
+    {
+      "epoch": 64.31,
+      "learning_rate": 0.00018716346153846154,
+      "loss": 0.2288,
+      "step": 7010
+    },
+    {
+      "epoch": 64.32,
+      "learning_rate": 0.00018711538461538464,
+      "loss": 0.1771,
+      "step": 7011
+    },
+    {
+      "epoch": 64.33,
+      "learning_rate": 0.0001870673076923077,
+      "loss": 0.2102,
+      "step": 7012
+    },
+    {
+      "epoch": 64.34,
+      "learning_rate": 0.00018701923076923078,
+      "loss": 0.2491,
+      "step": 7013
+    },
+    {
+      "epoch": 64.35,
+      "learning_rate": 0.00018697115384615386,
+      "loss": 0.231,
+      "step": 7014
+    },
+    {
+      "epoch": 64.36,
+      "learning_rate": 0.00018692307692307693,
+      "loss": 0.2688,
+      "step": 7015
+    },
+    {
+      "epoch": 64.37,
+      "learning_rate": 0.00018687500000000003,
+      "loss": 0.285,
+      "step": 7016
+    },
+    {
+      "epoch": 64.38,
+      "learning_rate": 0.0001868269230769231,
+      "loss": 0.2791,
+      "step": 7017
+    },
+    {
+      "epoch": 64.39,
+      "learning_rate": 0.00018677884615384615,
+      "loss": 0.3002,
+      "step": 7018
+    },
+    {
+      "epoch": 64.39,
+      "learning_rate": 0.00018673076923076922,
+      "loss": 0.3039,
+      "step": 7019
+    },
+    {
+      "epoch": 64.4,
+      "learning_rate": 0.0001866826923076923,
+      "loss": 0.1964,
+      "step": 7020
+    },
+    {
+      "epoch": 64.41,
+      "learning_rate": 0.00018663461538461537,
+      "loss": 0.27,
+      "step": 7021
+    },
+    {
+      "epoch": 64.42,
+      "learning_rate": 0.00018658653846153847,
+      "loss": 0.2641,
+      "step": 7022
+    },
+    {
+      "epoch": 64.43,
+      "learning_rate": 0.00018653846153846154,
+      "loss": 0.2571,
+      "step": 7023
+    },
+    {
+      "epoch": 64.44,
+      "learning_rate": 0.0001864903846153846,
+      "loss": 0.2426,
+      "step": 7024
+    },
+    {
+      "epoch": 64.45,
+      "learning_rate": 0.00018644230769230768,
+      "loss": 0.2353,
+      "step": 7025
+    },
+    {
+      "epoch": 64.46,
+      "learning_rate": 0.00018639423076923076,
+      "loss": 0.2144,
+      "step": 7026
+    },
+    {
+      "epoch": 64.47,
+      "learning_rate": 0.00018634615384615386,
+      "loss": 0.3515,
+      "step": 7027
+    },
+    {
+      "epoch": 64.48,
+      "learning_rate": 0.00018629807692307693,
+      "loss": 0.229,
+      "step": 7028
+    },
+    {
+      "epoch": 64.49,
+      "learning_rate": 0.00018625,
+      "loss": 0.2302,
+      "step": 7029
+    },
+    {
+      "epoch": 64.5,
+      "learning_rate": 0.00018620192307692307,
+      "loss": 0.1431,
+      "step": 7030
+    },
+    {
+      "epoch": 64.5,
+      "learning_rate": 0.00018615384615384615,
+      "loss": 0.3838,
+      "step": 7031
+    },
+    {
+      "epoch": 64.51,
+      "learning_rate": 0.00018610576923076925,
+      "loss": 0.3014,
+      "step": 7032
+    },
+    {
+      "epoch": 64.52,
+      "learning_rate": 0.00018605769230769232,
+      "loss": 0.2774,
+      "step": 7033
+    },
+    {
+      "epoch": 64.53,
+      "learning_rate": 0.0001860096153846154,
+      "loss": 0.2385,
+      "step": 7034
+    },
+    {
+      "epoch": 64.54,
+      "learning_rate": 0.00018596153846153846,
+      "loss": 0.2648,
+      "step": 7035
+    },
+    {
+      "epoch": 64.55,
+      "learning_rate": 0.00018591346153846154,
+      "loss": 0.2311,
+      "step": 7036
+    },
+    {
+      "epoch": 64.56,
+      "learning_rate": 0.00018586538461538464,
+      "loss": 0.2458,
+      "step": 7037
+    },
+    {
+      "epoch": 64.57,
+      "learning_rate": 0.0001858173076923077,
+      "loss": 0.2505,
+      "step": 7038
+    },
+    {
+      "epoch": 64.58,
+      "learning_rate": 0.00018576923076923078,
+      "loss": 0.2177,
+      "step": 7039
+    },
+    {
+      "epoch": 64.59,
+      "learning_rate": 0.00018572115384615385,
+      "loss": 0.2519,
+      "step": 7040
+    },
+    {
+      "epoch": 64.6,
+      "learning_rate": 0.00018567307692307693,
+      "loss": 0.248,
+      "step": 7041
+    },
+    {
+      "epoch": 64.61,
+      "learning_rate": 0.00018562500000000003,
+      "loss": 0.2919,
+      "step": 7042
+    },
+    {
+      "epoch": 64.61,
+      "learning_rate": 0.0001855769230769231,
+      "loss": 0.3328,
+      "step": 7043
+    },
+    {
+      "epoch": 64.62,
+      "learning_rate": 0.00018552884615384617,
+      "loss": 0.356,
+      "step": 7044
+    },
+    {
+      "epoch": 64.63,
+      "learning_rate": 0.00018548076923076924,
+      "loss": 0.2558,
+      "step": 7045
+    },
+    {
+      "epoch": 64.64,
+      "learning_rate": 0.0001854326923076923,
+      "loss": 0.231,
+      "step": 7046
+    },
+    {
+      "epoch": 64.65,
+      "learning_rate": 0.0001853846153846154,
+      "loss": 0.2269,
+      "step": 7047
+    },
+    {
+      "epoch": 64.66,
+      "learning_rate": 0.00018533653846153846,
+      "loss": 0.2336,
+      "step": 7048
+    },
+    {
+      "epoch": 64.67,
+      "learning_rate": 0.00018528846153846153,
+      "loss": 0.3033,
+      "step": 7049
+    },
+    {
+      "epoch": 64.68,
+      "learning_rate": 0.0001852403846153846,
+      "loss": 0.2661,
+      "step": 7050
+    },
+    {
+      "epoch": 64.69,
+      "learning_rate": 0.00018519230769230768,
+      "loss": 0.2773,
+      "step": 7051
+    },
+    {
+      "epoch": 64.7,
+      "learning_rate": 0.00018514423076923075,
+      "loss": 0.2295,
+      "step": 7052
+    },
+    {
+      "epoch": 64.71,
+      "learning_rate": 0.00018509615384615385,
+      "loss": 0.2043,
+      "step": 7053
+    },
+    {
+      "epoch": 64.72,
+      "learning_rate": 0.00018504807692307692,
+      "loss": 0.1733,
+      "step": 7054
+    },
+    {
+      "epoch": 64.72,
+      "learning_rate": 0.000185,
+      "loss": 0.2628,
+      "step": 7055
+    },
+    {
+      "epoch": 64.73,
+      "learning_rate": 0.00018495192307692307,
+      "loss": 0.2649,
+      "step": 7056
+    },
+    {
+      "epoch": 64.74,
+      "learning_rate": 0.00018490384615384614,
+      "loss": 0.2251,
+      "step": 7057
+    },
+    {
+      "epoch": 64.75,
+      "learning_rate": 0.00018485576923076924,
+      "loss": 0.3364,
+      "step": 7058
+    },
+    {
+      "epoch": 64.76,
+      "learning_rate": 0.00018480769230769231,
+      "loss": 0.2542,
+      "step": 7059
+    },
+    {
+      "epoch": 64.77,
+      "learning_rate": 0.0001847596153846154,
+      "loss": 0.2799,
+      "step": 7060
+    },
+    {
+      "epoch": 64.78,
+      "learning_rate": 0.00018471153846153846,
+      "loss": 0.2657,
+      "step": 7061
+    },
+    {
+      "epoch": 64.79,
+      "learning_rate": 0.00018466346153846153,
+      "loss": 0.2961,
+      "step": 7062
+    },
+    {
+      "epoch": 64.8,
+      "learning_rate": 0.00018461538461538463,
+      "loss": 0.372,
+      "step": 7063
+    },
+    {
+      "epoch": 64.81,
+      "learning_rate": 0.0001845673076923077,
+      "loss": 0.27,
+      "step": 7064
+    },
+    {
+      "epoch": 64.82,
+      "learning_rate": 0.00018451923076923078,
+      "loss": 0.2559,
+      "step": 7065
+    },
+    {
+      "epoch": 64.83,
+      "learning_rate": 0.00018447115384615385,
+      "loss": 0.2335,
+      "step": 7066
+    },
+    {
+      "epoch": 64.83,
+      "learning_rate": 0.00018442307692307692,
+      "loss": 0.2401,
+      "step": 7067
+    },
+    {
+      "epoch": 64.84,
+      "learning_rate": 0.00018437500000000002,
+      "loss": 0.2496,
+      "step": 7068
+    },
+    {
+      "epoch": 64.85,
+      "learning_rate": 0.0001843269230769231,
+      "loss": 0.219,
+      "step": 7069
+    },
+    {
+      "epoch": 64.86,
+      "learning_rate": 0.00018427884615384617,
+      "loss": 0.1992,
+      "step": 7070
+    },
+    {
+      "epoch": 64.87,
+      "learning_rate": 0.00018423076923076924,
+      "loss": 0.2687,
+      "step": 7071
+    },
+    {
+      "epoch": 64.88,
+      "learning_rate": 0.0001841826923076923,
+      "loss": 0.2394,
+      "step": 7072
+    },
+    {
+      "epoch": 64.89,
+      "learning_rate": 0.0001841346153846154,
+      "loss": 0.326,
+      "step": 7073
+    },
+    {
+      "epoch": 64.9,
+      "learning_rate": 0.00018408653846153846,
+      "loss": 0.2424,
+      "step": 7074
+    },
+    {
+      "epoch": 64.91,
+      "learning_rate": 0.00018403846153846153,
+      "loss": 0.1879,
+      "step": 7075
+    },
+    {
+      "epoch": 64.92,
+      "learning_rate": 0.0001839903846153846,
+      "loss": 0.2733,
+      "step": 7076
+    },
+    {
+      "epoch": 64.93,
+      "learning_rate": 0.00018394230769230768,
+      "loss": 0.2148,
+      "step": 7077
+    },
+    {
+      "epoch": 64.94,
+      "learning_rate": 0.00018389423076923078,
+      "loss": 0.2131,
+      "step": 7078
+    },
+    {
+      "epoch": 64.94,
+      "learning_rate": 0.00018384615384615385,
+      "loss": 0.3097,
+      "step": 7079
+    },
+    {
+      "epoch": 64.95,
+      "learning_rate": 0.00018379807692307692,
+      "loss": 0.3052,
+      "step": 7080
+    },
+    {
+      "epoch": 64.96,
+      "learning_rate": 0.00018375,
+      "loss": 0.2689,
+      "step": 7081
+    },
+    {
+      "epoch": 64.97,
+      "learning_rate": 0.00018370192307692307,
+      "loss": 0.1987,
+      "step": 7082
+    },
+    {
+      "epoch": 64.98,
+      "learning_rate": 0.00018365384615384617,
+      "loss": 0.2329,
+      "step": 7083
+    },
+    {
+      "epoch": 64.99,
+      "learning_rate": 0.00018360576923076924,
+      "loss": 0.2565,
+      "step": 7084
+    },
+    {
+      "epoch": 65.0,
+      "learning_rate": 0.0001835576923076923,
+      "loss": 0.2225,
+      "step": 7085
+    },
+    {
+      "epoch": 65.01,
+      "learning_rate": 0.00018350961538461538,
+      "loss": 0.3077,
+      "step": 7086
+    },
+    {
+      "epoch": 65.02,
+      "learning_rate": 0.00018346153846153846,
+      "loss": 0.2789,
+      "step": 7087
+    },
+    {
+      "epoch": 65.03,
+      "learning_rate": 0.00018341346153846153,
+      "loss": 0.3207,
+      "step": 7088
+    },
+    {
+      "epoch": 65.04,
+      "learning_rate": 0.00018336538461538463,
+      "loss": 0.2703,
+      "step": 7089
+    },
+    {
+      "epoch": 65.05,
+      "learning_rate": 0.0001833173076923077,
+      "loss": 0.2802,
+      "step": 7090
+    },
+    {
+      "epoch": 65.06,
+      "learning_rate": 0.00018326923076923077,
+      "loss": 0.2536,
+      "step": 7091
+    },
+    {
+      "epoch": 65.06,
+      "learning_rate": 0.00018322115384615385,
+      "loss": 0.2556,
+      "step": 7092
+    },
+    {
+      "epoch": 65.07,
+      "learning_rate": 0.00018317307692307692,
+      "loss": 0.244,
+      "step": 7093
+    },
+    {
+      "epoch": 65.08,
+      "learning_rate": 0.00018312500000000002,
+      "loss": 0.2999,
+      "step": 7094
+    },
+    {
+      "epoch": 65.09,
+      "learning_rate": 0.0001830769230769231,
+      "loss": 0.3417,
+      "step": 7095
+    },
+    {
+      "epoch": 65.1,
+      "learning_rate": 0.00018302884615384616,
+      "loss": 0.3776,
+      "step": 7096
+    },
+    {
+      "epoch": 65.11,
+      "learning_rate": 0.00018298076923076924,
+      "loss": 0.1732,
+      "step": 7097
+    },
+    {
+      "epoch": 65.12,
+      "learning_rate": 0.0001829326923076923,
+      "loss": 0.2724,
+      "step": 7098
+    },
+    {
+      "epoch": 65.13,
+      "learning_rate": 0.0001828846153846154,
+      "loss": 0.2241,
+      "step": 7099
+    },
+    {
+      "epoch": 65.14,
+      "learning_rate": 0.00018283653846153848,
+      "loss": 0.2357,
+      "step": 7100
+    },
+    {
+      "epoch": 65.15,
+      "learning_rate": 0.00018278846153846156,
+      "loss": 0.3352,
+      "step": 7101
+    },
+    {
+      "epoch": 65.16,
+      "learning_rate": 0.0001827403846153846,
+      "loss": 0.3108,
+      "step": 7102
+    },
+    {
+      "epoch": 65.17,
+      "learning_rate": 0.00018269230769230767,
+      "loss": 0.3158,
+      "step": 7103
+    },
+    {
+      "epoch": 65.17,
+      "learning_rate": 0.00018264423076923077,
+      "loss": 0.2486,
+      "step": 7104
+    },
+    {
+      "epoch": 65.18,
+      "learning_rate": 0.00018259615384615385,
+      "loss": 0.2083,
+      "step": 7105
+    },
+    {
+      "epoch": 65.19,
+      "learning_rate": 0.00018254807692307692,
+      "loss": 0.222,
+      "step": 7106
+    },
+    {
+      "epoch": 65.2,
+      "learning_rate": 0.0001825,
+      "loss": 0.2422,
+      "step": 7107
+    },
+    {
+      "epoch": 65.21,
+      "learning_rate": 0.00018245192307692306,
+      "loss": 0.1673,
+      "step": 7108
+    },
+    {
+      "epoch": 65.22,
+      "learning_rate": 0.00018240384615384616,
+      "loss": 0.3146,
+      "step": 7109
+    },
+    {
+      "epoch": 65.23,
+      "learning_rate": 0.00018235576923076924,
+      "loss": 0.2526,
+      "step": 7110
+    },
+    {
+      "epoch": 65.24,
+      "learning_rate": 0.0001823076923076923,
+      "loss": 0.2607,
+      "step": 7111
+    },
+    {
+      "epoch": 65.25,
+      "learning_rate": 0.00018225961538461538,
+      "loss": 0.203,
+      "step": 7112
+    },
+    {
+      "epoch": 65.26,
+      "learning_rate": 0.00018221153846153845,
+      "loss": 0.285,
+      "step": 7113
+    },
+    {
+      "epoch": 65.27,
+      "learning_rate": 0.00018216346153846155,
+      "loss": 0.31,
+      "step": 7114
+    },
+    {
+      "epoch": 65.28,
+      "learning_rate": 0.00018211538461538463,
+      "loss": 0.3014,
+      "step": 7115
+    },
+    {
+      "epoch": 65.28,
+      "learning_rate": 0.0001820673076923077,
+      "loss": 0.3365,
+      "step": 7116
+    },
+    {
+      "epoch": 65.29,
+      "learning_rate": 0.00018201923076923077,
+      "loss": 0.2232,
+      "step": 7117
+    },
+    {
+      "epoch": 65.3,
+      "learning_rate": 0.00018197115384615384,
+      "loss": 0.2615,
+      "step": 7118
+    },
+    {
+      "epoch": 65.31,
+      "learning_rate": 0.00018192307692307692,
+      "loss": 0.1685,
+      "step": 7119
+    },
+    {
+      "epoch": 65.32,
+      "learning_rate": 0.00018187500000000002,
+      "loss": 0.2774,
+      "step": 7120
+    },
+    {
+      "epoch": 65.33,
+      "learning_rate": 0.0001818269230769231,
+      "loss": 0.2465,
+      "step": 7121
+    },
+    {
+      "epoch": 65.34,
+      "learning_rate": 0.00018177884615384616,
+      "loss": 0.3373,
+      "step": 7122
+    },
+    {
+      "epoch": 65.35,
+      "learning_rate": 0.00018173076923076923,
+      "loss": 0.2074,
+      "step": 7123
+    },
+    {
+      "epoch": 65.36,
+      "learning_rate": 0.0001816826923076923,
+      "loss": 0.2032,
+      "step": 7124
+    },
+    {
+      "epoch": 65.37,
+      "learning_rate": 0.0001816346153846154,
+      "loss": 0.2691,
+      "step": 7125
+    },
+    {
+      "epoch": 65.38,
+      "learning_rate": 0.00018158653846153848,
+      "loss": 0.3311,
+      "step": 7126
+    },
+    {
+      "epoch": 65.39,
+      "learning_rate": 0.00018153846153846155,
+      "loss": 0.2513,
+      "step": 7127
+    },
+    {
+      "epoch": 65.39,
+      "learning_rate": 0.00018149038461538462,
+      "loss": 0.2183,
+      "step": 7128
+    },
+    {
+      "epoch": 65.4,
+      "learning_rate": 0.0001814423076923077,
+      "loss": 0.2325,
+      "step": 7129
+    },
+    {
+      "epoch": 65.41,
+      "learning_rate": 0.0001813942307692308,
+      "loss": 0.2591,
+      "step": 7130
+    },
+    {
+      "epoch": 65.42,
+      "learning_rate": 0.00018134615384615384,
+      "loss": 0.1785,
+      "step": 7131
+    },
+    {
+      "epoch": 65.43,
+      "learning_rate": 0.00018129807692307691,
+      "loss": 0.219,
+      "step": 7132
+    },
+    {
+      "epoch": 65.44,
+      "learning_rate": 0.00018125,
+      "loss": 0.269,
+      "step": 7133
+    },
+    {
+      "epoch": 65.45,
+      "learning_rate": 0.00018120192307692306,
+      "loss": 0.252,
+      "step": 7134
+    },
+    {
+      "epoch": 65.46,
+      "learning_rate": 0.00018115384615384616,
+      "loss": 0.248,
+      "step": 7135
+    },
+    {
+      "epoch": 65.47,
+      "learning_rate": 0.00018110576923076923,
+      "loss": 0.2187,
+      "step": 7136
+    },
+    {
+      "epoch": 65.48,
+      "learning_rate": 0.0001810576923076923,
+      "loss": 0.2354,
+      "step": 7137
+    },
+    {
+      "epoch": 65.49,
+      "learning_rate": 0.00018100961538461538,
+      "loss": 0.2384,
+      "step": 7138
+    },
+    {
+      "epoch": 65.5,
+      "learning_rate": 0.00018096153846153845,
+      "loss": 0.1114,
+      "step": 7139
+    },
+    {
+      "epoch": 65.5,
+      "learning_rate": 0.00018091346153846155,
+      "loss": 0.3339,
+      "step": 7140
+    },
+    {
+      "epoch": 65.51,
+      "learning_rate": 0.00018086538461538462,
+      "loss": 0.253,
+      "step": 7141
+    },
+    {
+      "epoch": 65.52,
+      "learning_rate": 0.0001808173076923077,
+      "loss": 0.2891,
+      "step": 7142
+    },
+    {
+      "epoch": 65.53,
+      "learning_rate": 0.00018076923076923077,
+      "loss": 0.2315,
+      "step": 7143
+    },
+    {
+      "epoch": 65.54,
+      "learning_rate": 0.00018072115384615384,
+      "loss": 0.23,
+      "step": 7144
+    },
+    {
+      "epoch": 65.55,
+      "learning_rate": 0.00018067307692307694,
+      "loss": 0.2694,
+      "step": 7145
+    },
+    {
+      "epoch": 65.56,
+      "learning_rate": 0.000180625,
+      "loss": 0.2155,
+      "step": 7146
+    },
+    {
+      "epoch": 65.57,
+      "learning_rate": 0.00018057692307692309,
+      "loss": 0.2189,
+      "step": 7147
+    },
+    {
+      "epoch": 65.58,
+      "learning_rate": 0.00018052884615384616,
+      "loss": 0.2758,
+      "step": 7148
+    },
+    {
+      "epoch": 65.59,
+      "learning_rate": 0.00018048076923076923,
+      "loss": 0.2188,
+      "step": 7149
+    },
+    {
+      "epoch": 65.6,
+      "learning_rate": 0.0001804326923076923,
+      "loss": 0.2862,
+      "step": 7150
+    },
+    {
+      "epoch": 65.61,
+      "learning_rate": 0.0001803846153846154,
+      "loss": 0.2629,
+      "step": 7151
+    },
+    {
+      "epoch": 65.61,
+      "learning_rate": 0.00018033653846153848,
+      "loss": 0.34,
+      "step": 7152
+    },
+    {
+      "epoch": 65.62,
+      "learning_rate": 0.00018028846153846155,
+      "loss": 0.1913,
+      "step": 7153
+    },
+    {
+      "epoch": 65.63,
+      "learning_rate": 0.00018024038461538462,
+      "loss": 0.1977,
+      "step": 7154
+    },
+    {
+      "epoch": 65.64,
+      "learning_rate": 0.0001801923076923077,
+      "loss": 0.25,
+      "step": 7155
+    },
+    {
+      "epoch": 65.65,
+      "learning_rate": 0.0001801442307692308,
+      "loss": 0.2007,
+      "step": 7156
+    },
+    {
+      "epoch": 65.66,
+      "learning_rate": 0.00018009615384615387,
+      "loss": 0.252,
+      "step": 7157
+    },
+    {
+      "epoch": 65.67,
+      "learning_rate": 0.00018004807692307694,
+      "loss": 0.2479,
+      "step": 7158
+    },
+    {
+      "epoch": 65.68,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.155,
+      "step": 7159
+    },
+    {
+      "epoch": 65.69,
+      "learning_rate": 0.00017995192307692306,
+      "loss": 0.2263,
+      "step": 7160
+    },
+    {
+      "epoch": 65.7,
+      "learning_rate": 0.00017990384615384616,
+      "loss": 0.2437,
+      "step": 7161
+    },
+    {
+      "epoch": 65.71,
+      "learning_rate": 0.00017985576923076923,
+      "loss": 0.1806,
+      "step": 7162
+    },
+    {
+      "epoch": 65.72,
+      "learning_rate": 0.0001798076923076923,
+      "loss": 0.3327,
+      "step": 7163
+    },
+    {
+      "epoch": 65.72,
+      "learning_rate": 0.00017975961538461537,
+      "loss": 0.191,
+      "step": 7164
+    },
+    {
+      "epoch": 65.73,
+      "learning_rate": 0.00017971153846153845,
+      "loss": 0.2896,
+      "step": 7165
+    },
+    {
+      "epoch": 65.74,
+      "learning_rate": 0.00017966346153846155,
+      "loss": 0.1449,
+      "step": 7166
+    },
+    {
+      "epoch": 65.75,
+      "learning_rate": 0.00017961538461538462,
+      "loss": 0.3135,
+      "step": 7167
+    },
+    {
+      "epoch": 65.76,
+      "learning_rate": 0.0001795673076923077,
+      "loss": 0.3209,
+      "step": 7168
+    },
+    {
+      "epoch": 65.77,
+      "learning_rate": 0.00017951923076923076,
+      "loss": 0.2931,
+      "step": 7169
+    },
+    {
+      "epoch": 65.78,
+      "learning_rate": 0.00017947115384615384,
+      "loss": 0.2944,
+      "step": 7170
+    },
+    {
+      "epoch": 65.79,
+      "learning_rate": 0.00017942307692307694,
+      "loss": 0.3102,
+      "step": 7171
+    },
+    {
+      "epoch": 65.8,
+      "learning_rate": 0.000179375,
+      "loss": 0.3227,
+      "step": 7172
+    },
+    {
+      "epoch": 65.81,
+      "learning_rate": 0.00017932692307692308,
+      "loss": 0.2829,
+      "step": 7173
+    },
+    {
+      "epoch": 65.82,
+      "learning_rate": 0.00017927884615384615,
+      "loss": 0.1907,
+      "step": 7174
+    },
+    {
+      "epoch": 65.83,
+      "learning_rate": 0.00017923076923076923,
+      "loss": 0.2753,
+      "step": 7175
+    },
+    {
+      "epoch": 65.83,
+      "learning_rate": 0.00017918269230769233,
+      "loss": 0.2322,
+      "step": 7176
+    },
+    {
+      "epoch": 65.84,
+      "learning_rate": 0.0001791346153846154,
+      "loss": 0.236,
+      "step": 7177
+    },
+    {
+      "epoch": 65.85,
+      "learning_rate": 0.00017908653846153847,
+      "loss": 0.253,
+      "step": 7178
+    },
+    {
+      "epoch": 65.86,
+      "learning_rate": 0.00017903846153846155,
+      "loss": 0.258,
+      "step": 7179
+    },
+    {
+      "epoch": 65.87,
+      "learning_rate": 0.00017899038461538462,
+      "loss": 0.2577,
+      "step": 7180
+    },
+    {
+      "epoch": 65.88,
+      "learning_rate": 0.00017894230769230772,
+      "loss": 0.1907,
+      "step": 7181
+    },
+    {
+      "epoch": 65.89,
+      "learning_rate": 0.0001788942307692308,
+      "loss": 0.3185,
+      "step": 7182
+    },
+    {
+      "epoch": 65.9,
+      "learning_rate": 0.00017884615384615386,
+      "loss": 0.2033,
+      "step": 7183
+    },
+    {
+      "epoch": 65.91,
+      "learning_rate": 0.00017879807692307694,
+      "loss": 0.2048,
+      "step": 7184
+    },
+    {
+      "epoch": 65.92,
+      "learning_rate": 0.00017875,
+      "loss": 0.2917,
+      "step": 7185
+    },
+    {
+      "epoch": 65.93,
+      "learning_rate": 0.00017870192307692308,
+      "loss": 0.211,
+      "step": 7186
+    },
+    {
+      "epoch": 65.94,
+      "learning_rate": 0.00017865384615384615,
+      "loss": 0.2497,
+      "step": 7187
+    },
+    {
+      "epoch": 65.94,
+      "learning_rate": 0.00017860576923076923,
+      "loss": 0.1224,
+      "step": 7188
+    },
+    {
+      "epoch": 65.95,
+      "learning_rate": 0.0001785576923076923,
+      "loss": 0.3071,
+      "step": 7189
+    },
+    {
+      "epoch": 65.96,
+      "learning_rate": 0.00017850961538461537,
+      "loss": 0.1634,
+      "step": 7190
+    },
+    {
+      "epoch": 65.97,
+      "learning_rate": 0.00017846153846153844,
+      "loss": 0.2641,
+      "step": 7191
+    },
+    {
+      "epoch": 65.98,
+      "learning_rate": 0.00017841346153846154,
+      "loss": 0.2445,
+      "step": 7192
+    },
+    {
+      "epoch": 65.99,
+      "learning_rate": 0.00017836538461538462,
+      "loss": 0.2283,
+      "step": 7193
+    },
+    {
+      "epoch": 66.0,
+      "learning_rate": 0.0001783173076923077,
+      "loss": 0.2444,
+      "step": 7194
+    },
+    {
+      "epoch": 66.01,
+      "learning_rate": 0.00017826923076923076,
+      "loss": 0.3452,
+      "step": 7195
+    },
+    {
+      "epoch": 66.02,
+      "learning_rate": 0.00017822115384615383,
+      "loss": 0.3098,
+      "step": 7196
+    },
+    {
+      "epoch": 66.03,
+      "learning_rate": 0.00017817307692307693,
+      "loss": 0.2573,
+      "step": 7197
+    },
+    {
+      "epoch": 66.04,
+      "learning_rate": 0.000178125,
+      "loss": 0.2176,
+      "step": 7198
+    },
+    {
+      "epoch": 66.05,
+      "learning_rate": 0.00017807692307692308,
+      "loss": 0.2844,
+      "step": 7199
+    },
+    {
+      "epoch": 66.06,
+      "learning_rate": 0.00017802884615384615,
+      "loss": 0.2965,
+      "step": 7200
+    },
+    {
+      "epoch": 66.06,
+      "learning_rate": 0.00017798076923076922,
+      "loss": 0.1708,
+      "step": 7201
+    },
+    {
+      "epoch": 66.07,
+      "learning_rate": 0.00017793269230769232,
+      "loss": 0.2043,
+      "step": 7202
+    },
+    {
+      "epoch": 66.08,
+      "learning_rate": 0.0001778846153846154,
+      "loss": 0.2176,
+      "step": 7203
+    },
+    {
+      "epoch": 66.09,
+      "learning_rate": 0.00017783653846153847,
+      "loss": 0.2131,
+      "step": 7204
+    },
+    {
+      "epoch": 66.1,
+      "learning_rate": 0.00017778846153846154,
+      "loss": 0.2523,
+      "step": 7205
+    },
+    {
+      "epoch": 66.11,
+      "learning_rate": 0.00017774038461538461,
+      "loss": 0.2119,
+      "step": 7206
+    },
+    {
+      "epoch": 66.12,
+      "learning_rate": 0.00017769230769230771,
+      "loss": 0.2171,
+      "step": 7207
+    },
+    {
+      "epoch": 66.13,
+      "learning_rate": 0.0001776442307692308,
+      "loss": 0.2395,
+      "step": 7208
+    },
+    {
+      "epoch": 66.14,
+      "learning_rate": 0.00017759615384615386,
+      "loss": 0.2118,
+      "step": 7209
+    },
+    {
+      "epoch": 66.15,
+      "learning_rate": 0.00017754807692307693,
+      "loss": 0.2515,
+      "step": 7210
+    },
+    {
+      "epoch": 66.16,
+      "learning_rate": 0.0001775,
+      "loss": 0.1586,
+      "step": 7211
+    },
+    {
+      "epoch": 66.17,
+      "learning_rate": 0.0001774519230769231,
+      "loss": 0.3099,
+      "step": 7212
+    },
+    {
+      "epoch": 66.17,
+      "learning_rate": 0.00017740384615384618,
+      "loss": 0.2718,
+      "step": 7213
+    },
+    {
+      "epoch": 66.18,
+      "learning_rate": 0.00017735576923076925,
+      "loss": 0.2442,
+      "step": 7214
+    },
+    {
+      "epoch": 66.19,
+      "learning_rate": 0.0001773076923076923,
+      "loss": 0.2002,
+      "step": 7215
+    },
+    {
+      "epoch": 66.2,
+      "learning_rate": 0.00017725961538461537,
+      "loss": 0.2074,
+      "step": 7216
+    },
+    {
+      "epoch": 66.21,
+      "learning_rate": 0.00017721153846153844,
+      "loss": 0.1831,
+      "step": 7217
+    },
+    {
+      "epoch": 66.22,
+      "learning_rate": 0.00017716346153846154,
+      "loss": 0.1833,
+      "step": 7218
+    },
+    {
+      "epoch": 66.23,
+      "learning_rate": 0.0001771153846153846,
+      "loss": 0.2038,
+      "step": 7219
+    },
+    {
+      "epoch": 66.24,
+      "learning_rate": 0.00017706730769230769,
+      "loss": 0.2229,
+      "step": 7220
+    },
+    {
+      "epoch": 66.25,
+      "learning_rate": 0.00017701923076923076,
+      "loss": 0.1112,
+      "step": 7221
+    },
+    {
+      "epoch": 66.26,
+      "learning_rate": 0.00017697115384615383,
+      "loss": 0.4488,
+      "step": 7222
+    },
+    {
+      "epoch": 66.27,
+      "learning_rate": 0.00017692307692307693,
+      "loss": 0.2798,
+      "step": 7223
+    },
+    {
+      "epoch": 66.28,
+      "learning_rate": 0.000176875,
+      "loss": 0.3139,
+      "step": 7224
+    },
+    {
+      "epoch": 66.28,
+      "learning_rate": 0.00017682692307692308,
+      "loss": 0.2222,
+      "step": 7225
+    },
+    {
+      "epoch": 66.29,
+      "learning_rate": 0.00017677884615384615,
+      "loss": 0.2098,
+      "step": 7226
+    },
+    {
+      "epoch": 66.3,
+      "learning_rate": 0.00017673076923076922,
+      "loss": 0.1918,
+      "step": 7227
+    },
+    {
+      "epoch": 66.31,
+      "learning_rate": 0.00017668269230769232,
+      "loss": 0.2365,
+      "step": 7228
+    },
+    {
+      "epoch": 66.32,
+      "learning_rate": 0.0001766346153846154,
+      "loss": 0.301,
+      "step": 7229
+    },
+    {
+      "epoch": 66.33,
+      "learning_rate": 0.00017658653846153847,
+      "loss": 0.294,
+      "step": 7230
+    },
+    {
+      "epoch": 66.34,
+      "learning_rate": 0.00017653846153846154,
+      "loss": 0.1816,
+      "step": 7231
+    },
+    {
+      "epoch": 66.35,
+      "learning_rate": 0.0001764903846153846,
+      "loss": 0.3189,
+      "step": 7232
+    },
+    {
+      "epoch": 66.36,
+      "learning_rate": 0.0001764423076923077,
+      "loss": 0.2294,
+      "step": 7233
+    },
+    {
+      "epoch": 66.37,
+      "learning_rate": 0.00017639423076923078,
+      "loss": 0.2036,
+      "step": 7234
+    },
+    {
+      "epoch": 66.38,
+      "learning_rate": 0.00017634615384615386,
+      "loss": 0.284,
+      "step": 7235
+    },
+    {
+      "epoch": 66.39,
+      "learning_rate": 0.00017629807692307693,
+      "loss": 0.269,
+      "step": 7236
+    },
+    {
+      "epoch": 66.39,
+      "learning_rate": 0.00017625,
+      "loss": 0.295,
+      "step": 7237
+    },
+    {
+      "epoch": 66.4,
+      "learning_rate": 0.0001762019230769231,
+      "loss": 0.2151,
+      "step": 7238
+    },
+    {
+      "epoch": 66.41,
+      "learning_rate": 0.00017615384615384617,
+      "loss": 0.2538,
+      "step": 7239
+    },
+    {
+      "epoch": 66.42,
+      "learning_rate": 0.00017610576923076925,
+      "loss": 0.1879,
+      "step": 7240
+    },
+    {
+      "epoch": 66.43,
+      "learning_rate": 0.00017605769230769232,
+      "loss": 0.2896,
+      "step": 7241
+    },
+    {
+      "epoch": 66.44,
+      "learning_rate": 0.0001760096153846154,
+      "loss": 0.1841,
+      "step": 7242
+    },
+    {
+      "epoch": 66.45,
+      "learning_rate": 0.0001759615384615385,
+      "loss": 0.2628,
+      "step": 7243
+    },
+    {
+      "epoch": 66.46,
+      "learning_rate": 0.00017591346153846154,
+      "loss": 0.1832,
+      "step": 7244
+    },
+    {
+      "epoch": 66.47,
+      "learning_rate": 0.0001758653846153846,
+      "loss": 0.1762,
+      "step": 7245
+    },
+    {
+      "epoch": 66.48,
+      "learning_rate": 0.00017581730769230768,
+      "loss": 0.1912,
+      "step": 7246
+    },
+    {
+      "epoch": 66.49,
+      "learning_rate": 0.00017576923076923075,
+      "loss": 0.2378,
+      "step": 7247
+    },
+    {
+      "epoch": 66.5,
+      "learning_rate": 0.00017572115384615383,
+      "loss": 0.2458,
+      "step": 7248
+    },
+    {
+      "epoch": 66.5,
+      "learning_rate": 0.00017567307692307693,
+      "loss": 0.3179,
+      "step": 7249
+    },
+    {
+      "epoch": 66.51,
+      "learning_rate": 0.000175625,
+      "loss": 0.3216,
+      "step": 7250
+    },
+    {
+      "epoch": 66.52,
+      "learning_rate": 0.00017557692307692307,
+      "loss": 0.2762,
+      "step": 7251
+    },
+    {
+      "epoch": 66.53,
+      "learning_rate": 0.00017552884615384615,
+      "loss": 0.2289,
+      "step": 7252
+    },
+    {
+      "epoch": 66.54,
+      "learning_rate": 0.00017548076923076922,
+      "loss": 0.2465,
+      "step": 7253
+    },
+    {
+      "epoch": 66.55,
+      "learning_rate": 0.00017543269230769232,
+      "loss": 0.2564,
+      "step": 7254
+    },
+    {
+      "epoch": 66.56,
+      "learning_rate": 0.0001753846153846154,
+      "loss": 0.3082,
+      "step": 7255
+    },
+    {
+      "epoch": 66.57,
+      "learning_rate": 0.00017533653846153846,
+      "loss": 0.2099,
+      "step": 7256
+    },
+    {
+      "epoch": 66.58,
+      "learning_rate": 0.00017528846153846154,
+      "loss": 0.236,
+      "step": 7257
+    },
+    {
+      "epoch": 66.59,
+      "learning_rate": 0.0001752403846153846,
+      "loss": 0.235,
+      "step": 7258
+    },
+    {
+      "epoch": 66.6,
+      "learning_rate": 0.0001751923076923077,
+      "loss": 0.2923,
+      "step": 7259
+    },
+    {
+      "epoch": 66.61,
+      "learning_rate": 0.00017514423076923078,
+      "loss": 0.2904,
+      "step": 7260
+    },
+    {
+      "epoch": 66.61,
+      "learning_rate": 0.00017509615384615385,
+      "loss": 0.3037,
+      "step": 7261
+    },
+    {
+      "epoch": 66.62,
+      "learning_rate": 0.00017504807692307693,
+      "loss": 0.2981,
+      "step": 7262
+    },
+    {
+      "epoch": 66.63,
+      "learning_rate": 0.000175,
+      "loss": 0.236,
+      "step": 7263
+    },
+    {
+      "epoch": 66.64,
+      "learning_rate": 0.0001749519230769231,
+      "loss": 0.3069,
+      "step": 7264
+    },
+    {
+      "epoch": 66.65,
+      "learning_rate": 0.00017490384615384617,
+      "loss": 0.2671,
+      "step": 7265
+    },
+    {
+      "epoch": 66.66,
+      "learning_rate": 0.00017485576923076924,
+      "loss": 0.1506,
+      "step": 7266
+    },
+    {
+      "epoch": 66.67,
+      "learning_rate": 0.00017480769230769232,
+      "loss": 0.2276,
+      "step": 7267
+    },
+    {
+      "epoch": 66.68,
+      "learning_rate": 0.0001747596153846154,
+      "loss": 0.2376,
+      "step": 7268
+    },
+    {
+      "epoch": 66.69,
+      "learning_rate": 0.0001747115384615385,
+      "loss": 0.1813,
+      "step": 7269
+    },
+    {
+      "epoch": 66.7,
+      "learning_rate": 0.00017466346153846156,
+      "loss": 0.1788,
+      "step": 7270
+    },
+    {
+      "epoch": 66.71,
+      "learning_rate": 0.00017461538461538463,
+      "loss": 0.175,
+      "step": 7271
+    },
+    {
+      "epoch": 66.72,
+      "learning_rate": 0.00017456730769230768,
+      "loss": 0.1894,
+      "step": 7272
+    },
+    {
+      "epoch": 66.72,
+      "learning_rate": 0.00017451923076923075,
+      "loss": 0.3924,
+      "step": 7273
+    },
+    {
+      "epoch": 66.73,
+      "learning_rate": 0.00017447115384615385,
+      "loss": 0.2149,
+      "step": 7274
+    },
+    {
+      "epoch": 66.74,
+      "learning_rate": 0.00017442307692307692,
+      "loss": 0.2656,
+      "step": 7275
+    },
+    {
+      "epoch": 66.75,
+      "learning_rate": 0.000174375,
+      "loss": 0.3688,
+      "step": 7276
+    },
+    {
+      "epoch": 66.76,
+      "learning_rate": 0.00017432692307692307,
+      "loss": 0.2509,
+      "step": 7277
+    },
+    {
+      "epoch": 66.77,
+      "learning_rate": 0.00017427884615384614,
+      "loss": 0.3503,
+      "step": 7278
+    },
+    {
+      "epoch": 66.78,
+      "learning_rate": 0.00017423076923076924,
+      "loss": 0.3337,
+      "step": 7279
+    },
+    {
+      "epoch": 66.79,
+      "learning_rate": 0.00017418269230769231,
+      "loss": 0.305,
+      "step": 7280
+    },
+    {
+      "epoch": 66.8,
+      "learning_rate": 0.0001741346153846154,
+      "loss": 0.2597,
+      "step": 7281
+    },
+    {
+      "epoch": 66.81,
+      "learning_rate": 0.00017408653846153846,
+      "loss": 0.2858,
+      "step": 7282
+    },
+    {
+      "epoch": 66.82,
+      "learning_rate": 0.00017403846153846153,
+      "loss": 0.272,
+      "step": 7283
+    },
+    {
+      "epoch": 66.83,
+      "learning_rate": 0.0001739903846153846,
+      "loss": 0.2997,
+      "step": 7284
+    },
+    {
+      "epoch": 66.83,
+      "learning_rate": 0.0001739423076923077,
+      "loss": 0.1704,
+      "step": 7285
+    },
+    {
+      "epoch": 66.84,
+      "learning_rate": 0.00017389423076923078,
+      "loss": 0.2238,
+      "step": 7286
+    },
+    {
+      "epoch": 66.85,
+      "learning_rate": 0.00017384615384615385,
+      "loss": 0.1724,
+      "step": 7287
+    },
+    {
+      "epoch": 66.86,
+      "learning_rate": 0.00017379807692307692,
+      "loss": 0.2856,
+      "step": 7288
+    },
+    {
+      "epoch": 66.87,
+      "learning_rate": 0.00017375,
+      "loss": 0.3091,
+      "step": 7289
+    },
+    {
+      "epoch": 66.88,
+      "learning_rate": 0.0001737019230769231,
+      "loss": 0.2748,
+      "step": 7290
+    },
+    {
+      "epoch": 66.89,
+      "learning_rate": 0.00017365384615384617,
+      "loss": 0.225,
+      "step": 7291
+    },
+    {
+      "epoch": 66.9,
+      "learning_rate": 0.00017360576923076924,
+      "loss": 0.279,
+      "step": 7292
+    },
+    {
+      "epoch": 66.91,
+      "learning_rate": 0.0001735576923076923,
+      "loss": 0.2479,
+      "step": 7293
+    },
+    {
+      "epoch": 66.92,
+      "learning_rate": 0.00017350961538461539,
+      "loss": 0.2075,
+      "step": 7294
+    },
+    {
+      "epoch": 66.93,
+      "learning_rate": 0.00017346153846153848,
+      "loss": 0.2769,
+      "step": 7295
+    },
+    {
+      "epoch": 66.94,
+      "learning_rate": 0.00017341346153846156,
+      "loss": 0.2444,
+      "step": 7296
+    },
+    {
+      "epoch": 66.94,
+      "learning_rate": 0.00017336538461538463,
+      "loss": 0.2904,
+      "step": 7297
+    },
+    {
+      "epoch": 66.95,
+      "learning_rate": 0.0001733173076923077,
+      "loss": 0.1634,
+      "step": 7298
+    },
+    {
+      "epoch": 66.96,
+      "learning_rate": 0.00017326923076923078,
+      "loss": 0.3005,
+      "step": 7299
+    },
+    {
+      "epoch": 66.97,
+      "learning_rate": 0.00017322115384615388,
+      "loss": 0.264,
+      "step": 7300
+    },
+    {
+      "epoch": 66.98,
+      "learning_rate": 0.00017317307692307692,
+      "loss": 0.282,
+      "step": 7301
+    },
+    {
+      "epoch": 66.99,
+      "learning_rate": 0.000173125,
+      "loss": 0.1414,
+      "step": 7302
+    },
+    {
+      "epoch": 67.0,
+      "learning_rate": 0.00017307692307692307,
+      "loss": 0.2874,
+      "step": 7303
+    },
+    {
+      "epoch": 67.01,
+      "learning_rate": 0.00017302884615384614,
+      "loss": 0.3514,
+      "step": 7304
+    },
+    {
+      "epoch": 67.02,
+      "learning_rate": 0.00017298076923076924,
+      "loss": 0.2505,
+      "step": 7305
+    },
+    {
+      "epoch": 67.03,
+      "learning_rate": 0.0001729326923076923,
+      "loss": 0.3137,
+      "step": 7306
+    },
+    {
+      "epoch": 67.04,
+      "learning_rate": 0.00017288461538461538,
+      "loss": 0.233,
+      "step": 7307
+    },
+    {
+      "epoch": 67.05,
+      "learning_rate": 0.00017283653846153846,
+      "loss": 0.2335,
+      "step": 7308
+    },
+    {
+      "epoch": 67.06,
+      "learning_rate": 0.00017278846153846153,
+      "loss": 0.2528,
+      "step": 7309
+    },
+    {
+      "epoch": 67.06,
+      "learning_rate": 0.00017274038461538463,
+      "loss": 0.235,
+      "step": 7310
+    },
+    {
+      "epoch": 67.07,
+      "learning_rate": 0.0001726923076923077,
+      "loss": 0.2267,
+      "step": 7311
+    },
+    {
+      "epoch": 67.08,
+      "learning_rate": 0.00017264423076923077,
+      "loss": 0.2453,
+      "step": 7312
+    },
+    {
+      "epoch": 67.09,
+      "learning_rate": 0.00017259615384615385,
+      "loss": 0.1494,
+      "step": 7313
+    },
+    {
+      "epoch": 67.1,
+      "learning_rate": 0.00017254807692307692,
+      "loss": 0.2552,
+      "step": 7314
+    },
+    {
+      "epoch": 67.11,
+      "learning_rate": 0.0001725,
+      "loss": 0.197,
+      "step": 7315
+    },
+    {
+      "epoch": 67.12,
+      "learning_rate": 0.0001724519230769231,
+      "loss": 0.2531,
+      "step": 7316
+    },
+    {
+      "epoch": 67.13,
+      "learning_rate": 0.00017240384615384616,
+      "loss": 0.2394,
+      "step": 7317
+    },
+    {
+      "epoch": 67.14,
+      "learning_rate": 0.00017235576923076924,
+      "loss": 0.2518,
+      "step": 7318
+    },
+    {
+      "epoch": 67.15,
+      "learning_rate": 0.0001723076923076923,
+      "loss": 0.2351,
+      "step": 7319
+    },
+    {
+      "epoch": 67.16,
+      "learning_rate": 0.00017225961538461538,
+      "loss": 0.1955,
+      "step": 7320
+    },
+    {
+      "epoch": 67.17,
+      "learning_rate": 0.00017221153846153848,
+      "loss": 0.2446,
+      "step": 7321
+    },
+    {
+      "epoch": 67.17,
+      "learning_rate": 0.00017216346153846155,
+      "loss": 0.2008,
+      "step": 7322
+    },
+    {
+      "epoch": 67.18,
+      "learning_rate": 0.00017211538461538463,
+      "loss": 0.1766,
+      "step": 7323
+    },
+    {
+      "epoch": 67.19,
+      "learning_rate": 0.0001720673076923077,
+      "loss": 0.1913,
+      "step": 7324
+    },
+    {
+      "epoch": 67.2,
+      "learning_rate": 0.00017201923076923077,
+      "loss": 0.1496,
+      "step": 7325
+    },
+    {
+      "epoch": 67.21,
+      "learning_rate": 0.00017197115384615387,
+      "loss": 0.3398,
+      "step": 7326
+    },
+    {
+      "epoch": 67.22,
+      "learning_rate": 0.00017192307692307694,
+      "loss": 0.2293,
+      "step": 7327
+    },
+    {
+      "epoch": 67.23,
+      "learning_rate": 0.000171875,
+      "loss": 0.2997,
+      "step": 7328
+    },
+    {
+      "epoch": 67.24,
+      "learning_rate": 0.00017182692307692306,
+      "loss": 0.2092,
+      "step": 7329
+    },
+    {
+      "epoch": 67.25,
+      "learning_rate": 0.00017177884615384614,
+      "loss": 0.2068,
+      "step": 7330
+    },
+    {
+      "epoch": 67.26,
+      "learning_rate": 0.00017173076923076924,
+      "loss": 0.321,
+      "step": 7331
+    },
+    {
+      "epoch": 67.27,
+      "learning_rate": 0.0001716826923076923,
+      "loss": 0.2827,
+      "step": 7332
+    },
+    {
+      "epoch": 67.28,
+      "learning_rate": 0.00017163461538461538,
+      "loss": 0.1984,
+      "step": 7333
+    },
+    {
+      "epoch": 67.28,
+      "learning_rate": 0.00017158653846153845,
+      "loss": 0.2088,
+      "step": 7334
+    },
+    {
+      "epoch": 67.29,
+      "learning_rate": 0.00017153846153846153,
+      "loss": 0.2765,
+      "step": 7335
+    },
+    {
+      "epoch": 67.3,
+      "learning_rate": 0.00017149038461538463,
+      "loss": 0.2453,
+      "step": 7336
+    },
+    {
+      "epoch": 67.31,
+      "learning_rate": 0.0001714423076923077,
+      "loss": 0.2065,
+      "step": 7337
+    },
+    {
+      "epoch": 67.32,
+      "learning_rate": 0.00017139423076923077,
+      "loss": 0.2873,
+      "step": 7338
+    },
+    {
+      "epoch": 67.33,
+      "learning_rate": 0.00017134615384615384,
+      "loss": 0.2043,
+      "step": 7339
+    },
+    {
+      "epoch": 67.34,
+      "learning_rate": 0.00017129807692307692,
+      "loss": 0.2573,
+      "step": 7340
+    },
+    {
+      "epoch": 67.35,
+      "learning_rate": 0.00017125000000000002,
+      "loss": 0.2173,
+      "step": 7341
+    },
+    {
+      "epoch": 67.36,
+      "learning_rate": 0.0001712019230769231,
+      "loss": 0.2679,
+      "step": 7342
+    },
+    {
+      "epoch": 67.37,
+      "learning_rate": 0.00017115384615384616,
+      "loss": 0.2479,
+      "step": 7343
+    },
+    {
+      "epoch": 67.38,
+      "learning_rate": 0.00017110576923076923,
+      "loss": 0.2883,
+      "step": 7344
+    },
+    {
+      "epoch": 67.39,
+      "learning_rate": 0.0001710576923076923,
+      "loss": 0.245,
+      "step": 7345
+    },
+    {
+      "epoch": 67.39,
+      "learning_rate": 0.00017100961538461538,
+      "loss": 0.3449,
+      "step": 7346
+    },
+    {
+      "epoch": 67.4,
+      "learning_rate": 0.00017096153846153848,
+      "loss": 0.1922,
+      "step": 7347
+    },
+    {
+      "epoch": 67.41,
+      "learning_rate": 0.00017091346153846155,
+      "loss": 0.2385,
+      "step": 7348
+    },
+    {
+      "epoch": 67.42,
+      "learning_rate": 0.00017086538461538462,
+      "loss": 0.2076,
+      "step": 7349
+    },
+    {
+      "epoch": 67.43,
+      "learning_rate": 0.0001708173076923077,
+      "loss": 0.2795,
+      "step": 7350
+    },
+    {
+      "epoch": 67.44,
+      "learning_rate": 0.00017076923076923077,
+      "loss": 0.2793,
+      "step": 7351
+    },
+    {
+      "epoch": 67.45,
+      "learning_rate": 0.00017072115384615387,
+      "loss": 0.2341,
+      "step": 7352
+    },
+    {
+      "epoch": 67.46,
+      "learning_rate": 0.00017067307692307694,
+      "loss": 0.3357,
+      "step": 7353
+    },
+    {
+      "epoch": 67.47,
+      "learning_rate": 0.00017062500000000001,
+      "loss": 0.1507,
+      "step": 7354
+    },
+    {
+      "epoch": 67.48,
+      "learning_rate": 0.0001705769230769231,
+      "loss": 0.2114,
+      "step": 7355
+    },
+    {
+      "epoch": 67.49,
+      "learning_rate": 0.00017052884615384613,
+      "loss": 0.2429,
+      "step": 7356
+    },
+    {
+      "epoch": 67.5,
+      "learning_rate": 0.00017048076923076923,
+      "loss": 0.17,
+      "step": 7357
+    },
+    {
+      "epoch": 67.5,
+      "learning_rate": 0.0001704326923076923,
+      "loss": 0.2445,
+      "step": 7358
+    },
+    {
+      "epoch": 67.51,
+      "learning_rate": 0.00017038461538461538,
+      "loss": 0.2349,
+      "step": 7359
+    },
+    {
+      "epoch": 67.52,
+      "learning_rate": 0.00017033653846153845,
+      "loss": 0.3026,
+      "step": 7360
+    },
+    {
+      "epoch": 67.53,
+      "learning_rate": 0.00017028846153846152,
+      "loss": 0.3006,
+      "step": 7361
+    },
+    {
+      "epoch": 67.54,
+      "learning_rate": 0.00017024038461538462,
+      "loss": 0.2731,
+      "step": 7362
+    },
+    {
+      "epoch": 67.55,
+      "learning_rate": 0.0001701923076923077,
+      "loss": 0.2873,
+      "step": 7363
+    },
+    {
+      "epoch": 67.56,
+      "learning_rate": 0.00017014423076923077,
+      "loss": 0.3738,
+      "step": 7364
+    },
+    {
+      "epoch": 67.57,
+      "learning_rate": 0.00017009615384615384,
+      "loss": 0.2346,
+      "step": 7365
+    },
+    {
+      "epoch": 67.58,
+      "learning_rate": 0.0001700480769230769,
+      "loss": 0.2283,
+      "step": 7366
+    },
+    {
+      "epoch": 67.59,
+      "learning_rate": 0.00017,
+      "loss": 0.1507,
+      "step": 7367
+    },
+    {
+      "epoch": 67.6,
+      "learning_rate": 0.00016995192307692308,
+      "loss": 0.2552,
+      "step": 7368
+    },
+    {
+      "epoch": 67.61,
+      "learning_rate": 0.00016990384615384616,
+      "loss": 0.1812,
+      "step": 7369
+    },
+    {
+      "epoch": 67.61,
+      "learning_rate": 0.00016985576923076923,
+      "loss": 0.2623,
+      "step": 7370
+    },
+    {
+      "epoch": 67.62,
+      "learning_rate": 0.0001698076923076923,
+      "loss": 0.1731,
+      "step": 7371
+    },
+    {
+      "epoch": 67.63,
+      "learning_rate": 0.0001697596153846154,
+      "loss": 0.2203,
+      "step": 7372
+    },
+    {
+      "epoch": 67.64,
+      "learning_rate": 0.00016971153846153848,
+      "loss": 0.1951,
+      "step": 7373
+    },
+    {
+      "epoch": 67.65,
+      "learning_rate": 0.00016966346153846155,
+      "loss": 0.2612,
+      "step": 7374
+    },
+    {
+      "epoch": 67.66,
+      "learning_rate": 0.00016961538461538462,
+      "loss": 0.2232,
+      "step": 7375
+    },
+    {
+      "epoch": 67.67,
+      "learning_rate": 0.0001695673076923077,
+      "loss": 0.2078,
+      "step": 7376
+    },
+    {
+      "epoch": 67.68,
+      "learning_rate": 0.00016951923076923077,
+      "loss": 0.1552,
+      "step": 7377
+    },
+    {
+      "epoch": 67.69,
+      "learning_rate": 0.00016947115384615387,
+      "loss": 0.209,
+      "step": 7378
+    },
+    {
+      "epoch": 67.7,
+      "learning_rate": 0.00016942307692307694,
+      "loss": 0.2386,
+      "step": 7379
+    },
+    {
+      "epoch": 67.71,
+      "learning_rate": 0.000169375,
+      "loss": 0.2162,
+      "step": 7380
+    },
+    {
+      "epoch": 67.72,
+      "learning_rate": 0.00016932692307692308,
+      "loss": 0.3015,
+      "step": 7381
+    },
+    {
+      "epoch": 67.72,
+      "learning_rate": 0.00016927884615384616,
+      "loss": 0.304,
+      "step": 7382
+    },
+    {
+      "epoch": 67.73,
+      "learning_rate": 0.00016923076923076926,
+      "loss": 0.2543,
+      "step": 7383
+    },
+    {
+      "epoch": 67.74,
+      "learning_rate": 0.00016918269230769233,
+      "loss": 0.1715,
+      "step": 7384
+    },
+    {
+      "epoch": 67.75,
+      "learning_rate": 0.00016913461538461537,
+      "loss": 0.3025,
+      "step": 7385
+    },
+    {
+      "epoch": 67.76,
+      "learning_rate": 0.00016908653846153845,
+      "loss": 0.2958,
+      "step": 7386
+    },
+    {
+      "epoch": 67.77,
+      "learning_rate": 0.00016903846153846152,
+      "loss": 0.2663,
+      "step": 7387
+    },
+    {
+      "epoch": 67.78,
+      "learning_rate": 0.00016899038461538462,
+      "loss": 0.235,
+      "step": 7388
+    },
+    {
+      "epoch": 67.79,
+      "learning_rate": 0.0001689423076923077,
+      "loss": 0.2888,
+      "step": 7389
+    },
+    {
+      "epoch": 67.8,
+      "learning_rate": 0.00016889423076923076,
+      "loss": 0.3345,
+      "step": 7390
+    },
+    {
+      "epoch": 67.81,
+      "learning_rate": 0.00016884615384615384,
+      "loss": 0.2856,
+      "step": 7391
+    },
+    {
+      "epoch": 67.82,
+      "learning_rate": 0.0001687980769230769,
+      "loss": 0.2156,
+      "step": 7392
+    },
+    {
+      "epoch": 67.83,
+      "learning_rate": 0.00016875,
+      "loss": 0.3029,
+      "step": 7393
+    },
+    {
+      "epoch": 67.83,
+      "learning_rate": 0.00016870192307692308,
+      "loss": 0.246,
+      "step": 7394
+    },
+    {
+      "epoch": 67.84,
+      "learning_rate": 0.00016865384615384615,
+      "loss": 0.244,
+      "step": 7395
+    },
+    {
+      "epoch": 67.85,
+      "learning_rate": 0.00016860576923076923,
+      "loss": 0.2619,
+      "step": 7396
+    },
+    {
+      "epoch": 67.86,
+      "learning_rate": 0.0001685576923076923,
+      "loss": 0.2561,
+      "step": 7397
+    },
+    {
+      "epoch": 67.87,
+      "learning_rate": 0.0001685096153846154,
+      "loss": 0.193,
+      "step": 7398
+    },
+    {
+      "epoch": 67.88,
+      "learning_rate": 0.0001685096153846154,
+      "loss": 0.2211,
+      "step": 7399
+    },
+    {
+      "epoch": 67.89,
+      "learning_rate": 0.00016846153846153847,
+      "loss": 0.2402,
+      "step": 7400
+    },
+    {
+      "epoch": 67.9,
+      "learning_rate": 0.00016841346153846154,
+      "loss": 0.1852,
+      "step": 7401
+    },
+    {
+      "epoch": 67.91,
+      "learning_rate": 0.00016836538461538462,
+      "loss": 0.2218,
+      "step": 7402
+    },
+    {
+      "epoch": 67.92,
+      "learning_rate": 0.0001683173076923077,
+      "loss": 0.1801,
+      "step": 7403
+    },
+    {
+      "epoch": 67.93,
+      "learning_rate": 0.0001682692307692308,
+      "loss": 0.2047,
+      "step": 7404
+    },
+    {
+      "epoch": 67.94,
+      "learning_rate": 0.00016822115384615386,
+      "loss": 0.2721,
+      "step": 7405
+    },
+    {
+      "epoch": 67.94,
+      "learning_rate": 0.00016817307692307693,
+      "loss": 0.2922,
+      "step": 7406
+    },
+    {
+      "epoch": 67.95,
+      "learning_rate": 0.000168125,
+      "loss": 0.2386,
+      "step": 7407
+    },
+    {
+      "epoch": 67.96,
+      "learning_rate": 0.00016807692307692308,
+      "loss": 0.2475,
+      "step": 7408
+    },
+    {
+      "epoch": 67.97,
+      "learning_rate": 0.00016802884615384618,
+      "loss": 0.2289,
+      "step": 7409
+    },
+    {
+      "epoch": 67.98,
+      "learning_rate": 0.00016798076923076925,
+      "loss": 0.2758,
+      "step": 7410
+    },
+    {
+      "epoch": 67.99,
+      "learning_rate": 0.00016793269230769232,
+      "loss": 0.1382,
+      "step": 7411
+    },
+    {
+      "epoch": 68.0,
+      "learning_rate": 0.0001678846153846154,
+      "loss": 0.2263,
+      "step": 7412
+    },
+    {
+      "epoch": 68.01,
+      "learning_rate": 0.00016783653846153847,
+      "loss": 0.2883,
+      "step": 7413
+    },
+    {
+      "epoch": 68.02,
+      "learning_rate": 0.00016778846153846152,
+      "loss": 0.1926,
+      "step": 7414
+    },
+    {
+      "epoch": 68.03,
+      "learning_rate": 0.00016774038461538462,
+      "loss": 0.3086,
+      "step": 7415
+    },
+    {
+      "epoch": 68.04,
+      "learning_rate": 0.0001676923076923077,
+      "loss": 0.2992,
+      "step": 7416
+    },
+    {
+      "epoch": 68.05,
+      "learning_rate": 0.00016764423076923076,
+      "loss": 0.2899,
+      "step": 7417
+    },
+    {
+      "epoch": 68.06,
+      "learning_rate": 0.00016759615384615383,
+      "loss": 0.2517,
+      "step": 7418
+    },
+    {
+      "epoch": 68.06,
+      "learning_rate": 0.0001675480769230769,
+      "loss": 0.253,
+      "step": 7419
+    },
+    {
+      "epoch": 68.07,
+      "learning_rate": 0.0001675,
+      "loss": 0.2373,
+      "step": 7420
+    },
+    {
+      "epoch": 68.08,
+      "learning_rate": 0.00016745192307692308,
+      "loss": 0.1971,
+      "step": 7421
+    },
+    {
+      "epoch": 68.09,
+      "learning_rate": 0.00016740384615384615,
+      "loss": 0.1721,
+      "step": 7422
+    },
+    {
+      "epoch": 68.1,
+      "learning_rate": 0.00016735576923076922,
+      "loss": 0.2733,
+      "step": 7423
+    },
+    {
+      "epoch": 68.11,
+      "learning_rate": 0.0001673076923076923,
+      "loss": 0.1774,
+      "step": 7424
+    },
+    {
+      "epoch": 68.12,
+      "learning_rate": 0.0001672596153846154,
+      "loss": 0.2363,
+      "step": 7425
+    },
+    {
+      "epoch": 68.13,
+      "learning_rate": 0.00016721153846153847,
+      "loss": 0.1838,
+      "step": 7426
+    },
+    {
+      "epoch": 68.14,
+      "learning_rate": 0.00016716346153846154,
+      "loss": 0.2123,
+      "step": 7427
+    },
+    {
+      "epoch": 68.15,
+      "learning_rate": 0.00016711538461538461,
+      "loss": 0.279,
+      "step": 7428
+    },
+    {
+      "epoch": 68.16,
+      "learning_rate": 0.00016706730769230769,
+      "loss": 0.1906,
+      "step": 7429
+    },
+    {
+      "epoch": 68.17,
+      "learning_rate": 0.00016701923076923079,
+      "loss": 0.2189,
+      "step": 7430
+    },
+    {
+      "epoch": 68.17,
+      "learning_rate": 0.00016697115384615386,
+      "loss": 0.2638,
+      "step": 7431
+    },
+    {
+      "epoch": 68.18,
+      "learning_rate": 0.00016692307692307693,
+      "loss": 0.2359,
+      "step": 7432
+    },
+    {
+      "epoch": 68.19,
+      "learning_rate": 0.000166875,
+      "loss": 0.2295,
+      "step": 7433
+    },
+    {
+      "epoch": 68.2,
+      "learning_rate": 0.00016682692307692308,
+      "loss": 0.2241,
+      "step": 7434
+    },
+    {
+      "epoch": 68.21,
+      "learning_rate": 0.00016677884615384618,
+      "loss": 0.2333,
+      "step": 7435
+    },
+    {
+      "epoch": 68.22,
+      "learning_rate": 0.00016673076923076925,
+      "loss": 0.221,
+      "step": 7436
+    },
+    {
+      "epoch": 68.23,
+      "learning_rate": 0.00016668269230769232,
+      "loss": 0.2014,
+      "step": 7437
+    },
+    {
+      "epoch": 68.24,
+      "learning_rate": 0.0001666346153846154,
+      "loss": 0.1516,
+      "step": 7438
+    },
+    {
+      "epoch": 68.25,
+      "learning_rate": 0.00016658653846153847,
+      "loss": 0.1217,
+      "step": 7439
+    },
+    {
+      "epoch": 68.26,
+      "learning_rate": 0.00016653846153846157,
+      "loss": 0.263,
+      "step": 7440
+    },
+    {
+      "epoch": 68.27,
+      "learning_rate": 0.00016649038461538464,
+      "loss": 0.3714,
+      "step": 7441
+    },
+    {
+      "epoch": 68.28,
+      "learning_rate": 0.0001664423076923077,
+      "loss": 0.2598,
+      "step": 7442
+    },
+    {
+      "epoch": 68.28,
+      "learning_rate": 0.00016639423076923076,
+      "loss": 0.2582,
+      "step": 7443
+    },
+    {
+      "epoch": 68.29,
+      "learning_rate": 0.00016634615384615383,
+      "loss": 0.2192,
+      "step": 7444
+    },
+    {
+      "epoch": 68.3,
+      "learning_rate": 0.0001662980769230769,
+      "loss": 0.2461,
+      "step": 7445
+    },
+    {
+      "epoch": 68.31,
+      "learning_rate": 0.00016625,
+      "loss": 0.1464,
+      "step": 7446
+    },
+    {
+      "epoch": 68.32,
+      "learning_rate": 0.00016620192307692308,
+      "loss": 0.1744,
+      "step": 7447
+    },
+    {
+      "epoch": 68.33,
+      "learning_rate": 0.00016615384615384615,
+      "loss": 0.1806,
+      "step": 7448
+    },
+    {
+      "epoch": 68.34,
+      "learning_rate": 0.00016610576923076922,
+      "loss": 0.1998,
+      "step": 7449
+    },
+    {
+      "epoch": 68.35,
+      "learning_rate": 0.0001660576923076923,
+      "loss": 0.2482,
+      "step": 7450
+    },
+    {
+      "epoch": 68.36,
+      "learning_rate": 0.0001660096153846154,
+      "loss": 0.2354,
+      "step": 7451
+    },
+    {
+      "epoch": 68.37,
+      "learning_rate": 0.00016596153846153847,
+      "loss": 0.2212,
+      "step": 7452
+    },
+    {
+      "epoch": 68.38,
+      "learning_rate": 0.00016591346153846154,
+      "loss": 0.2069,
+      "step": 7453
+    },
+    {
+      "epoch": 68.39,
+      "learning_rate": 0.0001658653846153846,
+      "loss": 0.2928,
+      "step": 7454
+    },
+    {
+      "epoch": 68.39,
+      "learning_rate": 0.00016581730769230768,
+      "loss": 0.2293,
+      "step": 7455
+    },
+    {
+      "epoch": 68.4,
+      "learning_rate": 0.00016576923076923078,
+      "loss": 0.2103,
+      "step": 7456
+    },
+    {
+      "epoch": 68.41,
+      "learning_rate": 0.00016572115384615386,
+      "loss": 0.2638,
+      "step": 7457
+    },
+    {
+      "epoch": 68.42,
+      "learning_rate": 0.00016567307692307693,
+      "loss": 0.3388,
+      "step": 7458
+    },
+    {
+      "epoch": 68.43,
+      "learning_rate": 0.000165625,
+      "loss": 0.1924,
+      "step": 7459
+    },
+    {
+      "epoch": 68.44,
+      "learning_rate": 0.00016557692307692307,
+      "loss": 0.2668,
+      "step": 7460
+    },
+    {
+      "epoch": 68.45,
+      "learning_rate": 0.00016552884615384617,
+      "loss": 0.1945,
+      "step": 7461
+    },
+    {
+      "epoch": 68.46,
+      "learning_rate": 0.00016548076923076925,
+      "loss": 0.2481,
+      "step": 7462
+    },
+    {
+      "epoch": 68.47,
+      "learning_rate": 0.00016543269230769232,
+      "loss": 0.1994,
+      "step": 7463
+    },
+    {
+      "epoch": 68.48,
+      "learning_rate": 0.0001653846153846154,
+      "loss": 0.199,
+      "step": 7464
+    },
+    {
+      "epoch": 68.49,
+      "learning_rate": 0.00016533653846153846,
+      "loss": 0.2074,
+      "step": 7465
+    },
+    {
+      "epoch": 68.5,
+      "learning_rate": 0.00016528846153846156,
+      "loss": 0.1559,
+      "step": 7466
+    },
+    {
+      "epoch": 68.5,
+      "learning_rate": 0.00016524038461538464,
+      "loss": 0.3455,
+      "step": 7467
+    },
+    {
+      "epoch": 68.51,
+      "learning_rate": 0.0001651923076923077,
+      "loss": 0.2569,
+      "step": 7468
+    },
+    {
+      "epoch": 68.52,
+      "learning_rate": 0.00016514423076923078,
+      "loss": 0.2531,
+      "step": 7469
+    },
+    {
+      "epoch": 68.53,
+      "learning_rate": 0.00016509615384615385,
+      "loss": 0.2981,
+      "step": 7470
+    },
+    {
+      "epoch": 68.54,
+      "learning_rate": 0.00016504807692307693,
+      "loss": 0.2623,
+      "step": 7471
+    },
+    {
+      "epoch": 68.55,
+      "learning_rate": 0.000165,
+      "loss": 0.2729,
+      "step": 7472
+    },
+    {
+      "epoch": 68.56,
+      "learning_rate": 0.00016495192307692307,
+      "loss": 0.2047,
+      "step": 7473
+    },
+    {
+      "epoch": 68.57,
+      "learning_rate": 0.00016490384615384614,
+      "loss": 0.2701,
+      "step": 7474
+    },
+    {
+      "epoch": 68.58,
+      "learning_rate": 0.00016485576923076922,
+      "loss": 0.297,
+      "step": 7475
+    },
+    {
+      "epoch": 68.59,
+      "learning_rate": 0.0001648076923076923,
+      "loss": 0.1556,
+      "step": 7476
+    },
+    {
+      "epoch": 68.6,
+      "learning_rate": 0.0001647596153846154,
+      "loss": 0.2735,
+      "step": 7477
+    },
+    {
+      "epoch": 68.61,
+      "learning_rate": 0.00016471153846153846,
+      "loss": 0.2333,
+      "step": 7478
+    },
+    {
+      "epoch": 68.61,
+      "learning_rate": 0.00016466346153846153,
+      "loss": 0.3222,
+      "step": 7479
+    },
+    {
+      "epoch": 68.62,
+      "learning_rate": 0.0001646153846153846,
+      "loss": 0.2495,
+      "step": 7480
+    },
+    {
+      "epoch": 68.63,
+      "learning_rate": 0.00016456730769230768,
+      "loss": 0.2088,
+      "step": 7481
+    },
+    {
+      "epoch": 68.64,
+      "learning_rate": 0.00016451923076923078,
+      "loss": 0.2566,
+      "step": 7482
+    },
+    {
+      "epoch": 68.65,
+      "learning_rate": 0.00016447115384615385,
+      "loss": 0.2514,
+      "step": 7483
+    },
+    {
+      "epoch": 68.66,
+      "learning_rate": 0.00016442307692307692,
+      "loss": 0.2727,
+      "step": 7484
+    },
+    {
+      "epoch": 68.67,
+      "learning_rate": 0.000164375,
+      "loss": 0.2443,
+      "step": 7485
+    },
+    {
+      "epoch": 68.68,
+      "learning_rate": 0.00016432692307692307,
+      "loss": 0.2206,
+      "step": 7486
+    },
+    {
+      "epoch": 68.69,
+      "learning_rate": 0.00016427884615384617,
+      "loss": 0.2519,
+      "step": 7487
+    },
+    {
+      "epoch": 68.7,
+      "learning_rate": 0.00016423076923076924,
+      "loss": 0.1955,
+      "step": 7488
+    },
+    {
+      "epoch": 68.71,
+      "learning_rate": 0.00016418269230769232,
+      "loss": 0.2977,
+      "step": 7489
+    },
+    {
+      "epoch": 68.72,
+      "learning_rate": 0.0001641346153846154,
+      "loss": 0.1443,
+      "step": 7490
+    },
+    {
+      "epoch": 68.72,
+      "learning_rate": 0.00016408653846153846,
+      "loss": 0.1575,
+      "step": 7491
+    },
+    {
+      "epoch": 68.73,
+      "learning_rate": 0.00016403846153846156,
+      "loss": 0.1729,
+      "step": 7492
+    },
+    {
+      "epoch": 68.74,
+      "learning_rate": 0.00016399038461538463,
+      "loss": 0.1812,
+      "step": 7493
+    },
+    {
+      "epoch": 68.75,
+      "learning_rate": 0.0001639423076923077,
+      "loss": 0.3616,
+      "step": 7494
+    },
+    {
+      "epoch": 68.76,
+      "learning_rate": 0.00016389423076923078,
+      "loss": 0.2062,
+      "step": 7495
+    },
+    {
+      "epoch": 68.77,
+      "learning_rate": 0.00016384615384615385,
+      "loss": 0.2411,
+      "step": 7496
+    },
+    {
+      "epoch": 68.78,
+      "learning_rate": 0.00016379807692307695,
+      "loss": 0.1732,
+      "step": 7497
+    },
+    {
+      "epoch": 68.79,
+      "learning_rate": 0.00016375000000000002,
+      "loss": 0.2765,
+      "step": 7498
+    },
+    {
+      "epoch": 68.8,
+      "learning_rate": 0.00016370192307692307,
+      "loss": 0.2459,
+      "step": 7499
+    },
+    {
+      "epoch": 68.81,
+      "learning_rate": 0.00016365384615384614,
+      "loss": 0.2711,
+      "step": 7500
+    },
+    {
+      "epoch": 68.81,
+      "eval_cer": 0.14946251068231908,
+      "eval_loss": 0.6899726986885071,
+      "eval_runtime": 87.9797,
+      "eval_samples_per_second": 18.72,
+      "eval_steps_per_second": 2.341,
+      "eval_wer": 0.5176182208150342,
+      "step": 7500
+    },
+    {
+      "epoch": 68.82,
+      "learning_rate": 0.00016360576923076921,
+      "loss": 0.306,
+      "step": 7501
+    },
+    {
+      "epoch": 68.83,
+      "learning_rate": 0.0001635576923076923,
+      "loss": 0.1957,
+      "step": 7502
+    },
+    {
+      "epoch": 68.83,
+      "learning_rate": 0.00016350961538461539,
+      "loss": 0.1797,
+      "step": 7503
+    },
+    {
+      "epoch": 68.84,
+      "learning_rate": 0.00016346153846153846,
+      "loss": 0.2501,
+      "step": 7504
+    },
+    {
+      "epoch": 68.85,
+      "learning_rate": 0.00016341346153846153,
+      "loss": 0.233,
+      "step": 7505
+    },
+    {
+      "epoch": 68.86,
+      "learning_rate": 0.0001633653846153846,
+      "loss": 0.2,
+      "step": 7506
+    },
+    {
+      "epoch": 68.87,
+      "learning_rate": 0.0001633173076923077,
+      "loss": 0.3077,
+      "step": 7507
+    },
+    {
+      "epoch": 68.88,
+      "learning_rate": 0.00016326923076923078,
+      "loss": 0.2292,
+      "step": 7508
+    },
+    {
+      "epoch": 68.89,
+      "learning_rate": 0.00016322115384615385,
+      "loss": 0.2161,
+      "step": 7509
+    },
+    {
+      "epoch": 68.9,
+      "learning_rate": 0.00016317307692307692,
+      "loss": 0.1987,
+      "step": 7510
+    },
+    {
+      "epoch": 68.91,
+      "learning_rate": 0.000163125,
+      "loss": 0.3208,
+      "step": 7511
+    },
+    {
+      "epoch": 68.92,
+      "learning_rate": 0.00016307692307692307,
+      "loss": 0.2751,
+      "step": 7512
+    },
+    {
+      "epoch": 68.93,
+      "learning_rate": 0.00016302884615384617,
+      "loss": 0.2007,
+      "step": 7513
+    },
+    {
+      "epoch": 68.94,
+      "learning_rate": 0.00016298076923076924,
+      "loss": 0.1834,
+      "step": 7514
+    },
+    {
+      "epoch": 68.94,
+      "learning_rate": 0.0001629326923076923,
+      "loss": 0.2142,
+      "step": 7515
+    },
+    {
+      "epoch": 68.95,
+      "learning_rate": 0.00016288461538461538,
+      "loss": 0.12,
+      "step": 7516
+    },
+    {
+      "epoch": 68.96,
+      "learning_rate": 0.00016283653846153846,
+      "loss": 0.259,
+      "step": 7517
+    },
+    {
+      "epoch": 68.97,
+      "learning_rate": 0.00016278846153846156,
+      "loss": 0.3711,
+      "step": 7518
+    },
+    {
+      "epoch": 68.98,
+      "learning_rate": 0.00016274038461538463,
+      "loss": 0.1981,
+      "step": 7519
+    },
+    {
+      "epoch": 68.99,
+      "learning_rate": 0.0001626923076923077,
+      "loss": 0.1145,
+      "step": 7520
+    },
+    {
+      "epoch": 69.0,
+      "learning_rate": 0.00016264423076923077,
+      "loss": 0.1409,
+      "step": 7521
+    },
+    {
+      "epoch": 69.01,
+      "learning_rate": 0.00016259615384615385,
+      "loss": 0.2608,
+      "step": 7522
+    },
+    {
+      "epoch": 69.02,
+      "learning_rate": 0.00016254807692307695,
+      "loss": 0.2951,
+      "step": 7523
+    },
+    {
+      "epoch": 69.03,
+      "learning_rate": 0.00016250000000000002,
+      "loss": 0.2578,
+      "step": 7524
+    },
+    {
+      "epoch": 69.04,
+      "learning_rate": 0.0001624519230769231,
+      "loss": 0.2856,
+      "step": 7525
+    },
+    {
+      "epoch": 69.05,
+      "learning_rate": 0.00016240384615384616,
+      "loss": 0.2086,
+      "step": 7526
+    },
+    {
+      "epoch": 69.06,
+      "learning_rate": 0.0001623557692307692,
+      "loss": 0.2475,
+      "step": 7527
+    },
+    {
+      "epoch": 69.06,
+      "learning_rate": 0.0001623076923076923,
+      "loss": 0.174,
+      "step": 7528
+    },
+    {
+      "epoch": 69.07,
+      "learning_rate": 0.00016225961538461538,
+      "loss": 0.2597,
+      "step": 7529
+    },
+    {
+      "epoch": 69.08,
+      "learning_rate": 0.00016221153846153846,
+      "loss": 0.2626,
+      "step": 7530
+    },
+    {
+      "epoch": 69.09,
+      "learning_rate": 0.00016216346153846153,
+      "loss": 0.1604,
+      "step": 7531
+    },
+    {
+      "epoch": 69.1,
+      "learning_rate": 0.0001621153846153846,
+      "loss": 0.19,
+      "step": 7532
+    },
+    {
+      "epoch": 69.11,
+      "learning_rate": 0.0001620673076923077,
+      "loss": 0.1858,
+      "step": 7533
+    },
+    {
+      "epoch": 69.12,
+      "learning_rate": 0.00016201923076923077,
+      "loss": 0.1992,
+      "step": 7534
+    },
+    {
+      "epoch": 69.13,
+      "learning_rate": 0.00016197115384615385,
+      "loss": 0.2898,
+      "step": 7535
+    },
+    {
+      "epoch": 69.14,
+      "learning_rate": 0.00016192307692307692,
+      "loss": 0.1783,
+      "step": 7536
+    },
+    {
+      "epoch": 69.15,
+      "learning_rate": 0.000161875,
+      "loss": 0.1946,
+      "step": 7537
+    },
+    {
+      "epoch": 69.16,
+      "learning_rate": 0.0001618269230769231,
+      "loss": 0.2477,
+      "step": 7538
+    },
+    {
+      "epoch": 69.17,
+      "learning_rate": 0.00016177884615384616,
+      "loss": 0.1578,
+      "step": 7539
+    },
+    {
+      "epoch": 69.17,
+      "learning_rate": 0.00016173076923076924,
+      "loss": 0.3264,
+      "step": 7540
+    },
+    {
+      "epoch": 69.18,
+      "learning_rate": 0.0001616826923076923,
+      "loss": 0.2085,
+      "step": 7541
+    },
+    {
+      "epoch": 69.19,
+      "learning_rate": 0.00016163461538461538,
+      "loss": 0.22,
+      "step": 7542
+    },
+    {
+      "epoch": 69.2,
+      "learning_rate": 0.00016158653846153845,
+      "loss": 0.139,
+      "step": 7543
+    },
+    {
+      "epoch": 69.21,
+      "learning_rate": 0.00016153846153846155,
+      "loss": 0.2314,
+      "step": 7544
+    },
+    {
+      "epoch": 69.22,
+      "learning_rate": 0.00016149038461538463,
+      "loss": 0.2475,
+      "step": 7545
+    },
+    {
+      "epoch": 69.23,
+      "learning_rate": 0.0001614423076923077,
+      "loss": 0.2009,
+      "step": 7546
+    },
+    {
+      "epoch": 69.24,
+      "learning_rate": 0.00016139423076923077,
+      "loss": 0.1926,
+      "step": 7547
+    },
+    {
+      "epoch": 69.25,
+      "learning_rate": 0.00016134615384615384,
+      "loss": 0.1318,
+      "step": 7548
+    },
+    {
+      "epoch": 69.26,
+      "learning_rate": 0.00016129807692307694,
+      "loss": 0.3348,
+      "step": 7549
+    },
+    {
+      "epoch": 69.27,
+      "learning_rate": 0.00016125000000000002,
+      "loss": 0.2538,
+      "step": 7550
+    },
+    {
+      "epoch": 69.28,
+      "learning_rate": 0.0001612019230769231,
+      "loss": 0.237,
+      "step": 7551
+    },
+    {
+      "epoch": 69.28,
+      "learning_rate": 0.00016115384615384616,
+      "loss": 0.2272,
+      "step": 7552
+    },
+    {
+      "epoch": 69.29,
+      "learning_rate": 0.00016110576923076923,
+      "loss": 0.2119,
+      "step": 7553
+    },
+    {
+      "epoch": 69.3,
+      "learning_rate": 0.00016105769230769233,
+      "loss": 0.2278,
+      "step": 7554
+    },
+    {
+      "epoch": 69.31,
+      "learning_rate": 0.0001610096153846154,
+      "loss": 0.2719,
+      "step": 7555
+    },
+    {
+      "epoch": 69.32,
+      "learning_rate": 0.00016096153846153845,
+      "loss": 0.2446,
+      "step": 7556
+    },
+    {
+      "epoch": 69.33,
+      "learning_rate": 0.00016091346153846152,
+      "loss": 0.1592,
+      "step": 7557
+    },
+    {
+      "epoch": 69.34,
+      "learning_rate": 0.0001608653846153846,
+      "loss": 0.2749,
+      "step": 7558
+    },
+    {
+      "epoch": 69.35,
+      "learning_rate": 0.0001608173076923077,
+      "loss": 0.2214,
+      "step": 7559
+    },
+    {
+      "epoch": 69.36,
+      "learning_rate": 0.00016076923076923077,
+      "loss": 0.2152,
+      "step": 7560
+    },
+    {
+      "epoch": 69.37,
+      "learning_rate": 0.00016072115384615384,
+      "loss": 0.3107,
+      "step": 7561
+    },
+    {
+      "epoch": 69.38,
+      "learning_rate": 0.00016067307692307691,
+      "loss": 0.2419,
+      "step": 7562
+    },
+    {
+      "epoch": 69.39,
+      "learning_rate": 0.000160625,
+      "loss": 0.2501,
+      "step": 7563
+    },
+    {
+      "epoch": 69.39,
+      "learning_rate": 0.0001605769230769231,
+      "loss": 0.2496,
+      "step": 7564
+    },
+    {
+      "epoch": 69.4,
+      "learning_rate": 0.00016052884615384616,
+      "loss": 0.1932,
+      "step": 7565
+    },
+    {
+      "epoch": 69.41,
+      "learning_rate": 0.00016048076923076923,
+      "loss": 0.1485,
+      "step": 7566
+    },
+    {
+      "epoch": 69.42,
+      "learning_rate": 0.0001604326923076923,
+      "loss": 0.1928,
+      "step": 7567
+    },
+    {
+      "epoch": 69.43,
+      "learning_rate": 0.00016038461538461538,
+      "loss": 0.262,
+      "step": 7568
+    },
+    {
+      "epoch": 69.44,
+      "learning_rate": 0.00016033653846153848,
+      "loss": 0.2314,
+      "step": 7569
+    },
+    {
+      "epoch": 69.45,
+      "learning_rate": 0.00016028846153846155,
+      "loss": 0.2493,
+      "step": 7570
+    },
+    {
+      "epoch": 69.46,
+      "learning_rate": 0.00016024038461538462,
+      "loss": 0.1945,
+      "step": 7571
+    },
+    {
+      "epoch": 69.47,
+      "learning_rate": 0.0001601923076923077,
+      "loss": 0.1464,
+      "step": 7572
+    },
+    {
+      "epoch": 69.48,
+      "learning_rate": 0.00016014423076923077,
+      "loss": 0.2282,
+      "step": 7573
+    },
+    {
+      "epoch": 69.49,
+      "learning_rate": 0.00016009615384615384,
+      "loss": 0.1872,
+      "step": 7574
+    },
+    {
+      "epoch": 69.5,
+      "learning_rate": 0.00016004807692307694,
+      "loss": 0.1028,
+      "step": 7575
+    },
+    {
+      "epoch": 69.5,
+      "learning_rate": 0.00016,
+      "loss": 0.3132,
+      "step": 7576
+    },
+    {
+      "epoch": 69.51,
+      "learning_rate": 0.00015995192307692309,
+      "loss": 0.266,
+      "step": 7577
+    },
+    {
+      "epoch": 69.52,
+      "learning_rate": 0.00015990384615384616,
+      "loss": 0.234,
+      "step": 7578
+    },
+    {
+      "epoch": 69.53,
+      "learning_rate": 0.00015985576923076923,
+      "loss": 0.2869,
+      "step": 7579
+    },
+    {
+      "epoch": 69.54,
+      "learning_rate": 0.00015980769230769233,
+      "loss": 0.2158,
+      "step": 7580
+    },
+    {
+      "epoch": 69.55,
+      "learning_rate": 0.0001597596153846154,
+      "loss": 0.2519,
+      "step": 7581
+    },
+    {
+      "epoch": 69.56,
+      "learning_rate": 0.00015971153846153848,
+      "loss": 0.2648,
+      "step": 7582
+    },
+    {
+      "epoch": 69.57,
+      "learning_rate": 0.00015966346153846155,
+      "loss": 0.2259,
+      "step": 7583
+    },
+    {
+      "epoch": 69.58,
+      "learning_rate": 0.0001596153846153846,
+      "loss": 0.3207,
+      "step": 7584
+    },
+    {
+      "epoch": 69.59,
+      "learning_rate": 0.0001595673076923077,
+      "loss": 0.3108,
+      "step": 7585
+    },
+    {
+      "epoch": 69.6,
+      "learning_rate": 0.00015951923076923077,
+      "loss": 0.2724,
+      "step": 7586
+    },
+    {
+      "epoch": 69.61,
+      "learning_rate": 0.00015947115384615384,
+      "loss": 0.2523,
+      "step": 7587
+    },
+    {
+      "epoch": 69.61,
+      "learning_rate": 0.0001594230769230769,
+      "loss": 0.2759,
+      "step": 7588
+    },
+    {
+      "epoch": 69.62,
+      "learning_rate": 0.00015937499999999998,
+      "loss": 0.2058,
+      "step": 7589
+    },
+    {
+      "epoch": 69.63,
+      "learning_rate": 0.00015932692307692308,
+      "loss": 0.2518,
+      "step": 7590
+    },
+    {
+      "epoch": 69.64,
+      "learning_rate": 0.00015927884615384616,
+      "loss": 0.1348,
+      "step": 7591
+    },
+    {
+      "epoch": 69.65,
+      "learning_rate": 0.00015923076923076923,
+      "loss": 0.2535,
+      "step": 7592
+    },
+    {
+      "epoch": 69.66,
+      "learning_rate": 0.0001591826923076923,
+      "loss": 0.2031,
+      "step": 7593
+    },
+    {
+      "epoch": 69.67,
+      "learning_rate": 0.00015913461538461537,
+      "loss": 0.3083,
+      "step": 7594
+    },
+    {
+      "epoch": 69.68,
+      "learning_rate": 0.00015908653846153847,
+      "loss": 0.2158,
+      "step": 7595
+    },
+    {
+      "epoch": 69.69,
+      "learning_rate": 0.00015903846153846155,
+      "loss": 0.2262,
+      "step": 7596
+    },
+    {
+      "epoch": 69.7,
+      "learning_rate": 0.00015899038461538462,
+      "loss": 0.2485,
+      "step": 7597
+    },
+    {
+      "epoch": 69.71,
+      "learning_rate": 0.0001589423076923077,
+      "loss": 0.2121,
+      "step": 7598
+    },
+    {
+      "epoch": 69.72,
+      "learning_rate": 0.00015889423076923076,
+      "loss": 0.2141,
+      "step": 7599
+    },
+    {
+      "epoch": 69.72,
+      "learning_rate": 0.00015884615384615386,
+      "loss": 0.1732,
+      "step": 7600
+    },
+    {
+      "epoch": 69.73,
+      "learning_rate": 0.00015879807692307694,
+      "loss": 0.2034,
+      "step": 7601
+    },
+    {
+      "epoch": 69.74,
+      "learning_rate": 0.00015875,
+      "loss": 0.2488,
+      "step": 7602
+    },
+    {
+      "epoch": 69.75,
+      "learning_rate": 0.00015870192307692308,
+      "loss": 0.2809,
+      "step": 7603
+    },
+    {
+      "epoch": 69.76,
+      "learning_rate": 0.00015865384615384616,
+      "loss": 0.2762,
+      "step": 7604
+    },
+    {
+      "epoch": 69.77,
+      "learning_rate": 0.00015860576923076925,
+      "loss": 0.2938,
+      "step": 7605
+    },
+    {
+      "epoch": 69.78,
+      "learning_rate": 0.00015855769230769233,
+      "loss": 0.2758,
+      "step": 7606
+    },
+    {
+      "epoch": 69.79,
+      "learning_rate": 0.0001585096153846154,
+      "loss": 0.2175,
+      "step": 7607
+    },
+    {
+      "epoch": 69.8,
+      "learning_rate": 0.00015846153846153847,
+      "loss": 0.1929,
+      "step": 7608
+    },
+    {
+      "epoch": 69.81,
+      "learning_rate": 0.00015841346153846155,
+      "loss": 0.2648,
+      "step": 7609
+    },
+    {
+      "epoch": 69.82,
+      "learning_rate": 0.00015836538461538462,
+      "loss": 0.2287,
+      "step": 7610
+    },
+    {
+      "epoch": 69.83,
+      "learning_rate": 0.00015831730769230772,
+      "loss": 0.3064,
+      "step": 7611
+    },
+    {
+      "epoch": 69.83,
+      "learning_rate": 0.00015826923076923076,
+      "loss": 0.2067,
+      "step": 7612
+    },
+    {
+      "epoch": 69.84,
+      "learning_rate": 0.00015822115384615384,
+      "loss": 0.2468,
+      "step": 7613
+    },
+    {
+      "epoch": 69.85,
+      "learning_rate": 0.0001581730769230769,
+      "loss": 0.2566,
+      "step": 7614
+    },
+    {
+      "epoch": 69.86,
+      "learning_rate": 0.00015812499999999998,
+      "loss": 0.2318,
+      "step": 7615
+    },
+    {
+      "epoch": 69.87,
+      "learning_rate": 0.00015807692307692308,
+      "loss": 0.2284,
+      "step": 7616
+    },
+    {
+      "epoch": 69.88,
+      "learning_rate": 0.00015802884615384615,
+      "loss": 0.1793,
+      "step": 7617
+    },
+    {
+      "epoch": 69.89,
+      "learning_rate": 0.00015798076923076923,
+      "loss": 0.2067,
+      "step": 7618
+    },
+    {
+      "epoch": 69.9,
+      "learning_rate": 0.0001579326923076923,
+      "loss": 0.2377,
+      "step": 7619
+    },
+    {
+      "epoch": 69.91,
+      "learning_rate": 0.00015788461538461537,
+      "loss": 0.2847,
+      "step": 7620
+    },
+    {
+      "epoch": 69.92,
+      "learning_rate": 0.00015783653846153847,
+      "loss": 0.2482,
+      "step": 7621
+    },
+    {
+      "epoch": 69.93,
+      "learning_rate": 0.00015778846153846154,
+      "loss": 0.2059,
+      "step": 7622
+    },
+    {
+      "epoch": 69.94,
+      "learning_rate": 0.00015774038461538462,
+      "loss": 0.1836,
+      "step": 7623
+    },
+    {
+      "epoch": 69.94,
+      "learning_rate": 0.0001576923076923077,
+      "loss": 0.154,
+      "step": 7624
+    },
+    {
+      "epoch": 69.95,
+      "learning_rate": 0.00015764423076923076,
+      "loss": 0.2026,
+      "step": 7625
+    },
+    {
+      "epoch": 69.96,
+      "learning_rate": 0.00015759615384615386,
+      "loss": 0.1334,
+      "step": 7626
+    },
+    {
+      "epoch": 69.97,
+      "learning_rate": 0.00015754807692307693,
+      "loss": 0.2089,
+      "step": 7627
+    },
+    {
+      "epoch": 69.98,
+      "learning_rate": 0.0001575,
+      "loss": 0.1577,
+      "step": 7628
+    },
+    {
+      "epoch": 69.99,
+      "learning_rate": 0.00015745192307692308,
+      "loss": 0.1256,
+      "step": 7629
+    },
+    {
+      "epoch": 70.0,
+      "learning_rate": 0.00015740384615384615,
+      "loss": 0.2684,
+      "step": 7630
+    },
+    {
+      "epoch": 70.01,
+      "learning_rate": 0.00015735576923076925,
+      "loss": 0.2893,
+      "step": 7631
+    },
+    {
+      "epoch": 70.02,
+      "learning_rate": 0.00015730769230769232,
+      "loss": 0.1976,
+      "step": 7632
+    },
+    {
+      "epoch": 70.03,
+      "learning_rate": 0.0001572596153846154,
+      "loss": 0.2394,
+      "step": 7633
+    },
+    {
+      "epoch": 70.04,
+      "learning_rate": 0.00015721153846153847,
+      "loss": 0.3078,
+      "step": 7634
+    },
+    {
+      "epoch": 70.05,
+      "learning_rate": 0.00015716346153846154,
+      "loss": 0.1777,
+      "step": 7635
+    },
+    {
+      "epoch": 70.06,
+      "learning_rate": 0.00015711538461538464,
+      "loss": 0.2148,
+      "step": 7636
+    },
+    {
+      "epoch": 70.06,
+      "learning_rate": 0.00015706730769230771,
+      "loss": 0.2344,
+      "step": 7637
+    },
+    {
+      "epoch": 70.07,
+      "learning_rate": 0.0001570192307692308,
+      "loss": 0.267,
+      "step": 7638
+    },
+    {
+      "epoch": 70.08,
+      "learning_rate": 0.00015697115384615386,
+      "loss": 0.2009,
+      "step": 7639
+    },
+    {
+      "epoch": 70.09,
+      "learning_rate": 0.0001569230769230769,
+      "loss": 0.2146,
+      "step": 7640
+    },
+    {
+      "epoch": 70.1,
+      "learning_rate": 0.00015687499999999998,
+      "loss": 0.257,
+      "step": 7641
+    },
+    {
+      "epoch": 70.11,
+      "learning_rate": 0.00015682692307692308,
+      "loss": 0.2443,
+      "step": 7642
+    },
+    {
+      "epoch": 70.12,
+      "learning_rate": 0.00015677884615384615,
+      "loss": 0.1646,
+      "step": 7643
+    },
+    {
+      "epoch": 70.13,
+      "learning_rate": 0.00015673076923076922,
+      "loss": 0.1434,
+      "step": 7644
+    },
+    {
+      "epoch": 70.14,
+      "learning_rate": 0.0001566826923076923,
+      "loss": 0.2868,
+      "step": 7645
+    },
+    {
+      "epoch": 70.15,
+      "learning_rate": 0.00015663461538461537,
+      "loss": 0.1692,
+      "step": 7646
+    },
+    {
+      "epoch": 70.16,
+      "learning_rate": 0.00015658653846153847,
+      "loss": 0.2155,
+      "step": 7647
+    },
+    {
+      "epoch": 70.17,
+      "learning_rate": 0.00015653846153846154,
+      "loss": 0.2189,
+      "step": 7648
+    },
+    {
+      "epoch": 70.17,
+      "learning_rate": 0.0001564903846153846,
+      "loss": 0.2157,
+      "step": 7649
+    },
+    {
+      "epoch": 70.18,
+      "learning_rate": 0.00015644230769230769,
+      "loss": 0.1934,
+      "step": 7650
+    },
+    {
+      "epoch": 70.19,
+      "learning_rate": 0.00015639423076923076,
+      "loss": 0.2159,
+      "step": 7651
+    },
+    {
+      "epoch": 70.2,
+      "learning_rate": 0.00015634615384615386,
+      "loss": 0.2236,
+      "step": 7652
+    },
+    {
+      "epoch": 70.21,
+      "learning_rate": 0.00015629807692307693,
+      "loss": 0.1779,
+      "step": 7653
+    },
+    {
+      "epoch": 70.22,
+      "learning_rate": 0.00015625,
+      "loss": 0.2606,
+      "step": 7654
+    },
+    {
+      "epoch": 70.23,
+      "learning_rate": 0.00015620192307692308,
+      "loss": 0.2245,
+      "step": 7655
+    },
+    {
+      "epoch": 70.24,
+      "learning_rate": 0.00015615384615384615,
+      "loss": 0.1744,
+      "step": 7656
+    },
+    {
+      "epoch": 70.25,
+      "learning_rate": 0.00015610576923076925,
+      "loss": 0.1675,
+      "step": 7657
+    },
+    {
+      "epoch": 70.26,
+      "learning_rate": 0.00015605769230769232,
+      "loss": 0.3036,
+      "step": 7658
+    },
+    {
+      "epoch": 70.27,
+      "learning_rate": 0.0001560096153846154,
+      "loss": 0.3433,
+      "step": 7659
+    },
+    {
+      "epoch": 70.28,
+      "learning_rate": 0.00015596153846153847,
+      "loss": 0.2387,
+      "step": 7660
+    },
+    {
+      "epoch": 70.28,
+      "learning_rate": 0.00015591346153846154,
+      "loss": 0.2494,
+      "step": 7661
+    },
+    {
+      "epoch": 70.29,
+      "learning_rate": 0.00015586538461538464,
+      "loss": 0.1854,
+      "step": 7662
+    },
+    {
+      "epoch": 70.3,
+      "learning_rate": 0.0001558173076923077,
+      "loss": 0.2979,
+      "step": 7663
+    },
+    {
+      "epoch": 70.31,
+      "learning_rate": 0.00015576923076923078,
+      "loss": 0.2679,
+      "step": 7664
+    },
+    {
+      "epoch": 70.32,
+      "learning_rate": 0.00015572115384615386,
+      "loss": 0.1864,
+      "step": 7665
+    },
+    {
+      "epoch": 70.33,
+      "learning_rate": 0.00015567307692307693,
+      "loss": 0.1874,
+      "step": 7666
+    },
+    {
+      "epoch": 70.34,
+      "learning_rate": 0.00015562500000000003,
+      "loss": 0.2371,
+      "step": 7667
+    },
+    {
+      "epoch": 70.35,
+      "learning_rate": 0.0001555769230769231,
+      "loss": 0.2168,
+      "step": 7668
+    },
+    {
+      "epoch": 70.36,
+      "learning_rate": 0.00015552884615384615,
+      "loss": 0.2404,
+      "step": 7669
+    },
+    {
+      "epoch": 70.37,
+      "learning_rate": 0.00015548076923076922,
+      "loss": 0.2231,
+      "step": 7670
+    },
+    {
+      "epoch": 70.38,
+      "learning_rate": 0.0001554326923076923,
+      "loss": 0.1451,
+      "step": 7671
+    },
+    {
+      "epoch": 70.39,
+      "learning_rate": 0.00015538461538461536,
+      "loss": 0.1746,
+      "step": 7672
+    },
+    {
+      "epoch": 70.39,
+      "learning_rate": 0.00015533653846153846,
+      "loss": 0.2109,
+      "step": 7673
+    },
+    {
+      "epoch": 70.4,
+      "learning_rate": 0.00015528846153846154,
+      "loss": 0.2666,
+      "step": 7674
+    },
+    {
+      "epoch": 70.41,
+      "learning_rate": 0.0001552403846153846,
+      "loss": 0.2428,
+      "step": 7675
+    },
+    {
+      "epoch": 70.42,
+      "learning_rate": 0.00015519230769230768,
+      "loss": 0.1733,
+      "step": 7676
+    },
+    {
+      "epoch": 70.43,
+      "learning_rate": 0.00015514423076923075,
+      "loss": 0.2057,
+      "step": 7677
+    },
+    {
+      "epoch": 70.44,
+      "learning_rate": 0.00015509615384615385,
+      "loss": 0.2159,
+      "step": 7678
+    },
+    {
+      "epoch": 70.45,
+      "learning_rate": 0.00015504807692307693,
+      "loss": 0.2129,
+      "step": 7679
+    },
+    {
+      "epoch": 70.46,
+      "learning_rate": 0.000155,
+      "loss": 0.1901,
+      "step": 7680
+    },
+    {
+      "epoch": 70.47,
+      "learning_rate": 0.00015495192307692307,
+      "loss": 0.2203,
+      "step": 7681
+    },
+    {
+      "epoch": 70.48,
+      "learning_rate": 0.00015490384615384615,
+      "loss": 0.1262,
+      "step": 7682
+    },
+    {
+      "epoch": 70.49,
+      "learning_rate": 0.00015485576923076924,
+      "loss": 0.1754,
+      "step": 7683
+    },
+    {
+      "epoch": 70.5,
+      "learning_rate": 0.00015480769230769232,
+      "loss": 0.184,
+      "step": 7684
+    },
+    {
+      "epoch": 70.5,
+      "learning_rate": 0.0001547596153846154,
+      "loss": 0.3672,
+      "step": 7685
+    },
+    {
+      "epoch": 70.51,
+      "learning_rate": 0.00015471153846153846,
+      "loss": 0.263,
+      "step": 7686
+    },
+    {
+      "epoch": 70.52,
+      "learning_rate": 0.00015466346153846154,
+      "loss": 0.221,
+      "step": 7687
+    },
+    {
+      "epoch": 70.53,
+      "learning_rate": 0.00015461538461538464,
+      "loss": 0.2361,
+      "step": 7688
+    },
+    {
+      "epoch": 70.54,
+      "learning_rate": 0.0001545673076923077,
+      "loss": 0.2249,
+      "step": 7689
+    },
+    {
+      "epoch": 70.55,
+      "learning_rate": 0.00015451923076923078,
+      "loss": 0.2797,
+      "step": 7690
+    },
+    {
+      "epoch": 70.56,
+      "learning_rate": 0.00015447115384615385,
+      "loss": 0.2826,
+      "step": 7691
+    },
+    {
+      "epoch": 70.57,
+      "learning_rate": 0.00015442307692307693,
+      "loss": 0.1871,
+      "step": 7692
+    },
+    {
+      "epoch": 70.58,
+      "learning_rate": 0.00015437500000000003,
+      "loss": 0.2302,
+      "step": 7693
+    },
+    {
+      "epoch": 70.59,
+      "learning_rate": 0.0001543269230769231,
+      "loss": 0.1642,
+      "step": 7694
+    },
+    {
+      "epoch": 70.6,
+      "learning_rate": 0.00015427884615384617,
+      "loss": 0.2509,
+      "step": 7695
+    },
+    {
+      "epoch": 70.61,
+      "learning_rate": 0.00015423076923076924,
+      "loss": 0.1782,
+      "step": 7696
+    },
+    {
+      "epoch": 70.61,
+      "learning_rate": 0.0001541826923076923,
+      "loss": 0.1943,
+      "step": 7697
+    },
+    {
+      "epoch": 70.62,
+      "learning_rate": 0.0001541346153846154,
+      "loss": 0.2617,
+      "step": 7698
+    },
+    {
+      "epoch": 70.63,
+      "learning_rate": 0.00015408653846153846,
+      "loss": 0.1509,
+      "step": 7699
+    },
+    {
+      "epoch": 70.64,
+      "learning_rate": 0.00015403846153846153,
+      "loss": 0.1747,
+      "step": 7700
+    },
+    {
+      "epoch": 70.65,
+      "learning_rate": 0.0001539903846153846,
+      "loss": 0.1917,
+      "step": 7701
+    },
+    {
+      "epoch": 70.66,
+      "learning_rate": 0.00015394230769230768,
+      "loss": 0.2013,
+      "step": 7702
+    },
+    {
+      "epoch": 70.67,
+      "learning_rate": 0.00015389423076923075,
+      "loss": 0.2001,
+      "step": 7703
+    },
+    {
+      "epoch": 70.68,
+      "learning_rate": 0.00015384615384615385,
+      "loss": 0.1829,
+      "step": 7704
+    },
+    {
+      "epoch": 70.69,
+      "learning_rate": 0.00015379807692307692,
+      "loss": 0.2275,
+      "step": 7705
+    },
+    {
+      "epoch": 70.7,
+      "learning_rate": 0.00015375,
+      "loss": 0.2406,
+      "step": 7706
+    },
+    {
+      "epoch": 70.71,
+      "learning_rate": 0.00015370192307692307,
+      "loss": 0.1716,
+      "step": 7707
+    },
+    {
+      "epoch": 70.72,
+      "learning_rate": 0.00015365384615384614,
+      "loss": 0.1888,
+      "step": 7708
+    },
+    {
+      "epoch": 70.72,
+      "learning_rate": 0.00015360576923076924,
+      "loss": 0.1151,
+      "step": 7709
+    },
+    {
+      "epoch": 70.73,
+      "learning_rate": 0.00015355769230769231,
+      "loss": 0.207,
+      "step": 7710
+    },
+    {
+      "epoch": 70.74,
+      "learning_rate": 0.0001535096153846154,
+      "loss": 0.2168,
+      "step": 7711
+    },
+    {
+      "epoch": 70.75,
+      "learning_rate": 0.00015346153846153846,
+      "loss": 0.285,
+      "step": 7712
+    },
+    {
+      "epoch": 70.76,
+      "learning_rate": 0.00015341346153846153,
+      "loss": 0.2427,
+      "step": 7713
+    },
+    {
+      "epoch": 70.77,
+      "learning_rate": 0.00015336538461538463,
+      "loss": 0.2584,
+      "step": 7714
+    },
+    {
+      "epoch": 70.78,
+      "learning_rate": 0.0001533173076923077,
+      "loss": 0.1957,
+      "step": 7715
+    },
+    {
+      "epoch": 70.79,
+      "learning_rate": 0.00015326923076923078,
+      "loss": 0.1851,
+      "step": 7716
+    },
+    {
+      "epoch": 70.8,
+      "learning_rate": 0.00015322115384615385,
+      "loss": 0.2766,
+      "step": 7717
+    },
+    {
+      "epoch": 70.81,
+      "learning_rate": 0.00015317307692307692,
+      "loss": 0.1942,
+      "step": 7718
+    },
+    {
+      "epoch": 70.82,
+      "learning_rate": 0.00015312500000000002,
+      "loss": 0.1746,
+      "step": 7719
+    },
+    {
+      "epoch": 70.83,
+      "learning_rate": 0.0001530769230769231,
+      "loss": 0.1919,
+      "step": 7720
+    },
+    {
+      "epoch": 70.83,
+      "learning_rate": 0.00015302884615384617,
+      "loss": 0.1822,
+      "step": 7721
+    },
+    {
+      "epoch": 70.84,
+      "learning_rate": 0.00015298076923076924,
+      "loss": 0.2065,
+      "step": 7722
+    },
+    {
+      "epoch": 70.85,
+      "learning_rate": 0.0001529326923076923,
+      "loss": 0.1828,
+      "step": 7723
+    },
+    {
+      "epoch": 70.86,
+      "learning_rate": 0.0001528846153846154,
+      "loss": 0.1659,
+      "step": 7724
+    },
+    {
+      "epoch": 70.87,
+      "learning_rate": 0.00015283653846153846,
+      "loss": 0.1422,
+      "step": 7725
+    },
+    {
+      "epoch": 70.88,
+      "learning_rate": 0.00015278846153846153,
+      "loss": 0.2407,
+      "step": 7726
+    },
+    {
+      "epoch": 70.89,
+      "learning_rate": 0.0001527403846153846,
+      "loss": 0.1328,
+      "step": 7727
+    },
+    {
+      "epoch": 70.9,
+      "learning_rate": 0.00015269230769230768,
+      "loss": 0.2174,
+      "step": 7728
+    },
+    {
+      "epoch": 70.91,
+      "learning_rate": 0.00015264423076923078,
+      "loss": 0.212,
+      "step": 7729
+    },
+    {
+      "epoch": 70.92,
+      "learning_rate": 0.00015259615384615385,
+      "loss": 0.2881,
+      "step": 7730
+    },
+    {
+      "epoch": 70.93,
+      "learning_rate": 0.00015254807692307692,
+      "loss": 0.2344,
+      "step": 7731
+    },
+    {
+      "epoch": 70.94,
+      "learning_rate": 0.0001525,
+      "loss": 0.1855,
+      "step": 7732
+    },
+    {
+      "epoch": 70.94,
+      "learning_rate": 0.00015245192307692307,
+      "loss": 0.2443,
+      "step": 7733
+    },
+    {
+      "epoch": 70.95,
+      "learning_rate": 0.00015240384615384617,
+      "loss": 0.2154,
+      "step": 7734
+    },
+    {
+      "epoch": 70.96,
+      "learning_rate": 0.00015235576923076924,
+      "loss": 0.1587,
+      "step": 7735
+    },
+    {
+      "epoch": 70.97,
+      "learning_rate": 0.0001523076923076923,
+      "loss": 0.255,
+      "step": 7736
+    },
+    {
+      "epoch": 70.98,
+      "learning_rate": 0.00015225961538461538,
+      "loss": 0.123,
+      "step": 7737
+    },
+    {
+      "epoch": 70.99,
+      "learning_rate": 0.00015221153846153846,
+      "loss": 0.1986,
+      "step": 7738
+    },
+    {
+      "epoch": 71.0,
+      "learning_rate": 0.00015216346153846153,
+      "loss": 0.2506,
+      "step": 7739
+    },
+    {
+      "epoch": 71.01,
+      "learning_rate": 0.00015211538461538463,
+      "loss": 0.2734,
+      "step": 7740
+    },
+    {
+      "epoch": 71.02,
+      "learning_rate": 0.0001520673076923077,
+      "loss": 0.2831,
+      "step": 7741
+    },
+    {
+      "epoch": 71.03,
+      "learning_rate": 0.00015201923076923077,
+      "loss": 0.1955,
+      "step": 7742
+    },
+    {
+      "epoch": 71.04,
+      "learning_rate": 0.00015197115384615385,
+      "loss": 0.2397,
+      "step": 7743
+    },
+    {
+      "epoch": 71.05,
+      "learning_rate": 0.00015192307692307692,
+      "loss": 0.182,
+      "step": 7744
+    },
+    {
+      "epoch": 71.06,
+      "learning_rate": 0.00015187500000000002,
+      "loss": 0.26,
+      "step": 7745
+    },
+    {
+      "epoch": 71.06,
+      "learning_rate": 0.0001518269230769231,
+      "loss": 0.2331,
+      "step": 7746
+    },
+    {
+      "epoch": 71.07,
+      "learning_rate": 0.00015177884615384616,
+      "loss": 0.2032,
+      "step": 7747
+    },
+    {
+      "epoch": 71.08,
+      "learning_rate": 0.00015173076923076924,
+      "loss": 0.1455,
+      "step": 7748
+    },
+    {
+      "epoch": 71.09,
+      "learning_rate": 0.0001516826923076923,
+      "loss": 0.2155,
+      "step": 7749
+    },
+    {
+      "epoch": 71.1,
+      "learning_rate": 0.0001516346153846154,
+      "loss": 0.1921,
+      "step": 7750
+    },
+    {
+      "epoch": 71.11,
+      "learning_rate": 0.00015158653846153848,
+      "loss": 0.1956,
+      "step": 7751
+    },
+    {
+      "epoch": 71.12,
+      "learning_rate": 0.00015153846153846155,
+      "loss": 0.2143,
+      "step": 7752
+    },
+    {
+      "epoch": 71.13,
+      "learning_rate": 0.0001514903846153846,
+      "loss": 0.1043,
+      "step": 7753
+    },
+    {
+      "epoch": 71.14,
+      "learning_rate": 0.00015144230769230767,
+      "loss": 0.2476,
+      "step": 7754
+    },
+    {
+      "epoch": 71.15,
+      "learning_rate": 0.00015139423076923077,
+      "loss": 0.1665,
+      "step": 7755
+    },
+    {
+      "epoch": 71.16,
+      "learning_rate": 0.00015134615384615384,
+      "loss": 0.1216,
+      "step": 7756
+    },
+    {
+      "epoch": 71.17,
+      "learning_rate": 0.00015129807692307692,
+      "loss": 0.2122,
+      "step": 7757
+    },
+    {
+      "epoch": 71.17,
+      "learning_rate": 0.00015125,
+      "loss": 0.2133,
+      "step": 7758
+    },
+    {
+      "epoch": 71.18,
+      "learning_rate": 0.00015120192307692306,
+      "loss": 0.2447,
+      "step": 7759
+    },
+    {
+      "epoch": 71.19,
+      "learning_rate": 0.00015115384615384616,
+      "loss": 0.202,
+      "step": 7760
+    },
+    {
+      "epoch": 71.2,
+      "learning_rate": 0.00015110576923076924,
+      "loss": 0.1436,
+      "step": 7761
+    },
+    {
+      "epoch": 71.21,
+      "learning_rate": 0.0001510576923076923,
+      "loss": 0.1401,
+      "step": 7762
+    },
+    {
+      "epoch": 71.22,
+      "learning_rate": 0.00015100961538461538,
+      "loss": 0.1536,
+      "step": 7763
+    },
+    {
+      "epoch": 71.23,
+      "learning_rate": 0.00015096153846153845,
+      "loss": 0.1981,
+      "step": 7764
+    },
+    {
+      "epoch": 71.24,
+      "learning_rate": 0.00015091346153846155,
+      "loss": 0.1849,
+      "step": 7765
+    },
+    {
+      "epoch": 71.25,
+      "learning_rate": 0.00015086538461538463,
+      "loss": 0.1702,
+      "step": 7766
+    },
+    {
+      "epoch": 71.26,
+      "learning_rate": 0.0001508173076923077,
+      "loss": 0.2703,
+      "step": 7767
+    },
+    {
+      "epoch": 71.27,
+      "learning_rate": 0.00015076923076923077,
+      "loss": 0.2683,
+      "step": 7768
+    },
+    {
+      "epoch": 71.28,
+      "learning_rate": 0.00015072115384615384,
+      "loss": 0.2372,
+      "step": 7769
+    },
+    {
+      "epoch": 71.28,
+      "learning_rate": 0.00015067307692307692,
+      "loss": 0.142,
+      "step": 7770
+    },
+    {
+      "epoch": 71.29,
+      "learning_rate": 0.00015062500000000002,
+      "loss": 0.2063,
+      "step": 7771
+    },
+    {
+      "epoch": 71.3,
+      "learning_rate": 0.0001505769230769231,
+      "loss": 0.216,
+      "step": 7772
+    },
+    {
+      "epoch": 71.31,
+      "learning_rate": 0.00015052884615384616,
+      "loss": 0.203,
+      "step": 7773
+    },
+    {
+      "epoch": 71.32,
+      "learning_rate": 0.00015048076923076923,
+      "loss": 0.1672,
+      "step": 7774
+    },
+    {
+      "epoch": 71.33,
+      "learning_rate": 0.0001504326923076923,
+      "loss": 0.2223,
+      "step": 7775
+    },
+    {
+      "epoch": 71.34,
+      "learning_rate": 0.0001503846153846154,
+      "loss": 0.1891,
+      "step": 7776
+    },
+    {
+      "epoch": 71.35,
+      "learning_rate": 0.00015033653846153848,
+      "loss": 0.2113,
+      "step": 7777
+    },
+    {
+      "epoch": 71.36,
+      "learning_rate": 0.00015028846153846155,
+      "loss": 0.1868,
+      "step": 7778
+    },
+    {
+      "epoch": 71.37,
+      "learning_rate": 0.00015024038461538462,
+      "loss": 0.3009,
+      "step": 7779
+    },
+    {
+      "epoch": 71.38,
+      "learning_rate": 0.0001501923076923077,
+      "loss": 0.1728,
+      "step": 7780
+    },
+    {
+      "epoch": 71.39,
+      "learning_rate": 0.0001501442307692308,
+      "loss": 0.2317,
+      "step": 7781
+    },
+    {
+      "epoch": 71.39,
+      "learning_rate": 0.00015009615384615384,
+      "loss": 0.269,
+      "step": 7782
+    },
+    {
+      "epoch": 71.4,
+      "learning_rate": 0.00015004807692307691,
+      "loss": 0.1469,
+      "step": 7783
+    },
+    {
+      "epoch": 71.41,
+      "learning_rate": 0.00015,
+      "loss": 0.2046,
+      "step": 7784
+    },
+    {
+      "epoch": 71.42,
+      "learning_rate": 0.00014995192307692306,
+      "loss": 0.2343,
+      "step": 7785
+    },
+    {
+      "epoch": 71.43,
+      "learning_rate": 0.00014990384615384616,
+      "loss": 0.2743,
+      "step": 7786
+    },
+    {
+      "epoch": 71.44,
+      "learning_rate": 0.00014985576923076923,
+      "loss": 0.3465,
+      "step": 7787
+    },
+    {
+      "epoch": 71.45,
+      "learning_rate": 0.0001498076923076923,
+      "loss": 0.1784,
+      "step": 7788
+    },
+    {
+      "epoch": 71.46,
+      "learning_rate": 0.00014975961538461538,
+      "loss": 0.2038,
+      "step": 7789
+    },
+    {
+      "epoch": 71.47,
+      "learning_rate": 0.00014971153846153845,
+      "loss": 0.2996,
+      "step": 7790
+    },
+    {
+      "epoch": 71.48,
+      "learning_rate": 0.00014966346153846155,
+      "loss": 0.2072,
+      "step": 7791
+    },
+    {
+      "epoch": 71.49,
+      "learning_rate": 0.00014961538461538462,
+      "loss": 0.1902,
+      "step": 7792
+    },
+    {
+      "epoch": 71.5,
+      "learning_rate": 0.0001495673076923077,
+      "loss": 0.2308,
+      "step": 7793
+    },
+    {
+      "epoch": 71.5,
+      "learning_rate": 0.00014951923076923077,
+      "loss": 0.3137,
+      "step": 7794
+    },
+    {
+      "epoch": 71.51,
+      "learning_rate": 0.00014947115384615384,
+      "loss": 0.2935,
+      "step": 7795
+    },
+    {
+      "epoch": 71.52,
+      "learning_rate": 0.00014942307692307694,
+      "loss": 0.244,
+      "step": 7796
+    },
+    {
+      "epoch": 71.53,
+      "learning_rate": 0.000149375,
+      "loss": 0.2166,
+      "step": 7797
+    },
+    {
+      "epoch": 71.54,
+      "learning_rate": 0.00014932692307692308,
+      "loss": 0.1978,
+      "step": 7798
+    },
+    {
+      "epoch": 71.55,
+      "learning_rate": 0.00014927884615384616,
+      "loss": 0.191,
+      "step": 7799
+    },
+    {
+      "epoch": 71.56,
+      "learning_rate": 0.00014923076923076923,
+      "loss": 0.1626,
+      "step": 7800
+    },
+    {
+      "epoch": 71.57,
+      "learning_rate": 0.0001491826923076923,
+      "loss": 0.1844,
+      "step": 7801
+    },
+    {
+      "epoch": 71.58,
+      "learning_rate": 0.0001491346153846154,
+      "loss": 0.1939,
+      "step": 7802
+    },
+    {
+      "epoch": 71.59,
+      "learning_rate": 0.00014908653846153848,
+      "loss": 0.2199,
+      "step": 7803
+    },
+    {
+      "epoch": 71.6,
+      "learning_rate": 0.00014903846153846155,
+      "loss": 0.2584,
+      "step": 7804
+    },
+    {
+      "epoch": 71.61,
+      "learning_rate": 0.00014899038461538462,
+      "loss": 0.2536,
+      "step": 7805
+    },
+    {
+      "epoch": 71.61,
+      "learning_rate": 0.0001489423076923077,
+      "loss": 0.1588,
+      "step": 7806
+    },
+    {
+      "epoch": 71.62,
+      "learning_rate": 0.0001488942307692308,
+      "loss": 0.2019,
+      "step": 7807
+    },
+    {
+      "epoch": 71.63,
+      "learning_rate": 0.00014884615384615387,
+      "loss": 0.2462,
+      "step": 7808
+    },
+    {
+      "epoch": 71.64,
+      "learning_rate": 0.00014879807692307694,
+      "loss": 0.2387,
+      "step": 7809
+    },
+    {
+      "epoch": 71.65,
+      "learning_rate": 0.00014874999999999998,
+      "loss": 0.2794,
+      "step": 7810
+    },
+    {
+      "epoch": 71.66,
+      "learning_rate": 0.00014870192307692306,
+      "loss": 0.2561,
+      "step": 7811
+    },
+    {
+      "epoch": 71.67,
+      "learning_rate": 0.00014865384615384616,
+      "loss": 0.2083,
+      "step": 7812
+    },
+    {
+      "epoch": 71.68,
+      "learning_rate": 0.00014860576923076923,
+      "loss": 0.1851,
+      "step": 7813
+    },
+    {
+      "epoch": 71.69,
+      "learning_rate": 0.0001485576923076923,
+      "loss": 0.2282,
+      "step": 7814
+    },
+    {
+      "epoch": 71.7,
+      "learning_rate": 0.00014850961538461537,
+      "loss": 0.2663,
+      "step": 7815
+    },
+    {
+      "epoch": 71.71,
+      "learning_rate": 0.00014846153846153845,
+      "loss": 0.2029,
+      "step": 7816
+    },
+    {
+      "epoch": 71.72,
+      "learning_rate": 0.00014841346153846155,
+      "loss": 0.2821,
+      "step": 7817
+    },
+    {
+      "epoch": 71.72,
+      "learning_rate": 0.00014836538461538462,
+      "loss": 0.2685,
+      "step": 7818
+    },
+    {
+      "epoch": 71.73,
+      "learning_rate": 0.0001483173076923077,
+      "loss": 0.203,
+      "step": 7819
+    },
+    {
+      "epoch": 71.74,
+      "learning_rate": 0.00014826923076923076,
+      "loss": 0.1493,
+      "step": 7820
+    },
+    {
+      "epoch": 71.75,
+      "learning_rate": 0.00014822115384615384,
+      "loss": 0.2953,
+      "step": 7821
+    },
+    {
+      "epoch": 71.76,
+      "learning_rate": 0.00014817307692307694,
+      "loss": 0.3136,
+      "step": 7822
+    },
+    {
+      "epoch": 71.77,
+      "learning_rate": 0.000148125,
+      "loss": 0.1856,
+      "step": 7823
+    },
+    {
+      "epoch": 71.78,
+      "learning_rate": 0.00014807692307692308,
+      "loss": 0.2346,
+      "step": 7824
+    },
+    {
+      "epoch": 71.79,
+      "learning_rate": 0.00014802884615384615,
+      "loss": 0.1981,
+      "step": 7825
+    },
+    {
+      "epoch": 71.8,
+      "learning_rate": 0.00014798076923076923,
+      "loss": 0.2106,
+      "step": 7826
+    },
+    {
+      "epoch": 71.81,
+      "learning_rate": 0.00014793269230769233,
+      "loss": 0.2318,
+      "step": 7827
+    },
+    {
+      "epoch": 71.82,
+      "learning_rate": 0.0001478846153846154,
+      "loss": 0.1676,
+      "step": 7828
+    },
+    {
+      "epoch": 71.83,
+      "learning_rate": 0.00014783653846153847,
+      "loss": 0.1222,
+      "step": 7829
+    },
+    {
+      "epoch": 71.83,
+      "learning_rate": 0.00014778846153846154,
+      "loss": 0.205,
+      "step": 7830
+    },
+    {
+      "epoch": 71.84,
+      "learning_rate": 0.00014774038461538462,
+      "loss": 0.2679,
+      "step": 7831
+    },
+    {
+      "epoch": 71.85,
+      "learning_rate": 0.00014769230769230772,
+      "loss": 0.2218,
+      "step": 7832
+    },
+    {
+      "epoch": 71.86,
+      "learning_rate": 0.0001476442307692308,
+      "loss": 0.2258,
+      "step": 7833
+    },
+    {
+      "epoch": 71.87,
+      "learning_rate": 0.00014759615384615386,
+      "loss": 0.3078,
+      "step": 7834
+    },
+    {
+      "epoch": 71.88,
+      "learning_rate": 0.00014754807692307693,
+      "loss": 0.245,
+      "step": 7835
+    },
+    {
+      "epoch": 71.89,
+      "learning_rate": 0.0001475,
+      "loss": 0.2602,
+      "step": 7836
+    },
+    {
+      "epoch": 71.9,
+      "learning_rate": 0.00014745192307692308,
+      "loss": 0.1766,
+      "step": 7837
+    },
+    {
+      "epoch": 71.91,
+      "learning_rate": 0.00014740384615384615,
+      "loss": 0.157,
+      "step": 7838
+    },
+    {
+      "epoch": 71.92,
+      "learning_rate": 0.00014735576923076923,
+      "loss": 0.2921,
+      "step": 7839
+    },
+    {
+      "epoch": 71.93,
+      "learning_rate": 0.0001473076923076923,
+      "loss": 0.1931,
+      "step": 7840
+    },
+    {
+      "epoch": 71.94,
+      "learning_rate": 0.00014725961538461537,
+      "loss": 0.1947,
+      "step": 7841
+    },
+    {
+      "epoch": 71.94,
+      "learning_rate": 0.00014721153846153844,
+      "loss": 0.232,
+      "step": 7842
+    },
+    {
+      "epoch": 71.95,
+      "learning_rate": 0.00014716346153846154,
+      "loss": 0.1933,
+      "step": 7843
+    },
+    {
+      "epoch": 71.96,
+      "learning_rate": 0.00014711538461538462,
+      "loss": 0.2048,
+      "step": 7844
+    },
+    {
+      "epoch": 71.97,
+      "learning_rate": 0.0001470673076923077,
+      "loss": 0.2,
+      "step": 7845
+    },
+    {
+      "epoch": 71.98,
+      "learning_rate": 0.00014701923076923076,
+      "loss": 0.1539,
+      "step": 7846
+    },
+    {
+      "epoch": 71.99,
+      "learning_rate": 0.00014697115384615383,
+      "loss": 0.1294,
+      "step": 7847
+    },
+    {
+      "epoch": 72.0,
+      "learning_rate": 0.00014692307692307693,
+      "loss": 0.1889,
+      "step": 7848
+    },
+    {
+      "epoch": 72.01,
+      "learning_rate": 0.000146875,
+      "loss": 0.2765,
+      "step": 7849
+    },
+    {
+      "epoch": 72.02,
+      "learning_rate": 0.00014682692307692308,
+      "loss": 0.2109,
+      "step": 7850
+    },
+    {
+      "epoch": 72.03,
+      "learning_rate": 0.00014677884615384615,
+      "loss": 0.2193,
+      "step": 7851
+    },
+    {
+      "epoch": 72.04,
+      "learning_rate": 0.00014673076923076922,
+      "loss": 0.3388,
+      "step": 7852
+    },
+    {
+      "epoch": 72.05,
+      "learning_rate": 0.00014668269230769232,
+      "loss": 0.1981,
+      "step": 7853
+    },
+    {
+      "epoch": 72.06,
+      "learning_rate": 0.0001466346153846154,
+      "loss": 0.1617,
+      "step": 7854
+    },
+    {
+      "epoch": 72.06,
+      "learning_rate": 0.00014658653846153847,
+      "loss": 0.2177,
+      "step": 7855
+    },
+    {
+      "epoch": 72.07,
+      "learning_rate": 0.00014653846153846154,
+      "loss": 0.1478,
+      "step": 7856
+    },
+    {
+      "epoch": 72.08,
+      "learning_rate": 0.00014649038461538461,
+      "loss": 0.1809,
+      "step": 7857
+    },
+    {
+      "epoch": 72.09,
+      "learning_rate": 0.00014644230769230771,
+      "loss": 0.1803,
+      "step": 7858
+    },
+    {
+      "epoch": 72.1,
+      "learning_rate": 0.00014639423076923079,
+      "loss": 0.1956,
+      "step": 7859
+    },
+    {
+      "epoch": 72.11,
+      "learning_rate": 0.00014634615384615386,
+      "loss": 0.1896,
+      "step": 7860
+    },
+    {
+      "epoch": 72.12,
+      "learning_rate": 0.00014629807692307693,
+      "loss": 0.2741,
+      "step": 7861
+    },
+    {
+      "epoch": 72.13,
+      "learning_rate": 0.00014625,
+      "loss": 0.2724,
+      "step": 7862
+    },
+    {
+      "epoch": 72.14,
+      "learning_rate": 0.0001462019230769231,
+      "loss": 0.1449,
+      "step": 7863
+    },
+    {
+      "epoch": 72.15,
+      "learning_rate": 0.00014615384615384618,
+      "loss": 0.2419,
+      "step": 7864
+    },
+    {
+      "epoch": 72.16,
+      "learning_rate": 0.00014610576923076925,
+      "loss": 0.1873,
+      "step": 7865
+    },
+    {
+      "epoch": 72.17,
+      "learning_rate": 0.0001460576923076923,
+      "loss": 0.2492,
+      "step": 7866
+    },
+    {
+      "epoch": 72.17,
+      "learning_rate": 0.00014600961538461537,
+      "loss": 0.2419,
+      "step": 7867
+    },
+    {
+      "epoch": 72.18,
+      "learning_rate": 0.00014596153846153844,
+      "loss": 0.228,
+      "step": 7868
+    },
+    {
+      "epoch": 72.19,
+      "learning_rate": 0.00014591346153846154,
+      "loss": 0.1612,
+      "step": 7869
+    },
+    {
+      "epoch": 72.2,
+      "learning_rate": 0.0001458653846153846,
+      "loss": 0.1876,
+      "step": 7870
+    },
+    {
+      "epoch": 72.21,
+      "learning_rate": 0.00014581730769230768,
+      "loss": 0.1253,
+      "step": 7871
+    },
+    {
+      "epoch": 72.22,
+      "learning_rate": 0.00014576923076923076,
+      "loss": 0.2024,
+      "step": 7872
+    },
+    {
+      "epoch": 72.23,
+      "learning_rate": 0.00014572115384615383,
+      "loss": 0.1482,
+      "step": 7873
+    },
+    {
+      "epoch": 72.24,
+      "learning_rate": 0.00014567307692307693,
+      "loss": 0.2657,
+      "step": 7874
+    },
+    {
+      "epoch": 72.25,
+      "learning_rate": 0.000145625,
+      "loss": 0.0883,
+      "step": 7875
+    },
+    {
+      "epoch": 72.26,
+      "learning_rate": 0.00014557692307692308,
+      "loss": 0.2742,
+      "step": 7876
+    },
+    {
+      "epoch": 72.27,
+      "learning_rate": 0.00014552884615384615,
+      "loss": 0.1971,
+      "step": 7877
+    },
+    {
+      "epoch": 72.28,
+      "learning_rate": 0.00014548076923076922,
+      "loss": 0.1966,
+      "step": 7878
+    },
+    {
+      "epoch": 72.28,
+      "learning_rate": 0.00014543269230769232,
+      "loss": 0.2366,
+      "step": 7879
+    },
+    {
+      "epoch": 72.29,
+      "learning_rate": 0.0001453846153846154,
+      "loss": 0.2962,
+      "step": 7880
+    },
+    {
+      "epoch": 72.3,
+      "learning_rate": 0.00014533653846153847,
+      "loss": 0.2301,
+      "step": 7881
+    },
+    {
+      "epoch": 72.31,
+      "learning_rate": 0.00014528846153846154,
+      "loss": 0.1549,
+      "step": 7882
+    },
+    {
+      "epoch": 72.32,
+      "learning_rate": 0.0001452403846153846,
+      "loss": 0.1661,
+      "step": 7883
+    },
+    {
+      "epoch": 72.33,
+      "learning_rate": 0.0001451923076923077,
+      "loss": 0.2407,
+      "step": 7884
+    },
+    {
+      "epoch": 72.34,
+      "learning_rate": 0.00014514423076923078,
+      "loss": 0.1986,
+      "step": 7885
+    },
+    {
+      "epoch": 72.35,
+      "learning_rate": 0.00014509615384615386,
+      "loss": 0.2148,
+      "step": 7886
+    },
+    {
+      "epoch": 72.36,
+      "learning_rate": 0.00014504807692307693,
+      "loss": 0.2196,
+      "step": 7887
+    },
+    {
+      "epoch": 72.37,
+      "learning_rate": 0.000145,
+      "loss": 0.1995,
+      "step": 7888
+    },
+    {
+      "epoch": 72.38,
+      "learning_rate": 0.0001449519230769231,
+      "loss": 0.1853,
+      "step": 7889
+    },
+    {
+      "epoch": 72.39,
+      "learning_rate": 0.00014490384615384617,
+      "loss": 0.2479,
+      "step": 7890
+    },
+    {
+      "epoch": 72.39,
+      "learning_rate": 0.00014485576923076925,
+      "loss": 0.2591,
+      "step": 7891
+    },
+    {
+      "epoch": 72.4,
+      "learning_rate": 0.00014480769230769232,
+      "loss": 0.2508,
+      "step": 7892
+    },
+    {
+      "epoch": 72.41,
+      "learning_rate": 0.0001447596153846154,
+      "loss": 0.1525,
+      "step": 7893
+    },
+    {
+      "epoch": 72.42,
+      "learning_rate": 0.0001447115384615385,
+      "loss": 0.2104,
+      "step": 7894
+    },
+    {
+      "epoch": 72.43,
+      "learning_rate": 0.00014466346153846154,
+      "loss": 0.2418,
+      "step": 7895
+    },
+    {
+      "epoch": 72.44,
+      "learning_rate": 0.0001446153846153846,
+      "loss": 0.1141,
+      "step": 7896
+    },
+    {
+      "epoch": 72.45,
+      "learning_rate": 0.00014456730769230768,
+      "loss": 0.1637,
+      "step": 7897
+    },
+    {
+      "epoch": 72.46,
+      "learning_rate": 0.00014451923076923075,
+      "loss": 0.2624,
+      "step": 7898
+    },
+    {
+      "epoch": 72.47,
+      "learning_rate": 0.00014447115384615383,
+      "loss": 0.2039,
+      "step": 7899
+    },
+    {
+      "epoch": 72.48,
+      "learning_rate": 0.00014442307692307693,
+      "loss": 0.2308,
+      "step": 7900
+    },
+    {
+      "epoch": 72.49,
+      "learning_rate": 0.000144375,
+      "loss": 0.1385,
+      "step": 7901
+    },
+    {
+      "epoch": 72.5,
+      "learning_rate": 0.00014432692307692307,
+      "loss": 0.0926,
+      "step": 7902
+    },
+    {
+      "epoch": 72.5,
+      "learning_rate": 0.00014427884615384614,
+      "loss": 0.2697,
+      "step": 7903
+    },
+    {
+      "epoch": 72.51,
+      "learning_rate": 0.00014423076923076922,
+      "loss": 0.3302,
+      "step": 7904
+    },
+    {
+      "epoch": 72.52,
+      "learning_rate": 0.00014418269230769232,
+      "loss": 0.1803,
+      "step": 7905
+    },
+    {
+      "epoch": 72.53,
+      "learning_rate": 0.0001441346153846154,
+      "loss": 0.2603,
+      "step": 7906
+    },
+    {
+      "epoch": 72.54,
+      "learning_rate": 0.00014408653846153846,
+      "loss": 0.194,
+      "step": 7907
+    },
+    {
+      "epoch": 72.55,
+      "learning_rate": 0.00014403846153846153,
+      "loss": 0.2416,
+      "step": 7908
+    },
+    {
+      "epoch": 72.56,
+      "learning_rate": 0.0001439903846153846,
+      "loss": 0.2526,
+      "step": 7909
+    },
+    {
+      "epoch": 72.57,
+      "learning_rate": 0.0001439423076923077,
+      "loss": 0.1382,
+      "step": 7910
+    },
+    {
+      "epoch": 72.58,
+      "learning_rate": 0.00014389423076923078,
+      "loss": 0.2999,
+      "step": 7911
+    },
+    {
+      "epoch": 72.59,
+      "learning_rate": 0.00014384615384615385,
+      "loss": 0.1967,
+      "step": 7912
+    },
+    {
+      "epoch": 72.6,
+      "learning_rate": 0.00014379807692307692,
+      "loss": 0.1628,
+      "step": 7913
+    },
+    {
+      "epoch": 72.61,
+      "learning_rate": 0.00014375,
+      "loss": 0.1878,
+      "step": 7914
+    },
+    {
+      "epoch": 72.61,
+      "learning_rate": 0.0001437019230769231,
+      "loss": 0.2128,
+      "step": 7915
+    },
+    {
+      "epoch": 72.62,
+      "learning_rate": 0.00014365384615384617,
+      "loss": 0.2576,
+      "step": 7916
+    },
+    {
+      "epoch": 72.63,
+      "learning_rate": 0.00014360576923076924,
+      "loss": 0.1961,
+      "step": 7917
+    },
+    {
+      "epoch": 72.64,
+      "learning_rate": 0.00014355769230769232,
+      "loss": 0.1969,
+      "step": 7918
+    },
+    {
+      "epoch": 72.65,
+      "learning_rate": 0.0001435096153846154,
+      "loss": 0.1938,
+      "step": 7919
+    },
+    {
+      "epoch": 72.66,
+      "learning_rate": 0.0001434615384615385,
+      "loss": 0.2061,
+      "step": 7920
+    },
+    {
+      "epoch": 72.67,
+      "learning_rate": 0.00014341346153846156,
+      "loss": 0.1381,
+      "step": 7921
+    },
+    {
+      "epoch": 72.68,
+      "learning_rate": 0.00014336538461538463,
+      "loss": 0.2431,
+      "step": 7922
+    },
+    {
+      "epoch": 72.69,
+      "learning_rate": 0.00014331730769230768,
+      "loss": 0.2135,
+      "step": 7923
+    },
+    {
+      "epoch": 72.7,
+      "learning_rate": 0.00014326923076923075,
+      "loss": 0.298,
+      "step": 7924
+    },
+    {
+      "epoch": 72.71,
+      "learning_rate": 0.00014322115384615385,
+      "loss": 0.2811,
+      "step": 7925
+    },
+    {
+      "epoch": 72.72,
+      "learning_rate": 0.00014317307692307692,
+      "loss": 0.2569,
+      "step": 7926
+    },
+    {
+      "epoch": 72.72,
+      "learning_rate": 0.000143125,
+      "loss": 0.1891,
+      "step": 7927
+    },
+    {
+      "epoch": 72.73,
+      "learning_rate": 0.00014307692307692307,
+      "loss": 0.141,
+      "step": 7928
+    },
+    {
+      "epoch": 72.74,
+      "learning_rate": 0.00014302884615384614,
+      "loss": 0.1712,
+      "step": 7929
+    },
+    {
+      "epoch": 72.75,
+      "learning_rate": 0.00014298076923076924,
+      "loss": 0.3618,
+      "step": 7930
+    },
+    {
+      "epoch": 72.76,
+      "learning_rate": 0.0001429326923076923,
+      "loss": 0.2383,
+      "step": 7931
+    },
+    {
+      "epoch": 72.77,
+      "learning_rate": 0.00014288461538461539,
+      "loss": 0.2307,
+      "step": 7932
+    },
+    {
+      "epoch": 72.78,
+      "learning_rate": 0.00014283653846153846,
+      "loss": 0.2384,
+      "step": 7933
+    },
+    {
+      "epoch": 72.79,
+      "learning_rate": 0.00014278846153846153,
+      "loss": 0.2781,
+      "step": 7934
+    },
+    {
+      "epoch": 72.8,
+      "learning_rate": 0.0001427403846153846,
+      "loss": 0.2401,
+      "step": 7935
+    },
+    {
+      "epoch": 72.81,
+      "learning_rate": 0.0001426923076923077,
+      "loss": 0.1904,
+      "step": 7936
+    },
+    {
+      "epoch": 72.82,
+      "learning_rate": 0.00014264423076923078,
+      "loss": 0.2276,
+      "step": 7937
+    },
+    {
+      "epoch": 72.83,
+      "learning_rate": 0.00014259615384615385,
+      "loss": 0.2176,
+      "step": 7938
+    },
+    {
+      "epoch": 72.83,
+      "learning_rate": 0.00014254807692307692,
+      "loss": 0.2305,
+      "step": 7939
+    },
+    {
+      "epoch": 72.84,
+      "learning_rate": 0.0001425,
+      "loss": 0.29,
+      "step": 7940
+    },
+    {
+      "epoch": 72.85,
+      "learning_rate": 0.0001424519230769231,
+      "loss": 0.2715,
+      "step": 7941
+    },
+    {
+      "epoch": 72.86,
+      "learning_rate": 0.00014240384615384617,
+      "loss": 0.1973,
+      "step": 7942
+    },
+    {
+      "epoch": 72.87,
+      "learning_rate": 0.00014235576923076924,
+      "loss": 0.1593,
+      "step": 7943
+    },
+    {
+      "epoch": 72.88,
+      "learning_rate": 0.0001423076923076923,
+      "loss": 0.1932,
+      "step": 7944
+    },
+    {
+      "epoch": 72.89,
+      "learning_rate": 0.00014225961538461538,
+      "loss": 0.3017,
+      "step": 7945
+    },
+    {
+      "epoch": 72.9,
+      "learning_rate": 0.00014221153846153848,
+      "loss": 0.2095,
+      "step": 7946
+    },
+    {
+      "epoch": 72.91,
+      "learning_rate": 0.00014216346153846156,
+      "loss": 0.1952,
+      "step": 7947
+    },
+    {
+      "epoch": 72.92,
+      "learning_rate": 0.00014211538461538463,
+      "loss": 0.1997,
+      "step": 7948
+    },
+    {
+      "epoch": 72.93,
+      "learning_rate": 0.0001420673076923077,
+      "loss": 0.1617,
+      "step": 7949
+    },
+    {
+      "epoch": 72.94,
+      "learning_rate": 0.00014201923076923077,
+      "loss": 0.2065,
+      "step": 7950
+    },
+    {
+      "epoch": 72.94,
+      "learning_rate": 0.00014197115384615387,
+      "loss": 0.2315,
+      "step": 7951
+    },
+    {
+      "epoch": 72.95,
+      "learning_rate": 0.00014192307692307692,
+      "loss": 0.1158,
+      "step": 7952
+    },
+    {
+      "epoch": 72.96,
+      "learning_rate": 0.000141875,
+      "loss": 0.2398,
+      "step": 7953
+    },
+    {
+      "epoch": 72.97,
+      "learning_rate": 0.00014182692307692307,
+      "loss": 0.2002,
+      "step": 7954
+    },
+    {
+      "epoch": 72.98,
+      "learning_rate": 0.00014177884615384614,
+      "loss": 0.2214,
+      "step": 7955
+    },
+    {
+      "epoch": 72.99,
+      "learning_rate": 0.00014173076923076924,
+      "loss": 0.0811,
+      "step": 7956
+    },
+    {
+      "epoch": 73.0,
+      "learning_rate": 0.0001416826923076923,
+      "loss": 0.2132,
+      "step": 7957
+    },
+    {
+      "epoch": 73.01,
+      "learning_rate": 0.00014163461538461538,
+      "loss": 0.2774,
+      "step": 7958
+    },
+    {
+      "epoch": 73.02,
+      "learning_rate": 0.00014158653846153846,
+      "loss": 0.2888,
+      "step": 7959
+    },
+    {
+      "epoch": 73.03,
+      "learning_rate": 0.00014153846153846153,
+      "loss": 0.2346,
+      "step": 7960
+    },
+    {
+      "epoch": 73.04,
+      "learning_rate": 0.00014149038461538463,
+      "loss": 0.277,
+      "step": 7961
+    },
+    {
+      "epoch": 73.05,
+      "learning_rate": 0.0001414423076923077,
+      "loss": 0.1684,
+      "step": 7962
+    },
+    {
+      "epoch": 73.06,
+      "learning_rate": 0.00014139423076923077,
+      "loss": 0.3088,
+      "step": 7963
+    },
+    {
+      "epoch": 73.06,
+      "learning_rate": 0.00014134615384615385,
+      "loss": 0.1743,
+      "step": 7964
+    },
+    {
+      "epoch": 73.07,
+      "learning_rate": 0.00014129807692307692,
+      "loss": 0.1479,
+      "step": 7965
+    },
+    {
+      "epoch": 73.08,
+      "learning_rate": 0.00014125,
+      "loss": 0.2011,
+      "step": 7966
+    },
+    {
+      "epoch": 73.09,
+      "learning_rate": 0.0001412019230769231,
+      "loss": 0.2007,
+      "step": 7967
+    },
+    {
+      "epoch": 73.1,
+      "learning_rate": 0.00014115384615384616,
+      "loss": 0.1673,
+      "step": 7968
+    },
+    {
+      "epoch": 73.11,
+      "learning_rate": 0.00014110576923076924,
+      "loss": 0.2352,
+      "step": 7969
+    },
+    {
+      "epoch": 73.12,
+      "learning_rate": 0.0001410576923076923,
+      "loss": 0.1913,
+      "step": 7970
+    },
+    {
+      "epoch": 73.13,
+      "learning_rate": 0.00014100961538461538,
+      "loss": 0.0961,
+      "step": 7971
+    },
+    {
+      "epoch": 73.14,
+      "learning_rate": 0.00014096153846153848,
+      "loss": 0.2172,
+      "step": 7972
+    },
+    {
+      "epoch": 73.15,
+      "learning_rate": 0.00014091346153846155,
+      "loss": 0.229,
+      "step": 7973
+    },
+    {
+      "epoch": 73.16,
+      "learning_rate": 0.00014086538461538463,
+      "loss": 0.152,
+      "step": 7974
+    },
+    {
+      "epoch": 73.17,
+      "learning_rate": 0.0001408173076923077,
+      "loss": 0.1683,
+      "step": 7975
+    },
+    {
+      "epoch": 73.17,
+      "learning_rate": 0.00014076923076923077,
+      "loss": 0.2263,
+      "step": 7976
+    },
+    {
+      "epoch": 73.18,
+      "learning_rate": 0.00014072115384615387,
+      "loss": 0.1348,
+      "step": 7977
+    },
+    {
+      "epoch": 73.19,
+      "learning_rate": 0.00014067307692307694,
+      "loss": 0.1664,
+      "step": 7978
+    },
+    {
+      "epoch": 73.2,
+      "learning_rate": 0.00014062500000000002,
+      "loss": 0.1249,
+      "step": 7979
+    },
+    {
+      "epoch": 73.21,
+      "learning_rate": 0.00014057692307692306,
+      "loss": 0.1614,
+      "step": 7980
+    },
+    {
+      "epoch": 73.22,
+      "learning_rate": 0.00014052884615384613,
+      "loss": 0.2194,
+      "step": 7981
+    },
+    {
+      "epoch": 73.23,
+      "learning_rate": 0.00014048076923076923,
+      "loss": 0.1982,
+      "step": 7982
+    },
+    {
+      "epoch": 73.24,
+      "learning_rate": 0.0001404326923076923,
+      "loss": 0.2176,
+      "step": 7983
+    },
+    {
+      "epoch": 73.25,
+      "learning_rate": 0.00014038461538461538,
+      "loss": 0.1779,
+      "step": 7984
+    },
+    {
+      "epoch": 73.26,
+      "learning_rate": 0.00014033653846153845,
+      "loss": 0.3007,
+      "step": 7985
+    },
+    {
+      "epoch": 73.27,
+      "learning_rate": 0.00014028846153846152,
+      "loss": 0.2179,
+      "step": 7986
+    },
+    {
+      "epoch": 73.28,
+      "learning_rate": 0.00014024038461538462,
+      "loss": 0.2548,
+      "step": 7987
+    },
+    {
+      "epoch": 73.28,
+      "learning_rate": 0.0001401923076923077,
+      "loss": 0.1818,
+      "step": 7988
+    },
+    {
+      "epoch": 73.29,
+      "learning_rate": 0.00014014423076923077,
+      "loss": 0.1975,
+      "step": 7989
+    },
+    {
+      "epoch": 73.3,
+      "learning_rate": 0.00014009615384615384,
+      "loss": 0.1809,
+      "step": 7990
+    },
+    {
+      "epoch": 73.31,
+      "learning_rate": 0.00014004807692307692,
+      "loss": 0.2404,
+      "step": 7991
+    },
+    {
+      "epoch": 73.32,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 0.1444,
+      "step": 7992
+    },
+    {
+      "epoch": 73.33,
+      "learning_rate": 0.0001399519230769231,
+      "loss": 0.2333,
+      "step": 7993
+    },
+    {
+      "epoch": 73.34,
+      "learning_rate": 0.00013990384615384616,
+      "loss": 0.1823,
+      "step": 7994
+    },
+    {
+      "epoch": 73.35,
+      "learning_rate": 0.00013985576923076923,
+      "loss": 0.1939,
+      "step": 7995
+    },
+    {
+      "epoch": 73.36,
+      "learning_rate": 0.0001398076923076923,
+      "loss": 0.1816,
+      "step": 7996
+    },
+    {
+      "epoch": 73.37,
+      "learning_rate": 0.00013975961538461538,
+      "loss": 0.1399,
+      "step": 7997
+    },
+    {
+      "epoch": 73.38,
+      "learning_rate": 0.00013971153846153848,
+      "loss": 0.2234,
+      "step": 7998
+    },
+    {
+      "epoch": 73.39,
+      "learning_rate": 0.00013966346153846155,
+      "loss": 0.1543,
+      "step": 7999
+    },
+    {
+      "epoch": 73.39,
+      "learning_rate": 0.00013961538461538462,
+      "loss": 0.2626,
+      "step": 8000
+    },
+    {
+      "epoch": 73.39,
+      "eval_cer": 0.14539948125159297,
+      "eval_loss": 0.6843499541282654,
+      "eval_runtime": 87.0509,
+      "eval_samples_per_second": 18.92,
+      "eval_steps_per_second": 2.366,
+      "eval_wer": 0.5043407210703708,
+      "step": 8000
+    },
+    {
+      "epoch": 73.4,
+      "learning_rate": 0.0001395673076923077,
+      "loss": 0.2407,
+      "step": 8001
+    },
+    {
+      "epoch": 73.41,
+      "learning_rate": 0.00013951923076923077,
+      "loss": 0.1714,
+      "step": 8002
+    },
+    {
+      "epoch": 73.42,
+      "learning_rate": 0.00013947115384615387,
+      "loss": 0.1235,
+      "step": 8003
+    },
+    {
+      "epoch": 73.43,
+      "learning_rate": 0.00013942307692307694,
+      "loss": 0.2294,
+      "step": 8004
+    },
+    {
+      "epoch": 73.44,
+      "learning_rate": 0.000139375,
+      "loss": 0.2152,
+      "step": 8005
+    },
+    {
+      "epoch": 73.45,
+      "learning_rate": 0.00013932692307692309,
+      "loss": 0.23,
+      "step": 8006
+    },
+    {
+      "epoch": 73.46,
+      "learning_rate": 0.00013927884615384613,
+      "loss": 0.2082,
+      "step": 8007
+    },
+    {
+      "epoch": 73.47,
+      "learning_rate": 0.00013923076923076923,
+      "loss": 0.1758,
+      "step": 8008
+    },
+    {
+      "epoch": 73.48,
+      "learning_rate": 0.0001391826923076923,
+      "loss": 0.1838,
+      "step": 8009
+    },
+    {
+      "epoch": 73.49,
+      "learning_rate": 0.00013913461538461538,
+      "loss": 0.2155,
+      "step": 8010
+    },
+    {
+      "epoch": 73.5,
+      "learning_rate": 0.00013908653846153845,
+      "loss": 0.0891,
+      "step": 8011
+    },
+    {
+      "epoch": 73.5,
+      "learning_rate": 0.00013903846153846152,
+      "loss": 0.2283,
+      "step": 8012
+    },
+    {
+      "epoch": 73.51,
+      "learning_rate": 0.00013899038461538462,
+      "loss": 0.2882,
+      "step": 8013
+    },
+    {
+      "epoch": 73.52,
+      "learning_rate": 0.0001389423076923077,
+      "loss": 0.1777,
+      "step": 8014
+    },
+    {
+      "epoch": 73.53,
+      "learning_rate": 0.00013889423076923077,
+      "loss": 0.1918,
+      "step": 8015
+    },
+    {
+      "epoch": 73.54,
+      "learning_rate": 0.00013884615384615384,
+      "loss": 0.1619,
+      "step": 8016
+    },
+    {
+      "epoch": 73.55,
+      "learning_rate": 0.0001387980769230769,
+      "loss": 0.1736,
+      "step": 8017
+    },
+    {
+      "epoch": 73.56,
+      "learning_rate": 0.00013875,
+      "loss": 0.1502,
+      "step": 8018
+    },
+    {
+      "epoch": 73.57,
+      "learning_rate": 0.00013870192307692308,
+      "loss": 0.1786,
+      "step": 8019
+    },
+    {
+      "epoch": 73.58,
+      "learning_rate": 0.00013865384615384616,
+      "loss": 0.2593,
+      "step": 8020
+    },
+    {
+      "epoch": 73.59,
+      "learning_rate": 0.00013860576923076923,
+      "loss": 0.1541,
+      "step": 8021
+    },
+    {
+      "epoch": 73.6,
+      "learning_rate": 0.0001385576923076923,
+      "loss": 0.3271,
+      "step": 8022
+    },
+    {
+      "epoch": 73.61,
+      "learning_rate": 0.0001385096153846154,
+      "loss": 0.175,
+      "step": 8023
+    },
+    {
+      "epoch": 73.61,
+      "learning_rate": 0.00013846153846153847,
+      "loss": 0.1749,
+      "step": 8024
+    },
+    {
+      "epoch": 73.62,
+      "learning_rate": 0.00013841346153846155,
+      "loss": 0.2655,
+      "step": 8025
+    },
+    {
+      "epoch": 73.63,
+      "learning_rate": 0.00013836538461538462,
+      "loss": 0.1247,
+      "step": 8026
+    },
+    {
+      "epoch": 73.64,
+      "learning_rate": 0.0001383173076923077,
+      "loss": 0.2464,
+      "step": 8027
+    },
+    {
+      "epoch": 73.65,
+      "learning_rate": 0.00013826923076923076,
+      "loss": 0.1656,
+      "step": 8028
+    },
+    {
+      "epoch": 73.66,
+      "learning_rate": 0.00013822115384615386,
+      "loss": 0.1174,
+      "step": 8029
+    },
+    {
+      "epoch": 73.67,
+      "learning_rate": 0.00013817307692307694,
+      "loss": 0.205,
+      "step": 8030
+    },
+    {
+      "epoch": 73.68,
+      "learning_rate": 0.000138125,
+      "loss": 0.1296,
+      "step": 8031
+    },
+    {
+      "epoch": 73.69,
+      "learning_rate": 0.00013807692307692308,
+      "loss": 0.2102,
+      "step": 8032
+    },
+    {
+      "epoch": 73.7,
+      "learning_rate": 0.00013802884615384616,
+      "loss": 0.1461,
+      "step": 8033
+    },
+    {
+      "epoch": 73.71,
+      "learning_rate": 0.00013798076923076925,
+      "loss": 0.1835,
+      "step": 8034
+    },
+    {
+      "epoch": 73.72,
+      "learning_rate": 0.00013793269230769233,
+      "loss": 0.2054,
+      "step": 8035
+    },
+    {
+      "epoch": 73.72,
+      "learning_rate": 0.00013788461538461537,
+      "loss": 0.2404,
+      "step": 8036
+    },
+    {
+      "epoch": 73.73,
+      "learning_rate": 0.00013783653846153845,
+      "loss": 0.1731,
+      "step": 8037
+    },
+    {
+      "epoch": 73.74,
+      "learning_rate": 0.00013778846153846152,
+      "loss": 0.1699,
+      "step": 8038
+    },
+    {
+      "epoch": 73.75,
+      "learning_rate": 0.00013774038461538462,
+      "loss": 0.2786,
+      "step": 8039
+    },
+    {
+      "epoch": 73.76,
+      "learning_rate": 0.0001376923076923077,
+      "loss": 0.2403,
+      "step": 8040
+    },
+    {
+      "epoch": 73.77,
+      "learning_rate": 0.00013764423076923076,
+      "loss": 0.2649,
+      "step": 8041
+    },
+    {
+      "epoch": 73.78,
+      "learning_rate": 0.00013759615384615384,
+      "loss": 0.268,
+      "step": 8042
+    },
+    {
+      "epoch": 73.79,
+      "learning_rate": 0.0001375480769230769,
+      "loss": 0.248,
+      "step": 8043
+    },
+    {
+      "epoch": 73.8,
+      "learning_rate": 0.0001375,
+      "loss": 0.1694,
+      "step": 8044
+    },
+    {
+      "epoch": 73.81,
+      "learning_rate": 0.00013745192307692308,
+      "loss": 0.2268,
+      "step": 8045
+    },
+    {
+      "epoch": 73.82,
+      "learning_rate": 0.00013740384615384615,
+      "loss": 0.1772,
+      "step": 8046
+    },
+    {
+      "epoch": 73.83,
+      "learning_rate": 0.00013735576923076923,
+      "loss": 0.1376,
+      "step": 8047
+    },
+    {
+      "epoch": 73.83,
+      "learning_rate": 0.0001373076923076923,
+      "loss": 0.2265,
+      "step": 8048
+    },
+    {
+      "epoch": 73.84,
+      "learning_rate": 0.0001372596153846154,
+      "loss": 0.2245,
+      "step": 8049
+    },
+    {
+      "epoch": 73.85,
+      "learning_rate": 0.00013721153846153847,
+      "loss": 0.186,
+      "step": 8050
+    },
+    {
+      "epoch": 73.86,
+      "learning_rate": 0.00013716346153846154,
+      "loss": 0.255,
+      "step": 8051
+    },
+    {
+      "epoch": 73.87,
+      "learning_rate": 0.00013711538461538462,
+      "loss": 0.1958,
+      "step": 8052
+    },
+    {
+      "epoch": 73.88,
+      "learning_rate": 0.0001370673076923077,
+      "loss": 0.2421,
+      "step": 8053
+    },
+    {
+      "epoch": 73.89,
+      "learning_rate": 0.0001370192307692308,
+      "loss": 0.1911,
+      "step": 8054
+    },
+    {
+      "epoch": 73.9,
+      "learning_rate": 0.00013697115384615386,
+      "loss": 0.1481,
+      "step": 8055
+    },
+    {
+      "epoch": 73.91,
+      "learning_rate": 0.00013692307692307693,
+      "loss": 0.2981,
+      "step": 8056
+    },
+    {
+      "epoch": 73.92,
+      "learning_rate": 0.000136875,
+      "loss": 0.2079,
+      "step": 8057
+    },
+    {
+      "epoch": 73.93,
+      "learning_rate": 0.00013682692307692308,
+      "loss": 0.2459,
+      "step": 8058
+    },
+    {
+      "epoch": 73.94,
+      "learning_rate": 0.00013677884615384618,
+      "loss": 0.1506,
+      "step": 8059
+    },
+    {
+      "epoch": 73.94,
+      "learning_rate": 0.00013673076923076925,
+      "loss": 0.2205,
+      "step": 8060
+    },
+    {
+      "epoch": 73.95,
+      "learning_rate": 0.00013668269230769232,
+      "loss": 0.149,
+      "step": 8061
+    },
+    {
+      "epoch": 73.96,
+      "learning_rate": 0.0001366346153846154,
+      "loss": 0.2445,
+      "step": 8062
+    },
+    {
+      "epoch": 73.97,
+      "learning_rate": 0.00013658653846153847,
+      "loss": 0.1503,
+      "step": 8063
+    },
+    {
+      "epoch": 73.98,
+      "learning_rate": 0.00013653846153846152,
+      "loss": 0.1378,
+      "step": 8064
+    },
+    {
+      "epoch": 73.99,
+      "learning_rate": 0.00013649038461538461,
+      "loss": 0.3169,
+      "step": 8065
+    },
+    {
+      "epoch": 74.0,
+      "learning_rate": 0.0001364423076923077,
+      "loss": 0.178,
+      "step": 8066
+    },
+    {
+      "epoch": 74.01,
+      "learning_rate": 0.00013639423076923076,
+      "loss": 0.2312,
+      "step": 8067
+    },
+    {
+      "epoch": 74.02,
+      "learning_rate": 0.00013634615384615383,
+      "loss": 0.3533,
+      "step": 8068
+    },
+    {
+      "epoch": 74.03,
+      "learning_rate": 0.0001362980769230769,
+      "loss": 0.1745,
+      "step": 8069
+    },
+    {
+      "epoch": 74.04,
+      "learning_rate": 0.00013625,
+      "loss": 0.207,
+      "step": 8070
+    },
+    {
+      "epoch": 74.05,
+      "learning_rate": 0.00013620192307692308,
+      "loss": 0.1802,
+      "step": 8071
+    },
+    {
+      "epoch": 74.06,
+      "learning_rate": 0.00013615384615384615,
+      "loss": 0.2392,
+      "step": 8072
+    },
+    {
+      "epoch": 74.06,
+      "learning_rate": 0.00013610576923076922,
+      "loss": 0.1562,
+      "step": 8073
+    },
+    {
+      "epoch": 74.07,
+      "learning_rate": 0.0001360576923076923,
+      "loss": 0.1979,
+      "step": 8074
+    },
+    {
+      "epoch": 74.08,
+      "learning_rate": 0.0001360096153846154,
+      "loss": 0.2163,
+      "step": 8075
+    },
+    {
+      "epoch": 74.09,
+      "learning_rate": 0.00013596153846153847,
+      "loss": 0.1485,
+      "step": 8076
+    },
+    {
+      "epoch": 74.1,
+      "learning_rate": 0.00013591346153846154,
+      "loss": 0.1528,
+      "step": 8077
+    },
+    {
+      "epoch": 74.11,
+      "learning_rate": 0.0001358653846153846,
+      "loss": 0.15,
+      "step": 8078
+    },
+    {
+      "epoch": 74.12,
+      "learning_rate": 0.00013581730769230769,
+      "loss": 0.2177,
+      "step": 8079
+    },
+    {
+      "epoch": 74.13,
+      "learning_rate": 0.00013576923076923079,
+      "loss": 0.1872,
+      "step": 8080
+    },
+    {
+      "epoch": 74.14,
+      "learning_rate": 0.00013572115384615386,
+      "loss": 0.1417,
+      "step": 8081
+    },
+    {
+      "epoch": 74.15,
+      "learning_rate": 0.00013567307692307693,
+      "loss": 0.148,
+      "step": 8082
+    },
+    {
+      "epoch": 74.16,
+      "learning_rate": 0.000135625,
+      "loss": 0.1294,
+      "step": 8083
+    },
+    {
+      "epoch": 74.17,
+      "learning_rate": 0.00013557692307692308,
+      "loss": 0.2105,
+      "step": 8084
+    },
+    {
+      "epoch": 74.17,
+      "learning_rate": 0.00013552884615384618,
+      "loss": 0.1603,
+      "step": 8085
+    },
+    {
+      "epoch": 74.18,
+      "learning_rate": 0.00013548076923076925,
+      "loss": 0.233,
+      "step": 8086
+    },
+    {
+      "epoch": 74.19,
+      "learning_rate": 0.00013543269230769232,
+      "loss": 0.1327,
+      "step": 8087
+    },
+    {
+      "epoch": 74.2,
+      "learning_rate": 0.0001353846153846154,
+      "loss": 0.2072,
+      "step": 8088
+    },
+    {
+      "epoch": 74.21,
+      "learning_rate": 0.00013533653846153847,
+      "loss": 0.1423,
+      "step": 8089
+    },
+    {
+      "epoch": 74.22,
+      "learning_rate": 0.00013528846153846157,
+      "loss": 0.2185,
+      "step": 8090
+    },
+    {
+      "epoch": 74.23,
+      "learning_rate": 0.00013524038461538464,
+      "loss": 0.1616,
+      "step": 8091
+    },
+    {
+      "epoch": 74.24,
+      "learning_rate": 0.0001351923076923077,
+      "loss": 0.2308,
+      "step": 8092
+    },
+    {
+      "epoch": 74.25,
+      "learning_rate": 0.00013514423076923076,
+      "loss": 0.1511,
+      "step": 8093
+    },
+    {
+      "epoch": 74.26,
+      "learning_rate": 0.00013509615384615383,
+      "loss": 0.2928,
+      "step": 8094
+    },
+    {
+      "epoch": 74.27,
+      "learning_rate": 0.0001350480769230769,
+      "loss": 0.1634,
+      "step": 8095
+    },
+    {
+      "epoch": 74.28,
+      "learning_rate": 0.000135,
+      "loss": 0.2226,
+      "step": 8096
+    },
+    {
+      "epoch": 74.28,
+      "learning_rate": 0.00013495192307692307,
+      "loss": 0.1701,
+      "step": 8097
+    },
+    {
+      "epoch": 74.29,
+      "learning_rate": 0.00013490384615384615,
+      "loss": 0.2066,
+      "step": 8098
+    },
+    {
+      "epoch": 74.3,
+      "learning_rate": 0.00013485576923076922,
+      "loss": 0.2021,
+      "step": 8099
+    },
+    {
+      "epoch": 74.31,
+      "learning_rate": 0.0001348076923076923,
+      "loss": 0.2088,
+      "step": 8100
+    },
+    {
+      "epoch": 74.32,
+      "learning_rate": 0.0001347596153846154,
+      "loss": 0.2052,
+      "step": 8101
+    },
+    {
+      "epoch": 74.33,
+      "learning_rate": 0.00013471153846153846,
+      "loss": 0.1272,
+      "step": 8102
+    },
+    {
+      "epoch": 74.34,
+      "learning_rate": 0.00013466346153846154,
+      "loss": 0.1778,
+      "step": 8103
+    },
+    {
+      "epoch": 74.35,
+      "learning_rate": 0.0001346153846153846,
+      "loss": 0.196,
+      "step": 8104
+    },
+    {
+      "epoch": 74.36,
+      "learning_rate": 0.00013456730769230768,
+      "loss": 0.2335,
+      "step": 8105
+    },
+    {
+      "epoch": 74.37,
+      "learning_rate": 0.00013451923076923078,
+      "loss": 0.1458,
+      "step": 8106
+    },
+    {
+      "epoch": 74.38,
+      "learning_rate": 0.00013447115384615385,
+      "loss": 0.1673,
+      "step": 8107
+    },
+    {
+      "epoch": 74.39,
+      "learning_rate": 0.00013442307692307693,
+      "loss": 0.2614,
+      "step": 8108
+    },
+    {
+      "epoch": 74.39,
+      "learning_rate": 0.000134375,
+      "loss": 0.2314,
+      "step": 8109
+    },
+    {
+      "epoch": 74.4,
+      "learning_rate": 0.00013432692307692307,
+      "loss": 0.2171,
+      "step": 8110
+    },
+    {
+      "epoch": 74.41,
+      "learning_rate": 0.00013427884615384617,
+      "loss": 0.1273,
+      "step": 8111
+    },
+    {
+      "epoch": 74.42,
+      "learning_rate": 0.00013423076923076925,
+      "loss": 0.2095,
+      "step": 8112
+    },
+    {
+      "epoch": 74.43,
+      "learning_rate": 0.00013418269230769232,
+      "loss": 0.168,
+      "step": 8113
+    },
+    {
+      "epoch": 74.44,
+      "learning_rate": 0.0001341346153846154,
+      "loss": 0.1915,
+      "step": 8114
+    },
+    {
+      "epoch": 74.45,
+      "learning_rate": 0.00013408653846153846,
+      "loss": 0.1822,
+      "step": 8115
+    },
+    {
+      "epoch": 74.46,
+      "learning_rate": 0.00013403846153846156,
+      "loss": 0.1887,
+      "step": 8116
+    },
+    {
+      "epoch": 74.47,
+      "learning_rate": 0.00013399038461538464,
+      "loss": 0.1549,
+      "step": 8117
+    },
+    {
+      "epoch": 74.48,
+      "learning_rate": 0.0001339423076923077,
+      "loss": 0.1867,
+      "step": 8118
+    },
+    {
+      "epoch": 74.49,
+      "learning_rate": 0.00013389423076923078,
+      "loss": 0.1526,
+      "step": 8119
+    },
+    {
+      "epoch": 74.5,
+      "learning_rate": 0.00013384615384615385,
+      "loss": 0.2393,
+      "step": 8120
+    },
+    {
+      "epoch": 74.5,
+      "learning_rate": 0.00013379807692307693,
+      "loss": 0.26,
+      "step": 8121
+    },
+    {
+      "epoch": 74.51,
+      "learning_rate": 0.00013375,
+      "loss": 0.2543,
+      "step": 8122
+    },
+    {
+      "epoch": 74.52,
+      "learning_rate": 0.00013370192307692307,
+      "loss": 0.21,
+      "step": 8123
+    },
+    {
+      "epoch": 74.53,
+      "learning_rate": 0.00013365384615384614,
+      "loss": 0.2815,
+      "step": 8124
+    },
+    {
+      "epoch": 74.54,
+      "learning_rate": 0.00013360576923076922,
+      "loss": 0.2151,
+      "step": 8125
+    },
+    {
+      "epoch": 74.55,
+      "learning_rate": 0.0001335576923076923,
+      "loss": 0.178,
+      "step": 8126
+    },
+    {
+      "epoch": 74.56,
+      "learning_rate": 0.0001335096153846154,
+      "loss": 0.2127,
+      "step": 8127
+    },
+    {
+      "epoch": 74.57,
+      "learning_rate": 0.00013346153846153846,
+      "loss": 0.1784,
+      "step": 8128
+    },
+    {
+      "epoch": 74.58,
+      "learning_rate": 0.00013341346153846153,
+      "loss": 0.2348,
+      "step": 8129
+    },
+    {
+      "epoch": 74.59,
+      "learning_rate": 0.0001333653846153846,
+      "loss": 0.1582,
+      "step": 8130
+    },
+    {
+      "epoch": 74.6,
+      "learning_rate": 0.00013331730769230768,
+      "loss": 0.3034,
+      "step": 8131
+    },
+    {
+      "epoch": 74.61,
+      "learning_rate": 0.00013326923076923078,
+      "loss": 0.1326,
+      "step": 8132
+    },
+    {
+      "epoch": 74.61,
+      "learning_rate": 0.00013322115384615385,
+      "loss": 0.1853,
+      "step": 8133
+    },
+    {
+      "epoch": 74.62,
+      "learning_rate": 0.00013317307692307692,
+      "loss": 0.1309,
+      "step": 8134
+    },
+    {
+      "epoch": 74.63,
+      "learning_rate": 0.000133125,
+      "loss": 0.2792,
+      "step": 8135
+    },
+    {
+      "epoch": 74.64,
+      "learning_rate": 0.00013307692307692307,
+      "loss": 0.1578,
+      "step": 8136
+    },
+    {
+      "epoch": 74.65,
+      "learning_rate": 0.00013302884615384617,
+      "loss": 0.1977,
+      "step": 8137
+    },
+    {
+      "epoch": 74.66,
+      "learning_rate": 0.00013298076923076924,
+      "loss": 0.1683,
+      "step": 8138
+    },
+    {
+      "epoch": 74.67,
+      "learning_rate": 0.00013293269230769231,
+      "loss": 0.2353,
+      "step": 8139
+    },
+    {
+      "epoch": 74.68,
+      "learning_rate": 0.0001328846153846154,
+      "loss": 0.198,
+      "step": 8140
+    },
+    {
+      "epoch": 74.69,
+      "learning_rate": 0.00013283653846153846,
+      "loss": 0.1786,
+      "step": 8141
+    },
+    {
+      "epoch": 74.7,
+      "learning_rate": 0.00013278846153846156,
+      "loss": 0.2103,
+      "step": 8142
+    },
+    {
+      "epoch": 74.71,
+      "learning_rate": 0.00013274038461538463,
+      "loss": 0.1529,
+      "step": 8143
+    },
+    {
+      "epoch": 74.72,
+      "learning_rate": 0.0001326923076923077,
+      "loss": 0.1301,
+      "step": 8144
+    },
+    {
+      "epoch": 74.72,
+      "learning_rate": 0.00013264423076923078,
+      "loss": 0.2483,
+      "step": 8145
+    },
+    {
+      "epoch": 74.73,
+      "learning_rate": 0.00013259615384615385,
+      "loss": 0.203,
+      "step": 8146
+    },
+    {
+      "epoch": 74.74,
+      "learning_rate": 0.00013254807692307695,
+      "loss": 0.1074,
+      "step": 8147
+    },
+    {
+      "epoch": 74.75,
+      "learning_rate": 0.00013250000000000002,
+      "loss": 0.2679,
+      "step": 8148
+    },
+    {
+      "epoch": 74.76,
+      "learning_rate": 0.00013245192307692307,
+      "loss": 0.2096,
+      "step": 8149
+    },
+    {
+      "epoch": 74.77,
+      "learning_rate": 0.00013240384615384614,
+      "loss": 0.2093,
+      "step": 8150
+    },
+    {
+      "epoch": 74.78,
+      "learning_rate": 0.0001323557692307692,
+      "loss": 0.2887,
+      "step": 8151
+    },
+    {
+      "epoch": 74.79,
+      "learning_rate": 0.0001323076923076923,
+      "loss": 0.2338,
+      "step": 8152
+    },
+    {
+      "epoch": 74.8,
+      "learning_rate": 0.00013225961538461539,
+      "loss": 0.2528,
+      "step": 8153
+    },
+    {
+      "epoch": 74.81,
+      "learning_rate": 0.00013221153846153846,
+      "loss": 0.2059,
+      "step": 8154
+    },
+    {
+      "epoch": 74.82,
+      "learning_rate": 0.00013216346153846153,
+      "loss": 0.2503,
+      "step": 8155
+    },
+    {
+      "epoch": 74.83,
+      "learning_rate": 0.0001321153846153846,
+      "loss": 0.1903,
+      "step": 8156
+    },
+    {
+      "epoch": 74.83,
+      "learning_rate": 0.0001320673076923077,
+      "loss": 0.1566,
+      "step": 8157
+    },
+    {
+      "epoch": 74.84,
+      "learning_rate": 0.00013201923076923078,
+      "loss": 0.1919,
+      "step": 8158
+    },
+    {
+      "epoch": 74.85,
+      "learning_rate": 0.00013197115384615385,
+      "loss": 0.1815,
+      "step": 8159
+    },
+    {
+      "epoch": 74.86,
+      "learning_rate": 0.00013192307692307692,
+      "loss": 0.1682,
+      "step": 8160
+    },
+    {
+      "epoch": 74.87,
+      "learning_rate": 0.000131875,
+      "loss": 0.1702,
+      "step": 8161
+    },
+    {
+      "epoch": 74.88,
+      "learning_rate": 0.00013182692307692307,
+      "loss": 0.2593,
+      "step": 8162
+    },
+    {
+      "epoch": 74.89,
+      "learning_rate": 0.00013177884615384617,
+      "loss": 0.2039,
+      "step": 8163
+    },
+    {
+      "epoch": 74.9,
+      "learning_rate": 0.00013173076923076924,
+      "loss": 0.1668,
+      "step": 8164
+    },
+    {
+      "epoch": 74.91,
+      "learning_rate": 0.0001316826923076923,
+      "loss": 0.2344,
+      "step": 8165
+    },
+    {
+      "epoch": 74.92,
+      "learning_rate": 0.00013163461538461538,
+      "loss": 0.1642,
+      "step": 8166
+    },
+    {
+      "epoch": 74.93,
+      "learning_rate": 0.00013158653846153846,
+      "loss": 0.1741,
+      "step": 8167
+    },
+    {
+      "epoch": 74.94,
+      "learning_rate": 0.00013153846153846156,
+      "loss": 0.1908,
+      "step": 8168
+    },
+    {
+      "epoch": 74.94,
+      "learning_rate": 0.00013149038461538463,
+      "loss": 0.2248,
+      "step": 8169
+    },
+    {
+      "epoch": 74.95,
+      "learning_rate": 0.0001314423076923077,
+      "loss": 0.1195,
+      "step": 8170
+    },
+    {
+      "epoch": 74.96,
+      "learning_rate": 0.00013139423076923077,
+      "loss": 0.1129,
+      "step": 8171
+    },
+    {
+      "epoch": 74.97,
+      "learning_rate": 0.00013134615384615385,
+      "loss": 0.2932,
+      "step": 8172
+    },
+    {
+      "epoch": 74.98,
+      "learning_rate": 0.00013129807692307695,
+      "loss": 0.0931,
+      "step": 8173
+    },
+    {
+      "epoch": 74.99,
+      "learning_rate": 0.00013125000000000002,
+      "loss": 0.1269,
+      "step": 8174
+    },
+    {
+      "epoch": 75.0,
+      "learning_rate": 0.0001312019230769231,
+      "loss": 0.2175,
+      "step": 8175
+    },
+    {
+      "epoch": 75.01,
+      "learning_rate": 0.00013115384615384616,
+      "loss": 0.2566,
+      "step": 8176
+    },
+    {
+      "epoch": 75.02,
+      "learning_rate": 0.0001311057692307692,
+      "loss": 0.2243,
+      "step": 8177
+    },
+    {
+      "epoch": 75.03,
+      "learning_rate": 0.0001310576923076923,
+      "loss": 0.2347,
+      "step": 8178
+    },
+    {
+      "epoch": 75.04,
+      "learning_rate": 0.00013100961538461538,
+      "loss": 0.2202,
+      "step": 8179
+    },
+    {
+      "epoch": 75.05,
+      "learning_rate": 0.00013096153846153845,
+      "loss": 0.2198,
+      "step": 8180
+    },
+    {
+      "epoch": 75.06,
+      "learning_rate": 0.00013091346153846153,
+      "loss": 0.2818,
+      "step": 8181
+    },
+    {
+      "epoch": 75.06,
+      "learning_rate": 0.0001308653846153846,
+      "loss": 0.2512,
+      "step": 8182
+    },
+    {
+      "epoch": 75.07,
+      "learning_rate": 0.0001308173076923077,
+      "loss": 0.2468,
+      "step": 8183
+    },
+    {
+      "epoch": 75.08,
+      "learning_rate": 0.00013076923076923077,
+      "loss": 0.2458,
+      "step": 8184
+    },
+    {
+      "epoch": 75.09,
+      "learning_rate": 0.00013072115384615384,
+      "loss": 0.2659,
+      "step": 8185
+    },
+    {
+      "epoch": 75.1,
+      "learning_rate": 0.00013067307692307692,
+      "loss": 0.1571,
+      "step": 8186
+    },
+    {
+      "epoch": 75.11,
+      "learning_rate": 0.000130625,
+      "loss": 0.2548,
+      "step": 8187
+    },
+    {
+      "epoch": 75.12,
+      "learning_rate": 0.0001305769230769231,
+      "loss": 0.181,
+      "step": 8188
+    },
+    {
+      "epoch": 75.13,
+      "learning_rate": 0.00013052884615384616,
+      "loss": 0.2327,
+      "step": 8189
+    },
+    {
+      "epoch": 75.14,
+      "learning_rate": 0.00013048076923076924,
+      "loss": 0.1928,
+      "step": 8190
+    },
+    {
+      "epoch": 75.15,
+      "learning_rate": 0.0001304326923076923,
+      "loss": 0.1869,
+      "step": 8191
+    },
+    {
+      "epoch": 75.16,
+      "learning_rate": 0.00013038461538461538,
+      "loss": 0.1742,
+      "step": 8192
+    },
+    {
+      "epoch": 75.17,
+      "learning_rate": 0.00013033653846153845,
+      "loss": 0.2203,
+      "step": 8193
+    },
+    {
+      "epoch": 75.17,
+      "learning_rate": 0.00013028846153846155,
+      "loss": 0.1676,
+      "step": 8194
+    },
+    {
+      "epoch": 75.18,
+      "learning_rate": 0.00013024038461538463,
+      "loss": 0.1755,
+      "step": 8195
+    },
+    {
+      "epoch": 75.19,
+      "learning_rate": 0.0001301923076923077,
+      "loss": 0.233,
+      "step": 8196
+    },
+    {
+      "epoch": 75.2,
+      "learning_rate": 0.00013014423076923077,
+      "loss": 0.149,
+      "step": 8197
+    },
+    {
+      "epoch": 75.21,
+      "learning_rate": 0.00013009615384615384,
+      "loss": 0.1968,
+      "step": 8198
+    },
+    {
+      "epoch": 75.22,
+      "learning_rate": 0.00013004807692307694,
+      "loss": 0.148,
+      "step": 8199
+    },
+    {
+      "epoch": 75.23,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 0.1634,
+      "step": 8200
+    },
+    {
+      "epoch": 75.24,
+      "learning_rate": 0.0001299519230769231,
+      "loss": 0.1742,
+      "step": 8201
+    },
+    {
+      "epoch": 75.25,
+      "learning_rate": 0.00012990384615384616,
+      "loss": 0.0412,
+      "step": 8202
+    },
+    {
+      "epoch": 75.26,
+      "learning_rate": 0.00012985576923076923,
+      "loss": 0.2224,
+      "step": 8203
+    },
+    {
+      "epoch": 75.27,
+      "learning_rate": 0.00012980769230769233,
+      "loss": 0.2067,
+      "step": 8204
+    },
+    {
+      "epoch": 75.28,
+      "learning_rate": 0.0001297596153846154,
+      "loss": 0.1464,
+      "step": 8205
+    },
+    {
+      "epoch": 75.28,
+      "learning_rate": 0.00012971153846153845,
+      "loss": 0.212,
+      "step": 8206
+    },
+    {
+      "epoch": 75.29,
+      "learning_rate": 0.00012966346153846152,
+      "loss": 0.2892,
+      "step": 8207
+    },
+    {
+      "epoch": 75.3,
+      "learning_rate": 0.0001296153846153846,
+      "loss": 0.2096,
+      "step": 8208
+    },
+    {
+      "epoch": 75.31,
+      "learning_rate": 0.0001295673076923077,
+      "loss": 0.1811,
+      "step": 8209
+    },
+    {
+      "epoch": 75.32,
+      "learning_rate": 0.00012951923076923077,
+      "loss": 0.2122,
+      "step": 8210
+    },
+    {
+      "epoch": 75.33,
+      "learning_rate": 0.00012947115384615384,
+      "loss": 0.2377,
+      "step": 8211
+    },
+    {
+      "epoch": 75.34,
+      "learning_rate": 0.00012942307692307691,
+      "loss": 0.231,
+      "step": 8212
+    },
+    {
+      "epoch": 75.35,
+      "learning_rate": 0.000129375,
+      "loss": 0.2563,
+      "step": 8213
+    },
+    {
+      "epoch": 75.36,
+      "learning_rate": 0.0001293269230769231,
+      "loss": 0.2025,
+      "step": 8214
+    },
+    {
+      "epoch": 75.37,
+      "learning_rate": 0.00012927884615384616,
+      "loss": 0.2091,
+      "step": 8215
+    },
+    {
+      "epoch": 75.38,
+      "learning_rate": 0.00012923076923076923,
+      "loss": 0.1568,
+      "step": 8216
+    },
+    {
+      "epoch": 75.39,
+      "learning_rate": 0.0001291826923076923,
+      "loss": 0.1513,
+      "step": 8217
+    },
+    {
+      "epoch": 75.39,
+      "learning_rate": 0.00012913461538461538,
+      "loss": 0.2116,
+      "step": 8218
+    },
+    {
+      "epoch": 75.4,
+      "learning_rate": 0.00012908653846153848,
+      "loss": 0.1469,
+      "step": 8219
+    },
+    {
+      "epoch": 75.41,
+      "learning_rate": 0.00012903846153846155,
+      "loss": 0.1468,
+      "step": 8220
+    },
+    {
+      "epoch": 75.42,
+      "learning_rate": 0.00012899038461538462,
+      "loss": 0.1227,
+      "step": 8221
+    },
+    {
+      "epoch": 75.43,
+      "learning_rate": 0.0001289423076923077,
+      "loss": 0.205,
+      "step": 8222
+    },
+    {
+      "epoch": 75.44,
+      "learning_rate": 0.00012889423076923077,
+      "loss": 0.25,
+      "step": 8223
+    },
+    {
+      "epoch": 75.45,
+      "learning_rate": 0.00012884615384615384,
+      "loss": 0.2437,
+      "step": 8224
+    },
+    {
+      "epoch": 75.46,
+      "learning_rate": 0.00012879807692307694,
+      "loss": 0.146,
+      "step": 8225
+    },
+    {
+      "epoch": 75.47,
+      "learning_rate": 0.00012875,
+      "loss": 0.141,
+      "step": 8226
+    },
+    {
+      "epoch": 75.48,
+      "learning_rate": 0.00012870192307692309,
+      "loss": 0.1809,
+      "step": 8227
+    },
+    {
+      "epoch": 75.49,
+      "learning_rate": 0.00012865384615384616,
+      "loss": 0.1577,
+      "step": 8228
+    },
+    {
+      "epoch": 75.5,
+      "learning_rate": 0.00012860576923076923,
+      "loss": 0.1809,
+      "step": 8229
+    },
+    {
+      "epoch": 75.5,
+      "learning_rate": 0.00012855769230769233,
+      "loss": 0.2775,
+      "step": 8230
+    },
+    {
+      "epoch": 75.51,
+      "learning_rate": 0.0001285096153846154,
+      "loss": 0.2546,
+      "step": 8231
+    },
+    {
+      "epoch": 75.52,
+      "learning_rate": 0.00012846153846153848,
+      "loss": 0.2399,
+      "step": 8232
+    },
+    {
+      "epoch": 75.53,
+      "learning_rate": 0.00012841346153846155,
+      "loss": 0.1847,
+      "step": 8233
+    },
+    {
+      "epoch": 75.54,
+      "learning_rate": 0.0001283653846153846,
+      "loss": 0.1828,
+      "step": 8234
+    },
+    {
+      "epoch": 75.55,
+      "learning_rate": 0.0001283173076923077,
+      "loss": 0.2443,
+      "step": 8235
+    },
+    {
+      "epoch": 75.56,
+      "learning_rate": 0.00012826923076923077,
+      "loss": 0.2302,
+      "step": 8236
+    },
+    {
+      "epoch": 75.57,
+      "learning_rate": 0.00012822115384615384,
+      "loss": 0.2013,
+      "step": 8237
+    },
+    {
+      "epoch": 75.58,
+      "learning_rate": 0.0001281730769230769,
+      "loss": 0.2238,
+      "step": 8238
+    },
+    {
+      "epoch": 75.59,
+      "learning_rate": 0.00012812499999999998,
+      "loss": 0.2198,
+      "step": 8239
+    },
+    {
+      "epoch": 75.6,
+      "learning_rate": 0.00012807692307692308,
+      "loss": 0.2159,
+      "step": 8240
+    },
+    {
+      "epoch": 75.61,
+      "learning_rate": 0.00012802884615384616,
+      "loss": 0.1588,
+      "step": 8241
+    },
+    {
+      "epoch": 75.61,
+      "learning_rate": 0.00012798076923076923,
+      "loss": 0.2292,
+      "step": 8242
+    },
+    {
+      "epoch": 75.62,
+      "learning_rate": 0.0001279326923076923,
+      "loss": 0.2314,
+      "step": 8243
+    },
+    {
+      "epoch": 75.63,
+      "learning_rate": 0.00012788461538461537,
+      "loss": 0.176,
+      "step": 8244
+    },
+    {
+      "epoch": 75.64,
+      "learning_rate": 0.00012783653846153847,
+      "loss": 0.245,
+      "step": 8245
+    },
+    {
+      "epoch": 75.65,
+      "learning_rate": 0.00012778846153846155,
+      "loss": 0.1621,
+      "step": 8246
+    },
+    {
+      "epoch": 75.66,
+      "learning_rate": 0.00012774038461538462,
+      "loss": 0.1711,
+      "step": 8247
+    },
+    {
+      "epoch": 75.67,
+      "learning_rate": 0.0001276923076923077,
+      "loss": 0.2034,
+      "step": 8248
+    },
+    {
+      "epoch": 75.68,
+      "learning_rate": 0.00012764423076923076,
+      "loss": 0.169,
+      "step": 8249
+    },
+    {
+      "epoch": 75.69,
+      "learning_rate": 0.00012759615384615386,
+      "loss": 0.1331,
+      "step": 8250
+    },
+    {
+      "epoch": 75.7,
+      "learning_rate": 0.00012754807692307694,
+      "loss": 0.1886,
+      "step": 8251
+    },
+    {
+      "epoch": 75.71,
+      "learning_rate": 0.0001275,
+      "loss": 0.2108,
+      "step": 8252
+    },
+    {
+      "epoch": 75.72,
+      "learning_rate": 0.00012745192307692308,
+      "loss": 0.1286,
+      "step": 8253
+    },
+    {
+      "epoch": 75.72,
+      "learning_rate": 0.00012740384615384615,
+      "loss": 0.1323,
+      "step": 8254
+    },
+    {
+      "epoch": 75.73,
+      "learning_rate": 0.00012735576923076925,
+      "loss": 0.1541,
+      "step": 8255
+    },
+    {
+      "epoch": 75.74,
+      "learning_rate": 0.00012730769230769233,
+      "loss": 0.1019,
+      "step": 8256
+    },
+    {
+      "epoch": 75.75,
+      "learning_rate": 0.0001272596153846154,
+      "loss": 0.3036,
+      "step": 8257
+    },
+    {
+      "epoch": 75.76,
+      "learning_rate": 0.00012721153846153847,
+      "loss": 0.2453,
+      "step": 8258
+    },
+    {
+      "epoch": 75.77,
+      "learning_rate": 0.00012716346153846154,
+      "loss": 0.2626,
+      "step": 8259
+    },
+    {
+      "epoch": 75.78,
+      "learning_rate": 0.00012711538461538462,
+      "loss": 0.2194,
+      "step": 8260
+    },
+    {
+      "epoch": 75.79,
+      "learning_rate": 0.00012706730769230772,
+      "loss": 0.2361,
+      "step": 8261
+    },
+    {
+      "epoch": 75.8,
+      "learning_rate": 0.00012701923076923076,
+      "loss": 0.3125,
+      "step": 8262
+    },
+    {
+      "epoch": 75.81,
+      "learning_rate": 0.00012697115384615384,
+      "loss": 0.1874,
+      "step": 8263
+    },
+    {
+      "epoch": 75.82,
+      "learning_rate": 0.0001269230769230769,
+      "loss": 0.1543,
+      "step": 8264
+    },
+    {
+      "epoch": 75.83,
+      "learning_rate": 0.00012687499999999998,
+      "loss": 0.1807,
+      "step": 8265
+    },
+    {
+      "epoch": 75.83,
+      "learning_rate": 0.00012682692307692308,
+      "loss": 0.2061,
+      "step": 8266
+    },
+    {
+      "epoch": 75.84,
+      "learning_rate": 0.00012677884615384615,
+      "loss": 0.1955,
+      "step": 8267
+    },
+    {
+      "epoch": 75.85,
+      "learning_rate": 0.00012673076923076923,
+      "loss": 0.1747,
+      "step": 8268
+    },
+    {
+      "epoch": 75.86,
+      "learning_rate": 0.0001266826923076923,
+      "loss": 0.1844,
+      "step": 8269
+    },
+    {
+      "epoch": 75.87,
+      "learning_rate": 0.00012663461538461537,
+      "loss": 0.2371,
+      "step": 8270
+    },
+    {
+      "epoch": 75.88,
+      "learning_rate": 0.00012658653846153847,
+      "loss": 0.2283,
+      "step": 8271
+    },
+    {
+      "epoch": 75.89,
+      "learning_rate": 0.00012653846153846154,
+      "loss": 0.2335,
+      "step": 8272
+    },
+    {
+      "epoch": 75.9,
+      "learning_rate": 0.00012649038461538462,
+      "loss": 0.1771,
+      "step": 8273
+    },
+    {
+      "epoch": 75.91,
+      "learning_rate": 0.0001264423076923077,
+      "loss": 0.1701,
+      "step": 8274
+    },
+    {
+      "epoch": 75.92,
+      "learning_rate": 0.00012639423076923076,
+      "loss": 0.1721,
+      "step": 8275
+    },
+    {
+      "epoch": 75.93,
+      "learning_rate": 0.00012634615384615386,
+      "loss": 0.1585,
+      "step": 8276
+    },
+    {
+      "epoch": 75.94,
+      "learning_rate": 0.00012629807692307693,
+      "loss": 0.1999,
+      "step": 8277
+    },
+    {
+      "epoch": 75.94,
+      "learning_rate": 0.00012625,
+      "loss": 0.2132,
+      "step": 8278
+    },
+    {
+      "epoch": 75.95,
+      "learning_rate": 0.00012620192307692308,
+      "loss": 0.1762,
+      "step": 8279
+    },
+    {
+      "epoch": 75.96,
+      "learning_rate": 0.00012615384615384615,
+      "loss": 0.2293,
+      "step": 8280
+    },
+    {
+      "epoch": 75.97,
+      "learning_rate": 0.00012610576923076925,
+      "loss": 0.3925,
+      "step": 8281
+    },
+    {
+      "epoch": 75.98,
+      "learning_rate": 0.00012605769230769232,
+      "loss": 0.2011,
+      "step": 8282
+    },
+    {
+      "epoch": 75.99,
+      "learning_rate": 0.0001260096153846154,
+      "loss": 0.076,
+      "step": 8283
+    },
+    {
+      "epoch": 76.0,
+      "learning_rate": 0.00012596153846153847,
+      "loss": 0.1613,
+      "step": 8284
+    },
+    {
+      "epoch": 76.01,
+      "learning_rate": 0.00012591346153846154,
+      "loss": 0.2534,
+      "step": 8285
+    },
+    {
+      "epoch": 76.02,
+      "learning_rate": 0.00012586538461538464,
+      "loss": 0.3102,
+      "step": 8286
+    },
+    {
+      "epoch": 76.03,
+      "learning_rate": 0.00012581730769230771,
+      "loss": 0.1746,
+      "step": 8287
+    },
+    {
+      "epoch": 76.04,
+      "learning_rate": 0.00012576923076923079,
+      "loss": 0.2267,
+      "step": 8288
+    },
+    {
+      "epoch": 76.05,
+      "learning_rate": 0.00012572115384615386,
+      "loss": 0.2343,
+      "step": 8289
+    },
+    {
+      "epoch": 76.06,
+      "learning_rate": 0.0001256730769230769,
+      "loss": 0.1133,
+      "step": 8290
+    },
+    {
+      "epoch": 76.06,
+      "learning_rate": 0.00012562499999999998,
+      "loss": 0.1259,
+      "step": 8291
+    },
+    {
+      "epoch": 76.07,
+      "learning_rate": 0.00012557692307692308,
+      "loss": 0.217,
+      "step": 8292
+    },
+    {
+      "epoch": 76.08,
+      "learning_rate": 0.00012552884615384615,
+      "loss": 0.1988,
+      "step": 8293
+    },
+    {
+      "epoch": 76.09,
+      "learning_rate": 0.00012548076923076922,
+      "loss": 0.1934,
+      "step": 8294
+    },
+    {
+      "epoch": 76.1,
+      "learning_rate": 0.0001254326923076923,
+      "loss": 0.237,
+      "step": 8295
+    },
+    {
+      "epoch": 76.11,
+      "learning_rate": 0.00012538461538461537,
+      "loss": 0.1917,
+      "step": 8296
+    },
+    {
+      "epoch": 76.12,
+      "learning_rate": 0.00012533653846153847,
+      "loss": 0.1975,
+      "step": 8297
+    },
+    {
+      "epoch": 76.13,
+      "learning_rate": 0.00012528846153846154,
+      "loss": 0.128,
+      "step": 8298
+    },
+    {
+      "epoch": 76.14,
+      "learning_rate": 0.0001252403846153846,
+      "loss": 0.1915,
+      "step": 8299
+    },
+    {
+      "epoch": 76.15,
+      "learning_rate": 0.00012519230769230768,
+      "loss": 0.2595,
+      "step": 8300
+    },
+    {
+      "epoch": 76.16,
+      "learning_rate": 0.00012514423076923076,
+      "loss": 0.1836,
+      "step": 8301
+    },
+    {
+      "epoch": 76.17,
+      "learning_rate": 0.00012509615384615386,
+      "loss": 0.1962,
+      "step": 8302
+    },
+    {
+      "epoch": 76.17,
+      "learning_rate": 0.00012504807692307693,
+      "loss": 0.2043,
+      "step": 8303
+    },
+    {
+      "epoch": 76.18,
+      "learning_rate": 0.000125,
+      "loss": 0.2213,
+      "step": 8304
+    },
+    {
+      "epoch": 76.19,
+      "learning_rate": 0.00012495192307692308,
+      "loss": 0.191,
+      "step": 8305
+    },
+    {
+      "epoch": 76.2,
+      "learning_rate": 0.00012490384615384615,
+      "loss": 0.1777,
+      "step": 8306
+    },
+    {
+      "epoch": 76.21,
+      "learning_rate": 0.00012485576923076922,
+      "loss": 0.1799,
+      "step": 8307
+    },
+    {
+      "epoch": 76.22,
+      "learning_rate": 0.00012480769230769232,
+      "loss": 0.0918,
+      "step": 8308
+    },
+    {
+      "epoch": 76.23,
+      "learning_rate": 0.0001247596153846154,
+      "loss": 0.1864,
+      "step": 8309
+    },
+    {
+      "epoch": 76.24,
+      "learning_rate": 0.00012471153846153847,
+      "loss": 0.149,
+      "step": 8310
+    },
+    {
+      "epoch": 76.25,
+      "learning_rate": 0.00012466346153846154,
+      "loss": 0.1551,
+      "step": 8311
+    },
+    {
+      "epoch": 76.26,
+      "learning_rate": 0.0001246153846153846,
+      "loss": 0.2455,
+      "step": 8312
+    },
+    {
+      "epoch": 76.27,
+      "learning_rate": 0.00012456730769230768,
+      "loss": 0.2285,
+      "step": 8313
+    },
+    {
+      "epoch": 76.28,
+      "learning_rate": 0.00012451923076923078,
+      "loss": 0.2873,
+      "step": 8314
+    },
+    {
+      "epoch": 76.28,
+      "learning_rate": 0.00012447115384615386,
+      "loss": 0.209,
+      "step": 8315
+    },
+    {
+      "epoch": 76.29,
+      "learning_rate": 0.00012442307692307693,
+      "loss": 0.1867,
+      "step": 8316
+    },
+    {
+      "epoch": 76.3,
+      "learning_rate": 0.000124375,
+      "loss": 0.265,
+      "step": 8317
+    },
+    {
+      "epoch": 76.31,
+      "learning_rate": 0.00012432692307692307,
+      "loss": 0.2301,
+      "step": 8318
+    },
+    {
+      "epoch": 76.32,
+      "learning_rate": 0.00012427884615384615,
+      "loss": 0.1953,
+      "step": 8319
+    },
+    {
+      "epoch": 76.33,
+      "learning_rate": 0.00012423076923076922,
+      "loss": 0.2253,
+      "step": 8320
+    },
+    {
+      "epoch": 76.34,
+      "learning_rate": 0.00012418269230769232,
+      "loss": 0.1057,
+      "step": 8321
+    },
+    {
+      "epoch": 76.35,
+      "learning_rate": 0.0001241346153846154,
+      "loss": 0.2056,
+      "step": 8322
+    },
+    {
+      "epoch": 76.36,
+      "learning_rate": 0.00012408653846153846,
+      "loss": 0.2001,
+      "step": 8323
+    },
+    {
+      "epoch": 76.37,
+      "learning_rate": 0.00012403846153846154,
+      "loss": 0.181,
+      "step": 8324
+    },
+    {
+      "epoch": 76.38,
+      "learning_rate": 0.0001239903846153846,
+      "loss": 0.2165,
+      "step": 8325
+    },
+    {
+      "epoch": 76.39,
+      "learning_rate": 0.0001239423076923077,
+      "loss": 0.2145,
+      "step": 8326
+    },
+    {
+      "epoch": 76.39,
+      "learning_rate": 0.00012389423076923078,
+      "loss": 0.1478,
+      "step": 8327
+    },
+    {
+      "epoch": 76.4,
+      "learning_rate": 0.00012384615384615385,
+      "loss": 0.1895,
+      "step": 8328
+    },
+    {
+      "epoch": 76.41,
+      "learning_rate": 0.00012379807692307693,
+      "loss": 0.2103,
+      "step": 8329
+    },
+    {
+      "epoch": 76.42,
+      "learning_rate": 0.00012375,
+      "loss": 0.1924,
+      "step": 8330
+    },
+    {
+      "epoch": 76.43,
+      "learning_rate": 0.00012370192307692307,
+      "loss": 0.1941,
+      "step": 8331
+    },
+    {
+      "epoch": 76.44,
+      "learning_rate": 0.00012365384615384617,
+      "loss": 0.1593,
+      "step": 8332
+    },
+    {
+      "epoch": 76.45,
+      "learning_rate": 0.00012360576923076922,
+      "loss": 0.2017,
+      "step": 8333
+    },
+    {
+      "epoch": 76.46,
+      "learning_rate": 0.00012355769230769232,
+      "loss": 0.2096,
+      "step": 8334
+    },
+    {
+      "epoch": 76.47,
+      "learning_rate": 0.0001235096153846154,
+      "loss": 0.1839,
+      "step": 8335
+    },
+    {
+      "epoch": 76.48,
+      "learning_rate": 0.00012346153846153846,
+      "loss": 0.1731,
+      "step": 8336
+    },
+    {
+      "epoch": 76.49,
+      "learning_rate": 0.00012341346153846153,
+      "loss": 0.1906,
+      "step": 8337
+    },
+    {
+      "epoch": 76.5,
+      "learning_rate": 0.0001233653846153846,
+      "loss": 0.0567,
+      "step": 8338
+    },
+    {
+      "epoch": 76.5,
+      "learning_rate": 0.0001233173076923077,
+      "loss": 0.2855,
+      "step": 8339
+    },
+    {
+      "epoch": 76.51,
+      "learning_rate": 0.00012326923076923078,
+      "loss": 0.2594,
+      "step": 8340
+    },
+    {
+      "epoch": 76.52,
+      "learning_rate": 0.00012322115384615385,
+      "loss": 0.1888,
+      "step": 8341
+    },
+    {
+      "epoch": 76.53,
+      "learning_rate": 0.00012317307692307693,
+      "loss": 0.2071,
+      "step": 8342
+    },
+    {
+      "epoch": 76.54,
+      "learning_rate": 0.000123125,
+      "loss": 0.2168,
+      "step": 8343
+    },
+    {
+      "epoch": 76.55,
+      "learning_rate": 0.0001230769230769231,
+      "loss": 0.2388,
+      "step": 8344
+    },
+    {
+      "epoch": 76.56,
+      "learning_rate": 0.00012302884615384617,
+      "loss": 0.2147,
+      "step": 8345
+    },
+    {
+      "epoch": 76.57,
+      "learning_rate": 0.00012298076923076924,
+      "loss": 0.1514,
+      "step": 8346
+    },
+    {
+      "epoch": 76.58,
+      "learning_rate": 0.00012293269230769232,
+      "loss": 0.1718,
+      "step": 8347
+    },
+    {
+      "epoch": 76.59,
+      "learning_rate": 0.0001228846153846154,
+      "loss": 0.1686,
+      "step": 8348
+    },
+    {
+      "epoch": 76.6,
+      "learning_rate": 0.00012283653846153846,
+      "loss": 0.1669,
+      "step": 8349
+    },
+    {
+      "epoch": 76.61,
+      "learning_rate": 0.00012278846153846153,
+      "loss": 0.1816,
+      "step": 8350
+    },
+    {
+      "epoch": 76.61,
+      "learning_rate": 0.0001227403846153846,
+      "loss": 0.189,
+      "step": 8351
+    },
+    {
+      "epoch": 76.62,
+      "learning_rate": 0.0001226923076923077,
+      "loss": 0.2652,
+      "step": 8352
+    },
+    {
+      "epoch": 76.63,
+      "learning_rate": 0.00012264423076923078,
+      "loss": 0.2077,
+      "step": 8353
+    },
+    {
+      "epoch": 76.64,
+      "learning_rate": 0.00012259615384615385,
+      "loss": 0.2092,
+      "step": 8354
+    },
+    {
+      "epoch": 76.65,
+      "learning_rate": 0.00012254807692307692,
+      "loss": 0.1253,
+      "step": 8355
+    },
+    {
+      "epoch": 76.66,
+      "learning_rate": 0.0001225,
+      "loss": 0.2539,
+      "step": 8356
+    },
+    {
+      "epoch": 76.67,
+      "learning_rate": 0.0001224519230769231,
+      "loss": 0.168,
+      "step": 8357
+    },
+    {
+      "epoch": 76.68,
+      "learning_rate": 0.00012240384615384617,
+      "loss": 0.1627,
+      "step": 8358
+    },
+    {
+      "epoch": 76.69,
+      "learning_rate": 0.00012235576923076924,
+      "loss": 0.1394,
+      "step": 8359
+    },
+    {
+      "epoch": 76.7,
+      "learning_rate": 0.00012230769230769231,
+      "loss": 0.1725,
+      "step": 8360
+    },
+    {
+      "epoch": 76.71,
+      "learning_rate": 0.00012225961538461539,
+      "loss": 0.1369,
+      "step": 8361
+    },
+    {
+      "epoch": 76.72,
+      "learning_rate": 0.00012221153846153846,
+      "loss": 0.1502,
+      "step": 8362
+    },
+    {
+      "epoch": 76.72,
+      "learning_rate": 0.00012216346153846153,
+      "loss": 0.2473,
+      "step": 8363
+    },
+    {
+      "epoch": 76.73,
+      "learning_rate": 0.0001221153846153846,
+      "loss": 0.1365,
+      "step": 8364
+    },
+    {
+      "epoch": 76.74,
+      "learning_rate": 0.0001220673076923077,
+      "loss": 0.1186,
+      "step": 8365
+    },
+    {
+      "epoch": 76.75,
+      "learning_rate": 0.00012201923076923078,
+      "loss": 0.24,
+      "step": 8366
+    },
+    {
+      "epoch": 76.76,
+      "learning_rate": 0.00012197115384615385,
+      "loss": 0.2429,
+      "step": 8367
+    },
+    {
+      "epoch": 76.77,
+      "learning_rate": 0.00012192307692307692,
+      "loss": 0.2243,
+      "step": 8368
+    },
+    {
+      "epoch": 76.78,
+      "learning_rate": 0.000121875,
+      "loss": 0.2543,
+      "step": 8369
+    },
+    {
+      "epoch": 76.79,
+      "learning_rate": 0.00012182692307692308,
+      "loss": 0.1952,
+      "step": 8370
+    },
+    {
+      "epoch": 76.8,
+      "learning_rate": 0.00012177884615384615,
+      "loss": 0.2605,
+      "step": 8371
+    },
+    {
+      "epoch": 76.81,
+      "learning_rate": 0.00012173076923076923,
+      "loss": 0.1261,
+      "step": 8372
+    },
+    {
+      "epoch": 76.82,
+      "learning_rate": 0.00012168269230769231,
+      "loss": 0.1597,
+      "step": 8373
+    },
+    {
+      "epoch": 76.83,
+      "learning_rate": 0.00012163461538461538,
+      "loss": 0.1349,
+      "step": 8374
+    },
+    {
+      "epoch": 76.83,
+      "learning_rate": 0.00012158653846153847,
+      "loss": 0.1951,
+      "step": 8375
+    },
+    {
+      "epoch": 76.84,
+      "learning_rate": 0.00012153846153846154,
+      "loss": 0.1973,
+      "step": 8376
+    },
+    {
+      "epoch": 76.85,
+      "learning_rate": 0.00012149038461538462,
+      "loss": 0.1663,
+      "step": 8377
+    },
+    {
+      "epoch": 76.86,
+      "learning_rate": 0.0001214423076923077,
+      "loss": 0.1484,
+      "step": 8378
+    },
+    {
+      "epoch": 76.87,
+      "learning_rate": 0.00012139423076923077,
+      "loss": 0.2209,
+      "step": 8379
+    },
+    {
+      "epoch": 76.88,
+      "learning_rate": 0.00012134615384615385,
+      "loss": 0.2084,
+      "step": 8380
+    },
+    {
+      "epoch": 76.89,
+      "learning_rate": 0.00012129807692307693,
+      "loss": 0.2299,
+      "step": 8381
+    },
+    {
+      "epoch": 76.9,
+      "learning_rate": 0.00012124999999999999,
+      "loss": 0.2067,
+      "step": 8382
+    },
+    {
+      "epoch": 76.91,
+      "learning_rate": 0.00012120192307692308,
+      "loss": 0.2329,
+      "step": 8383
+    },
+    {
+      "epoch": 76.92,
+      "learning_rate": 0.00012115384615384615,
+      "loss": 0.2249,
+      "step": 8384
+    },
+    {
+      "epoch": 76.93,
+      "learning_rate": 0.00012110576923076922,
+      "loss": 0.2241,
+      "step": 8385
+    },
+    {
+      "epoch": 76.94,
+      "learning_rate": 0.00012105769230769231,
+      "loss": 0.2157,
+      "step": 8386
+    },
+    {
+      "epoch": 76.94,
+      "learning_rate": 0.00012100961538461538,
+      "loss": 0.1893,
+      "step": 8387
+    },
+    {
+      "epoch": 76.95,
+      "learning_rate": 0.00012096153846153847,
+      "loss": 0.2111,
+      "step": 8388
+    },
+    {
+      "epoch": 76.96,
+      "learning_rate": 0.00012091346153846154,
+      "loss": 0.115,
+      "step": 8389
+    },
+    {
+      "epoch": 76.97,
+      "learning_rate": 0.00012086538461538461,
+      "loss": 0.1142,
+      "step": 8390
+    },
+    {
+      "epoch": 76.98,
+      "learning_rate": 0.0001208173076923077,
+      "loss": 0.1371,
+      "step": 8391
+    },
+    {
+      "epoch": 76.99,
+      "learning_rate": 0.00012076923076923077,
+      "loss": 0.1184,
+      "step": 8392
+    },
+    {
+      "epoch": 77.0,
+      "learning_rate": 0.00012072115384615386,
+      "loss": 0.1485,
+      "step": 8393
+    },
+    {
+      "epoch": 77.01,
+      "learning_rate": 0.00012067307692307693,
+      "loss": 0.2604,
+      "step": 8394
+    },
+    {
+      "epoch": 77.02,
+      "learning_rate": 0.000120625,
+      "loss": 0.1925,
+      "step": 8395
+    },
+    {
+      "epoch": 77.03,
+      "learning_rate": 0.00012057692307692308,
+      "loss": 0.2404,
+      "step": 8396
+    },
+    {
+      "epoch": 77.04,
+      "learning_rate": 0.00012052884615384615,
+      "loss": 0.1995,
+      "step": 8397
+    },
+    {
+      "epoch": 77.05,
+      "learning_rate": 0.00012048076923076922,
+      "loss": 0.1597,
+      "step": 8398
+    },
+    {
+      "epoch": 77.06,
+      "learning_rate": 0.00012043269230769231,
+      "loss": 0.1821,
+      "step": 8399
+    },
+    {
+      "epoch": 77.06,
+      "learning_rate": 0.00012038461538461538,
+      "loss": 0.2051,
+      "step": 8400
+    },
+    {
+      "epoch": 77.07,
+      "learning_rate": 0.00012033653846153847,
+      "loss": 0.2277,
+      "step": 8401
+    },
+    {
+      "epoch": 77.08,
+      "learning_rate": 0.00012028846153846154,
+      "loss": 0.231,
+      "step": 8402
+    },
+    {
+      "epoch": 77.09,
+      "learning_rate": 0.00012024038461538461,
+      "loss": 0.0932,
+      "step": 8403
+    },
+    {
+      "epoch": 77.1,
+      "learning_rate": 0.0001201923076923077,
+      "loss": 0.1481,
+      "step": 8404
+    },
+    {
+      "epoch": 77.11,
+      "learning_rate": 0.00012014423076923077,
+      "loss": 0.1684,
+      "step": 8405
+    },
+    {
+      "epoch": 77.12,
+      "learning_rate": 0.00012009615384615386,
+      "loss": 0.1569,
+      "step": 8406
+    },
+    {
+      "epoch": 77.13,
+      "learning_rate": 0.00012004807692307693,
+      "loss": 0.1132,
+      "step": 8407
+    },
+    {
+      "epoch": 77.14,
+      "learning_rate": 0.00012,
+      "loss": 0.1792,
+      "step": 8408
+    },
+    {
+      "epoch": 77.15,
+      "learning_rate": 0.00011995192307692309,
+      "loss": 0.2398,
+      "step": 8409
+    },
+    {
+      "epoch": 77.16,
+      "learning_rate": 0.00011990384615384615,
+      "loss": 0.1836,
+      "step": 8410
+    },
+    {
+      "epoch": 77.17,
+      "learning_rate": 0.00011985576923076922,
+      "loss": 0.209,
+      "step": 8411
+    },
+    {
+      "epoch": 77.17,
+      "learning_rate": 0.00011980769230769231,
+      "loss": 0.1739,
+      "step": 8412
+    },
+    {
+      "epoch": 77.18,
+      "learning_rate": 0.00011975961538461538,
+      "loss": 0.1732,
+      "step": 8413
+    },
+    {
+      "epoch": 77.19,
+      "learning_rate": 0.00011971153846153847,
+      "loss": 0.1481,
+      "step": 8414
+    },
+    {
+      "epoch": 77.2,
+      "learning_rate": 0.00011966346153846154,
+      "loss": 0.1971,
+      "step": 8415
+    },
+    {
+      "epoch": 77.21,
+      "learning_rate": 0.00011961538461538461,
+      "loss": 0.1454,
+      "step": 8416
+    },
+    {
+      "epoch": 77.22,
+      "learning_rate": 0.0001195673076923077,
+      "loss": 0.1412,
+      "step": 8417
+    },
+    {
+      "epoch": 77.23,
+      "learning_rate": 0.00011951923076923077,
+      "loss": 0.1985,
+      "step": 8418
+    },
+    {
+      "epoch": 77.24,
+      "learning_rate": 0.00011947115384615386,
+      "loss": 0.148,
+      "step": 8419
+    },
+    {
+      "epoch": 77.25,
+      "learning_rate": 0.00011942307692307693,
+      "loss": 0.0943,
+      "step": 8420
+    },
+    {
+      "epoch": 77.26,
+      "learning_rate": 0.000119375,
+      "loss": 0.2487,
+      "step": 8421
+    },
+    {
+      "epoch": 77.27,
+      "learning_rate": 0.00011932692307692309,
+      "loss": 0.2284,
+      "step": 8422
+    },
+    {
+      "epoch": 77.28,
+      "learning_rate": 0.00011927884615384616,
+      "loss": 0.1848,
+      "step": 8423
+    },
+    {
+      "epoch": 77.28,
+      "learning_rate": 0.00011923076923076925,
+      "loss": 0.2346,
+      "step": 8424
+    },
+    {
+      "epoch": 77.29,
+      "learning_rate": 0.0001191826923076923,
+      "loss": 0.2645,
+      "step": 8425
+    },
+    {
+      "epoch": 77.3,
+      "learning_rate": 0.00011913461538461538,
+      "loss": 0.2143,
+      "step": 8426
+    },
+    {
+      "epoch": 77.31,
+      "learning_rate": 0.00011908653846153846,
+      "loss": 0.1554,
+      "step": 8427
+    },
+    {
+      "epoch": 77.32,
+      "learning_rate": 0.00011903846153846154,
+      "loss": 0.1713,
+      "step": 8428
+    },
+    {
+      "epoch": 77.33,
+      "learning_rate": 0.00011899038461538461,
+      "loss": 0.2351,
+      "step": 8429
+    },
+    {
+      "epoch": 77.34,
+      "learning_rate": 0.0001189423076923077,
+      "loss": 0.2245,
+      "step": 8430
+    },
+    {
+      "epoch": 77.35,
+      "learning_rate": 0.00011889423076923077,
+      "loss": 0.2438,
+      "step": 8431
+    },
+    {
+      "epoch": 77.36,
+      "learning_rate": 0.00011884615384615385,
+      "loss": 0.1527,
+      "step": 8432
+    },
+    {
+      "epoch": 77.37,
+      "learning_rate": 0.00011879807692307693,
+      "loss": 0.2074,
+      "step": 8433
+    },
+    {
+      "epoch": 77.38,
+      "learning_rate": 0.00011875,
+      "loss": 0.2218,
+      "step": 8434
+    },
+    {
+      "epoch": 77.39,
+      "learning_rate": 0.00011870192307692309,
+      "loss": 0.1579,
+      "step": 8435
+    },
+    {
+      "epoch": 77.39,
+      "learning_rate": 0.00011865384615384616,
+      "loss": 0.173,
+      "step": 8436
+    },
+    {
+      "epoch": 77.4,
+      "learning_rate": 0.00011860576923076924,
+      "loss": 0.2064,
+      "step": 8437
+    },
+    {
+      "epoch": 77.41,
+      "learning_rate": 0.00011855769230769232,
+      "loss": 0.1195,
+      "step": 8438
+    },
+    {
+      "epoch": 77.42,
+      "learning_rate": 0.00011850961538461538,
+      "loss": 0.1672,
+      "step": 8439
+    },
+    {
+      "epoch": 77.43,
+      "learning_rate": 0.00011846153846153846,
+      "loss": 0.2316,
+      "step": 8440
+    },
+    {
+      "epoch": 77.44,
+      "learning_rate": 0.00011841346153846154,
+      "loss": 0.1617,
+      "step": 8441
+    },
+    {
+      "epoch": 77.45,
+      "learning_rate": 0.00011836538461538462,
+      "loss": 0.1212,
+      "step": 8442
+    },
+    {
+      "epoch": 77.46,
+      "learning_rate": 0.0001183173076923077,
+      "loss": 0.1608,
+      "step": 8443
+    },
+    {
+      "epoch": 77.47,
+      "learning_rate": 0.00011826923076923077,
+      "loss": 0.095,
+      "step": 8444
+    },
+    {
+      "epoch": 77.48,
+      "learning_rate": 0.00011822115384615385,
+      "loss": 0.0986,
+      "step": 8445
+    },
+    {
+      "epoch": 77.49,
+      "learning_rate": 0.00011817307692307693,
+      "loss": 0.1737,
+      "step": 8446
+    },
+    {
+      "epoch": 77.5,
+      "learning_rate": 0.000118125,
+      "loss": 0.0887,
+      "step": 8447
+    },
+    {
+      "epoch": 77.5,
+      "learning_rate": 0.00011807692307692308,
+      "loss": 0.2643,
+      "step": 8448
+    },
+    {
+      "epoch": 77.51,
+      "learning_rate": 0.00011802884615384616,
+      "loss": 0.18,
+      "step": 8449
+    },
+    {
+      "epoch": 77.52,
+      "learning_rate": 0.00011798076923076924,
+      "loss": 0.2555,
+      "step": 8450
+    },
+    {
+      "epoch": 77.53,
+      "learning_rate": 0.00011793269230769232,
+      "loss": 0.1465,
+      "step": 8451
+    },
+    {
+      "epoch": 77.54,
+      "learning_rate": 0.00011788461538461539,
+      "loss": 0.1387,
+      "step": 8452
+    },
+    {
+      "epoch": 77.55,
+      "learning_rate": 0.00011783653846153846,
+      "loss": 0.1836,
+      "step": 8453
+    },
+    {
+      "epoch": 77.56,
+      "learning_rate": 0.00011778846153846153,
+      "loss": 0.1897,
+      "step": 8454
+    },
+    {
+      "epoch": 77.57,
+      "learning_rate": 0.00011774038461538462,
+      "loss": 0.2476,
+      "step": 8455
+    },
+    {
+      "epoch": 77.58,
+      "learning_rate": 0.00011769230769230769,
+      "loss": 0.1481,
+      "step": 8456
+    },
+    {
+      "epoch": 77.59,
+      "learning_rate": 0.00011764423076923077,
+      "loss": 0.1311,
+      "step": 8457
+    },
+    {
+      "epoch": 77.6,
+      "learning_rate": 0.00011759615384615385,
+      "loss": 0.2023,
+      "step": 8458
+    },
+    {
+      "epoch": 77.61,
+      "learning_rate": 0.00011754807692307692,
+      "loss": 0.1688,
+      "step": 8459
+    },
+    {
+      "epoch": 77.61,
+      "learning_rate": 0.0001175,
+      "loss": 0.1742,
+      "step": 8460
+    },
+    {
+      "epoch": 77.62,
+      "learning_rate": 0.00011745192307692308,
+      "loss": 0.1763,
+      "step": 8461
+    },
+    {
+      "epoch": 77.63,
+      "learning_rate": 0.00011740384615384616,
+      "loss": 0.221,
+      "step": 8462
+    },
+    {
+      "epoch": 77.64,
+      "learning_rate": 0.00011735576923076924,
+      "loss": 0.2098,
+      "step": 8463
+    },
+    {
+      "epoch": 77.65,
+      "learning_rate": 0.00011730769230769231,
+      "loss": 0.1604,
+      "step": 8464
+    },
+    {
+      "epoch": 77.66,
+      "learning_rate": 0.00011725961538461539,
+      "loss": 0.2908,
+      "step": 8465
+    },
+    {
+      "epoch": 77.67,
+      "learning_rate": 0.00011721153846153847,
+      "loss": 0.1418,
+      "step": 8466
+    },
+    {
+      "epoch": 77.68,
+      "learning_rate": 0.00011716346153846153,
+      "loss": 0.1634,
+      "step": 8467
+    },
+    {
+      "epoch": 77.69,
+      "learning_rate": 0.00011711538461538462,
+      "loss": 0.1745,
+      "step": 8468
+    },
+    {
+      "epoch": 77.7,
+      "learning_rate": 0.00011706730769230769,
+      "loss": 0.2072,
+      "step": 8469
+    },
+    {
+      "epoch": 77.71,
+      "learning_rate": 0.00011701923076923076,
+      "loss": 0.1822,
+      "step": 8470
+    },
+    {
+      "epoch": 77.72,
+      "learning_rate": 0.00011697115384615385,
+      "loss": 0.101,
+      "step": 8471
+    },
+    {
+      "epoch": 77.72,
+      "learning_rate": 0.00011692307692307692,
+      "loss": 0.1273,
+      "step": 8472
+    },
+    {
+      "epoch": 77.73,
+      "learning_rate": 0.00011687500000000001,
+      "loss": 0.2214,
+      "step": 8473
+    },
+    {
+      "epoch": 77.74,
+      "learning_rate": 0.00011682692307692308,
+      "loss": 0.0539,
+      "step": 8474
+    },
+    {
+      "epoch": 77.75,
+      "learning_rate": 0.00011677884615384615,
+      "loss": 0.2456,
+      "step": 8475
+    },
+    {
+      "epoch": 77.76,
+      "learning_rate": 0.00011673076923076924,
+      "loss": 0.2322,
+      "step": 8476
+    },
+    {
+      "epoch": 77.77,
+      "learning_rate": 0.00011668269230769231,
+      "loss": 0.2222,
+      "step": 8477
+    },
+    {
+      "epoch": 77.78,
+      "learning_rate": 0.00011663461538461539,
+      "loss": 0.1846,
+      "step": 8478
+    },
+    {
+      "epoch": 77.79,
+      "learning_rate": 0.00011658653846153847,
+      "loss": 0.2433,
+      "step": 8479
+    },
+    {
+      "epoch": 77.8,
+      "learning_rate": 0.00011653846153846154,
+      "loss": 0.2441,
+      "step": 8480
+    },
+    {
+      "epoch": 77.81,
+      "learning_rate": 0.00011649038461538462,
+      "loss": 0.1978,
+      "step": 8481
+    },
+    {
+      "epoch": 77.82,
+      "learning_rate": 0.00011644230769230769,
+      "loss": 0.23,
+      "step": 8482
+    },
+    {
+      "epoch": 77.83,
+      "learning_rate": 0.00011639423076923076,
+      "loss": 0.2068,
+      "step": 8483
+    },
+    {
+      "epoch": 77.83,
+      "learning_rate": 0.00011634615384615385,
+      "loss": 0.1578,
+      "step": 8484
+    },
+    {
+      "epoch": 77.84,
+      "learning_rate": 0.00011629807692307692,
+      "loss": 0.222,
+      "step": 8485
+    },
+    {
+      "epoch": 77.85,
+      "learning_rate": 0.00011625,
+      "loss": 0.185,
+      "step": 8486
+    },
+    {
+      "epoch": 77.86,
+      "learning_rate": 0.00011620192307692308,
+      "loss": 0.1722,
+      "step": 8487
+    },
+    {
+      "epoch": 77.87,
+      "learning_rate": 0.00011615384615384615,
+      "loss": 0.2044,
+      "step": 8488
+    },
+    {
+      "epoch": 77.88,
+      "learning_rate": 0.00011610576923076924,
+      "loss": 0.2019,
+      "step": 8489
+    },
+    {
+      "epoch": 77.89,
+      "learning_rate": 0.00011605769230769231,
+      "loss": 0.1707,
+      "step": 8490
+    },
+    {
+      "epoch": 77.9,
+      "learning_rate": 0.00011600961538461538,
+      "loss": 0.2069,
+      "step": 8491
+    },
+    {
+      "epoch": 77.91,
+      "learning_rate": 0.00011596153846153847,
+      "loss": 0.1977,
+      "step": 8492
+    },
+    {
+      "epoch": 77.92,
+      "learning_rate": 0.00011591346153846154,
+      "loss": 0.1628,
+      "step": 8493
+    },
+    {
+      "epoch": 77.93,
+      "learning_rate": 0.00011586538461538463,
+      "loss": 0.2101,
+      "step": 8494
+    },
+    {
+      "epoch": 77.94,
+      "learning_rate": 0.00011581730769230769,
+      "loss": 0.1829,
+      "step": 8495
+    },
+    {
+      "epoch": 77.94,
+      "learning_rate": 0.00011576923076923076,
+      "loss": 0.2129,
+      "step": 8496
+    },
+    {
+      "epoch": 77.95,
+      "learning_rate": 0.00011572115384615385,
+      "loss": 0.1002,
+      "step": 8497
+    },
+    {
+      "epoch": 77.96,
+      "learning_rate": 0.00011567307692307692,
+      "loss": 0.1027,
+      "step": 8498
+    },
+    {
+      "epoch": 77.97,
+      "learning_rate": 0.000115625,
+      "loss": 0.18,
+      "step": 8499
+    },
+    {
+      "epoch": 77.98,
+      "learning_rate": 0.00011557692307692308,
+      "loss": 0.1377,
+      "step": 8500
+    },
+    {
+      "epoch": 77.98,
+      "eval_cer": 0.1470486813895261,
+      "eval_loss": 0.738295316696167,
+      "eval_runtime": 87.1913,
+      "eval_samples_per_second": 18.889,
+      "eval_steps_per_second": 2.363,
+      "eval_wer": 0.5100602594219181,
+      "step": 8500
+    },
+    {
+      "epoch": 77.99,
+      "learning_rate": 0.00011552884615384615,
+      "loss": 0.1759,
+      "step": 8501
+    },
+    {
+      "epoch": 78.0,
+      "learning_rate": 0.00011548076923076924,
+      "loss": 0.2284,
+      "step": 8502
+    },
+    {
+      "epoch": 78.01,
+      "learning_rate": 0.00011543269230769231,
+      "loss": 0.2303,
+      "step": 8503
+    },
+    {
+      "epoch": 78.02,
+      "learning_rate": 0.0001153846153846154,
+      "loss": 0.2254,
+      "step": 8504
+    },
+    {
+      "epoch": 78.03,
+      "learning_rate": 0.00011533653846153847,
+      "loss": 0.2629,
+      "step": 8505
+    },
+    {
+      "epoch": 78.04,
+      "learning_rate": 0.00011528846153846154,
+      "loss": 0.2089,
+      "step": 8506
+    },
+    {
+      "epoch": 78.05,
+      "learning_rate": 0.00011524038461538463,
+      "loss": 0.1578,
+      "step": 8507
+    },
+    {
+      "epoch": 78.06,
+      "learning_rate": 0.0001151923076923077,
+      "loss": 0.1978,
+      "step": 8508
+    },
+    {
+      "epoch": 78.06,
+      "learning_rate": 0.00011514423076923076,
+      "loss": 0.2352,
+      "step": 8509
+    },
+    {
+      "epoch": 78.07,
+      "learning_rate": 0.00011509615384615384,
+      "loss": 0.203,
+      "step": 8510
+    },
+    {
+      "epoch": 78.08,
+      "learning_rate": 0.00011504807692307692,
+      "loss": 0.1252,
+      "step": 8511
+    },
+    {
+      "epoch": 78.09,
+      "learning_rate": 0.000115,
+      "loss": 0.1092,
+      "step": 8512
+    },
+    {
+      "epoch": 78.1,
+      "learning_rate": 0.00011495192307692308,
+      "loss": 0.1726,
+      "step": 8513
+    },
+    {
+      "epoch": 78.11,
+      "learning_rate": 0.00011490384615384615,
+      "loss": 0.1344,
+      "step": 8514
+    },
+    {
+      "epoch": 78.12,
+      "learning_rate": 0.00011485576923076923,
+      "loss": 0.2389,
+      "step": 8515
+    },
+    {
+      "epoch": 78.13,
+      "learning_rate": 0.00011480769230769231,
+      "loss": 0.1881,
+      "step": 8516
+    },
+    {
+      "epoch": 78.14,
+      "learning_rate": 0.0001147596153846154,
+      "loss": 0.2241,
+      "step": 8517
+    },
+    {
+      "epoch": 78.15,
+      "learning_rate": 0.00011471153846153847,
+      "loss": 0.1405,
+      "step": 8518
+    },
+    {
+      "epoch": 78.16,
+      "learning_rate": 0.00011466346153846154,
+      "loss": 0.1137,
+      "step": 8519
+    },
+    {
+      "epoch": 78.17,
+      "learning_rate": 0.00011461538461538463,
+      "loss": 0.1846,
+      "step": 8520
+    },
+    {
+      "epoch": 78.17,
+      "learning_rate": 0.0001145673076923077,
+      "loss": 0.1269,
+      "step": 8521
+    },
+    {
+      "epoch": 78.18,
+      "learning_rate": 0.00011451923076923078,
+      "loss": 0.1817,
+      "step": 8522
+    },
+    {
+      "epoch": 78.19,
+      "learning_rate": 0.00011447115384615386,
+      "loss": 0.1441,
+      "step": 8523
+    },
+    {
+      "epoch": 78.2,
+      "learning_rate": 0.00011442307692307692,
+      "loss": 0.2316,
+      "step": 8524
+    },
+    {
+      "epoch": 78.21,
+      "learning_rate": 0.000114375,
+      "loss": 0.2048,
+      "step": 8525
+    },
+    {
+      "epoch": 78.22,
+      "learning_rate": 0.00011432692307692307,
+      "loss": 0.1498,
+      "step": 8526
+    },
+    {
+      "epoch": 78.23,
+      "learning_rate": 0.00011427884615384615,
+      "loss": 0.1416,
+      "step": 8527
+    },
+    {
+      "epoch": 78.24,
+      "learning_rate": 0.00011423076923076923,
+      "loss": 0.1308,
+      "step": 8528
+    },
+    {
+      "epoch": 78.25,
+      "learning_rate": 0.0001141826923076923,
+      "loss": 0.1276,
+      "step": 8529
+    },
+    {
+      "epoch": 78.26,
+      "learning_rate": 0.00011413461538461539,
+      "loss": 0.2682,
+      "step": 8530
+    },
+    {
+      "epoch": 78.27,
+      "learning_rate": 0.00011408653846153846,
+      "loss": 0.1786,
+      "step": 8531
+    },
+    {
+      "epoch": 78.28,
+      "learning_rate": 0.00011403846153846154,
+      "loss": 0.1943,
+      "step": 8532
+    },
+    {
+      "epoch": 78.28,
+      "learning_rate": 0.00011399038461538462,
+      "loss": 0.216,
+      "step": 8533
+    },
+    {
+      "epoch": 78.29,
+      "learning_rate": 0.0001139423076923077,
+      "loss": 0.225,
+      "step": 8534
+    },
+    {
+      "epoch": 78.3,
+      "learning_rate": 0.00011389423076923078,
+      "loss": 0.2646,
+      "step": 8535
+    },
+    {
+      "epoch": 78.31,
+      "learning_rate": 0.00011384615384615385,
+      "loss": 0.2123,
+      "step": 8536
+    },
+    {
+      "epoch": 78.32,
+      "learning_rate": 0.00011379807692307693,
+      "loss": 0.1455,
+      "step": 8537
+    },
+    {
+      "epoch": 78.33,
+      "learning_rate": 0.00011375,
+      "loss": 0.1423,
+      "step": 8538
+    },
+    {
+      "epoch": 78.34,
+      "learning_rate": 0.00011370192307692307,
+      "loss": 0.1888,
+      "step": 8539
+    },
+    {
+      "epoch": 78.35,
+      "learning_rate": 0.00011365384615384615,
+      "loss": 0.1759,
+      "step": 8540
+    },
+    {
+      "epoch": 78.36,
+      "learning_rate": 0.00011360576923076923,
+      "loss": 0.2101,
+      "step": 8541
+    },
+    {
+      "epoch": 78.37,
+      "learning_rate": 0.0001135576923076923,
+      "loss": 0.1487,
+      "step": 8542
+    },
+    {
+      "epoch": 78.38,
+      "learning_rate": 0.00011350961538461539,
+      "loss": 0.2278,
+      "step": 8543
+    },
+    {
+      "epoch": 78.39,
+      "learning_rate": 0.00011346153846153846,
+      "loss": 0.1832,
+      "step": 8544
+    },
+    {
+      "epoch": 78.39,
+      "learning_rate": 0.00011341346153846154,
+      "loss": 0.1283,
+      "step": 8545
+    },
+    {
+      "epoch": 78.4,
+      "learning_rate": 0.00011336538461538462,
+      "loss": 0.1634,
+      "step": 8546
+    },
+    {
+      "epoch": 78.41,
+      "learning_rate": 0.0001133173076923077,
+      "loss": 0.1656,
+      "step": 8547
+    },
+    {
+      "epoch": 78.42,
+      "learning_rate": 0.00011326923076923078,
+      "loss": 0.2882,
+      "step": 8548
+    },
+    {
+      "epoch": 78.43,
+      "learning_rate": 0.00011322115384615385,
+      "loss": 0.1612,
+      "step": 8549
+    },
+    {
+      "epoch": 78.44,
+      "learning_rate": 0.00011317307692307693,
+      "loss": 0.1548,
+      "step": 8550
+    },
+    {
+      "epoch": 78.45,
+      "learning_rate": 0.00011312500000000001,
+      "loss": 0.1381,
+      "step": 8551
+    },
+    {
+      "epoch": 78.46,
+      "learning_rate": 0.00011307692307692307,
+      "loss": 0.2184,
+      "step": 8552
+    },
+    {
+      "epoch": 78.47,
+      "learning_rate": 0.00011302884615384616,
+      "loss": 0.1954,
+      "step": 8553
+    },
+    {
+      "epoch": 78.48,
+      "learning_rate": 0.00011298076923076923,
+      "loss": 0.1794,
+      "step": 8554
+    },
+    {
+      "epoch": 78.49,
+      "learning_rate": 0.0001129326923076923,
+      "loss": 0.1348,
+      "step": 8555
+    },
+    {
+      "epoch": 78.5,
+      "learning_rate": 0.00011288461538461539,
+      "loss": 0.0609,
+      "step": 8556
+    },
+    {
+      "epoch": 78.5,
+      "learning_rate": 0.00011283653846153846,
+      "loss": 0.2162,
+      "step": 8557
+    },
+    {
+      "epoch": 78.51,
+      "learning_rate": 0.00011278846153846153,
+      "loss": 0.2441,
+      "step": 8558
+    },
+    {
+      "epoch": 78.52,
+      "learning_rate": 0.00011274038461538462,
+      "loss": 0.1777,
+      "step": 8559
+    },
+    {
+      "epoch": 78.53,
+      "learning_rate": 0.00011269230769230769,
+      "loss": 0.2357,
+      "step": 8560
+    },
+    {
+      "epoch": 78.54,
+      "learning_rate": 0.00011264423076923078,
+      "loss": 0.247,
+      "step": 8561
+    },
+    {
+      "epoch": 78.55,
+      "learning_rate": 0.00011259615384615385,
+      "loss": 0.2068,
+      "step": 8562
+    },
+    {
+      "epoch": 78.56,
+      "learning_rate": 0.00011254807692307692,
+      "loss": 0.1972,
+      "step": 8563
+    },
+    {
+      "epoch": 78.57,
+      "learning_rate": 0.00011250000000000001,
+      "loss": 0.1255,
+      "step": 8564
+    },
+    {
+      "epoch": 78.58,
+      "learning_rate": 0.00011245192307692308,
+      "loss": 0.1546,
+      "step": 8565
+    },
+    {
+      "epoch": 78.59,
+      "learning_rate": 0.00011240384615384616,
+      "loss": 0.2409,
+      "step": 8566
+    },
+    {
+      "epoch": 78.6,
+      "learning_rate": 0.00011235576923076923,
+      "loss": 0.2174,
+      "step": 8567
+    },
+    {
+      "epoch": 78.61,
+      "learning_rate": 0.0001123076923076923,
+      "loss": 0.1689,
+      "step": 8568
+    },
+    {
+      "epoch": 78.61,
+      "learning_rate": 0.00011225961538461539,
+      "loss": 0.1313,
+      "step": 8569
+    },
+    {
+      "epoch": 78.62,
+      "learning_rate": 0.00011221153846153846,
+      "loss": 0.2637,
+      "step": 8570
+    },
+    {
+      "epoch": 78.63,
+      "learning_rate": 0.00011216346153846155,
+      "loss": 0.175,
+      "step": 8571
+    },
+    {
+      "epoch": 78.64,
+      "learning_rate": 0.00011211538461538462,
+      "loss": 0.1119,
+      "step": 8572
+    },
+    {
+      "epoch": 78.65,
+      "learning_rate": 0.00011206730769230769,
+      "loss": 0.1567,
+      "step": 8573
+    },
+    {
+      "epoch": 78.66,
+      "learning_rate": 0.00011201923076923078,
+      "loss": 0.1629,
+      "step": 8574
+    },
+    {
+      "epoch": 78.67,
+      "learning_rate": 0.00011197115384615385,
+      "loss": 0.17,
+      "step": 8575
+    },
+    {
+      "epoch": 78.68,
+      "learning_rate": 0.00011192307692307692,
+      "loss": 0.1669,
+      "step": 8576
+    },
+    {
+      "epoch": 78.69,
+      "learning_rate": 0.00011187500000000001,
+      "loss": 0.2142,
+      "step": 8577
+    },
+    {
+      "epoch": 78.7,
+      "learning_rate": 0.00011182692307692308,
+      "loss": 0.2276,
+      "step": 8578
+    },
+    {
+      "epoch": 78.71,
+      "learning_rate": 0.00011177884615384617,
+      "loss": 0.1797,
+      "step": 8579
+    },
+    {
+      "epoch": 78.72,
+      "learning_rate": 0.00011173076923076923,
+      "loss": 0.2157,
+      "step": 8580
+    },
+    {
+      "epoch": 78.72,
+      "learning_rate": 0.0001116826923076923,
+      "loss": 0.2644,
+      "step": 8581
+    },
+    {
+      "epoch": 78.73,
+      "learning_rate": 0.00011163461538461539,
+      "loss": 0.1398,
+      "step": 8582
+    },
+    {
+      "epoch": 78.74,
+      "learning_rate": 0.00011158653846153846,
+      "loss": 0.09,
+      "step": 8583
+    },
+    {
+      "epoch": 78.75,
+      "learning_rate": 0.00011153846153846154,
+      "loss": 0.2624,
+      "step": 8584
+    },
+    {
+      "epoch": 78.76,
+      "learning_rate": 0.00011149038461538462,
+      "loss": 0.1605,
+      "step": 8585
+    },
+    {
+      "epoch": 78.77,
+      "learning_rate": 0.00011144230769230769,
+      "loss": 0.2149,
+      "step": 8586
+    },
+    {
+      "epoch": 78.78,
+      "learning_rate": 0.00011139423076923078,
+      "loss": 0.2386,
+      "step": 8587
+    },
+    {
+      "epoch": 78.79,
+      "learning_rate": 0.00011134615384615385,
+      "loss": 0.1911,
+      "step": 8588
+    },
+    {
+      "epoch": 78.8,
+      "learning_rate": 0.00011129807692307692,
+      "loss": 0.169,
+      "step": 8589
+    },
+    {
+      "epoch": 78.81,
+      "learning_rate": 0.00011125000000000001,
+      "loss": 0.1673,
+      "step": 8590
+    },
+    {
+      "epoch": 78.82,
+      "learning_rate": 0.00011120192307692308,
+      "loss": 0.1816,
+      "step": 8591
+    },
+    {
+      "epoch": 78.83,
+      "learning_rate": 0.00011115384615384617,
+      "loss": 0.1802,
+      "step": 8592
+    },
+    {
+      "epoch": 78.83,
+      "learning_rate": 0.00011110576923076924,
+      "loss": 0.1394,
+      "step": 8593
+    },
+    {
+      "epoch": 78.84,
+      "learning_rate": 0.0001110576923076923,
+      "loss": 0.2276,
+      "step": 8594
+    },
+    {
+      "epoch": 78.85,
+      "learning_rate": 0.00011100961538461538,
+      "loss": 0.2249,
+      "step": 8595
+    },
+    {
+      "epoch": 78.86,
+      "learning_rate": 0.00011096153846153846,
+      "loss": 0.1566,
+      "step": 8596
+    },
+    {
+      "epoch": 78.87,
+      "learning_rate": 0.00011091346153846154,
+      "loss": 0.2147,
+      "step": 8597
+    },
+    {
+      "epoch": 78.88,
+      "learning_rate": 0.00011086538461538462,
+      "loss": 0.1439,
+      "step": 8598
+    },
+    {
+      "epoch": 78.89,
+      "learning_rate": 0.00011081730769230769,
+      "loss": 0.1812,
+      "step": 8599
+    },
+    {
+      "epoch": 78.9,
+      "learning_rate": 0.00011076923076923077,
+      "loss": 0.1164,
+      "step": 8600
+    },
+    {
+      "epoch": 78.91,
+      "learning_rate": 0.00011072115384615385,
+      "loss": 0.1567,
+      "step": 8601
+    },
+    {
+      "epoch": 78.92,
+      "learning_rate": 0.00011067307692307693,
+      "loss": 0.2012,
+      "step": 8602
+    },
+    {
+      "epoch": 78.93,
+      "learning_rate": 0.000110625,
+      "loss": 0.141,
+      "step": 8603
+    },
+    {
+      "epoch": 78.94,
+      "learning_rate": 0.00011057692307692308,
+      "loss": 0.18,
+      "step": 8604
+    },
+    {
+      "epoch": 78.94,
+      "learning_rate": 0.00011052884615384616,
+      "loss": 0.1952,
+      "step": 8605
+    },
+    {
+      "epoch": 78.95,
+      "learning_rate": 0.00011048076923076924,
+      "loss": 0.1216,
+      "step": 8606
+    },
+    {
+      "epoch": 78.96,
+      "learning_rate": 0.00011043269230769231,
+      "loss": 0.1998,
+      "step": 8607
+    },
+    {
+      "epoch": 78.97,
+      "learning_rate": 0.00011038461538461538,
+      "loss": 0.1621,
+      "step": 8608
+    },
+    {
+      "epoch": 78.98,
+      "learning_rate": 0.00011033653846153845,
+      "loss": 0.1645,
+      "step": 8609
+    },
+    {
+      "epoch": 78.99,
+      "learning_rate": 0.00011028846153846154,
+      "loss": 0.1145,
+      "step": 8610
+    },
+    {
+      "epoch": 79.0,
+      "learning_rate": 0.00011024038461538461,
+      "loss": 0.1786,
+      "step": 8611
+    },
+    {
+      "epoch": 79.01,
+      "learning_rate": 0.00011019230769230769,
+      "loss": 0.2406,
+      "step": 8612
+    },
+    {
+      "epoch": 79.02,
+      "learning_rate": 0.00011014423076923077,
+      "loss": 0.1842,
+      "step": 8613
+    },
+    {
+      "epoch": 79.03,
+      "learning_rate": 0.00011009615384615385,
+      "loss": 0.1507,
+      "step": 8614
+    },
+    {
+      "epoch": 79.04,
+      "learning_rate": 0.00011004807692307693,
+      "loss": 0.1573,
+      "step": 8615
+    },
+    {
+      "epoch": 79.05,
+      "learning_rate": 0.00011,
+      "loss": 0.1389,
+      "step": 8616
+    },
+    {
+      "epoch": 79.06,
+      "learning_rate": 0.00010995192307692308,
+      "loss": 0.2142,
+      "step": 8617
+    },
+    {
+      "epoch": 79.06,
+      "learning_rate": 0.00010990384615384616,
+      "loss": 0.1989,
+      "step": 8618
+    },
+    {
+      "epoch": 79.07,
+      "learning_rate": 0.00010985576923076924,
+      "loss": 0.1579,
+      "step": 8619
+    },
+    {
+      "epoch": 79.08,
+      "learning_rate": 0.00010980769230769232,
+      "loss": 0.157,
+      "step": 8620
+    },
+    {
+      "epoch": 79.09,
+      "learning_rate": 0.0001097596153846154,
+      "loss": 0.1885,
+      "step": 8621
+    },
+    {
+      "epoch": 79.1,
+      "learning_rate": 0.00010971153846153845,
+      "loss": 0.1879,
+      "step": 8622
+    },
+    {
+      "epoch": 79.11,
+      "learning_rate": 0.00010966346153846154,
+      "loss": 0.1343,
+      "step": 8623
+    },
+    {
+      "epoch": 79.12,
+      "learning_rate": 0.00010961538461538461,
+      "loss": 0.1479,
+      "step": 8624
+    },
+    {
+      "epoch": 79.13,
+      "learning_rate": 0.00010956730769230768,
+      "loss": 0.2005,
+      "step": 8625
+    },
+    {
+      "epoch": 79.14,
+      "learning_rate": 0.00010951923076923077,
+      "loss": 0.2049,
+      "step": 8626
+    },
+    {
+      "epoch": 79.15,
+      "learning_rate": 0.00010947115384615384,
+      "loss": 0.1456,
+      "step": 8627
+    },
+    {
+      "epoch": 79.16,
+      "learning_rate": 0.00010942307692307693,
+      "loss": 0.2013,
+      "step": 8628
+    },
+    {
+      "epoch": 79.17,
+      "learning_rate": 0.000109375,
+      "loss": 0.1572,
+      "step": 8629
+    },
+    {
+      "epoch": 79.17,
+      "learning_rate": 0.00010932692307692307,
+      "loss": 0.288,
+      "step": 8630
+    },
+    {
+      "epoch": 79.18,
+      "learning_rate": 0.00010927884615384616,
+      "loss": 0.1268,
+      "step": 8631
+    },
+    {
+      "epoch": 79.19,
+      "learning_rate": 0.00010923076923076923,
+      "loss": 0.1663,
+      "step": 8632
+    },
+    {
+      "epoch": 79.2,
+      "learning_rate": 0.00010918269230769232,
+      "loss": 0.2062,
+      "step": 8633
+    },
+    {
+      "epoch": 79.21,
+      "learning_rate": 0.00010913461538461539,
+      "loss": 0.1473,
+      "step": 8634
+    },
+    {
+      "epoch": 79.22,
+      "learning_rate": 0.00010908653846153847,
+      "loss": 0.1661,
+      "step": 8635
+    },
+    {
+      "epoch": 79.23,
+      "learning_rate": 0.00010903846153846155,
+      "loss": 0.1603,
+      "step": 8636
+    },
+    {
+      "epoch": 79.24,
+      "learning_rate": 0.00010899038461538461,
+      "loss": 0.0939,
+      "step": 8637
+    },
+    {
+      "epoch": 79.25,
+      "learning_rate": 0.00010894230769230768,
+      "loss": 0.0888,
+      "step": 8638
+    },
+    {
+      "epoch": 79.26,
+      "learning_rate": 0.00010889423076923077,
+      "loss": 0.2599,
+      "step": 8639
+    },
+    {
+      "epoch": 79.27,
+      "learning_rate": 0.00010884615384615384,
+      "loss": 0.1964,
+      "step": 8640
+    },
+    {
+      "epoch": 79.28,
+      "learning_rate": 0.00010879807692307693,
+      "loss": 0.1759,
+      "step": 8641
+    },
+    {
+      "epoch": 79.28,
+      "learning_rate": 0.00010875,
+      "loss": 0.2203,
+      "step": 8642
+    },
+    {
+      "epoch": 79.29,
+      "learning_rate": 0.00010870192307692307,
+      "loss": 0.1873,
+      "step": 8643
+    },
+    {
+      "epoch": 79.3,
+      "learning_rate": 0.00010865384615384616,
+      "loss": 0.2174,
+      "step": 8644
+    },
+    {
+      "epoch": 79.31,
+      "learning_rate": 0.00010860576923076923,
+      "loss": 0.1933,
+      "step": 8645
+    },
+    {
+      "epoch": 79.32,
+      "learning_rate": 0.00010855769230769232,
+      "loss": 0.1558,
+      "step": 8646
+    },
+    {
+      "epoch": 79.33,
+      "learning_rate": 0.00010850961538461539,
+      "loss": 0.1304,
+      "step": 8647
+    },
+    {
+      "epoch": 79.34,
+      "learning_rate": 0.00010846153846153846,
+      "loss": 0.1134,
+      "step": 8648
+    },
+    {
+      "epoch": 79.35,
+      "learning_rate": 0.00010841346153846155,
+      "loss": 0.1128,
+      "step": 8649
+    },
+    {
+      "epoch": 79.36,
+      "learning_rate": 0.00010836538461538462,
+      "loss": 0.1303,
+      "step": 8650
+    },
+    {
+      "epoch": 79.37,
+      "learning_rate": 0.0001083173076923077,
+      "loss": 0.1709,
+      "step": 8651
+    },
+    {
+      "epoch": 79.38,
+      "learning_rate": 0.00010826923076923077,
+      "loss": 0.1286,
+      "step": 8652
+    },
+    {
+      "epoch": 79.39,
+      "learning_rate": 0.00010822115384615384,
+      "loss": 0.2037,
+      "step": 8653
+    },
+    {
+      "epoch": 79.39,
+      "learning_rate": 0.00010817307692307693,
+      "loss": 0.2152,
+      "step": 8654
+    },
+    {
+      "epoch": 79.4,
+      "learning_rate": 0.000108125,
+      "loss": 0.2038,
+      "step": 8655
+    },
+    {
+      "epoch": 79.41,
+      "learning_rate": 0.00010807692307692307,
+      "loss": 0.1822,
+      "step": 8656
+    },
+    {
+      "epoch": 79.42,
+      "learning_rate": 0.00010802884615384616,
+      "loss": 0.2154,
+      "step": 8657
+    },
+    {
+      "epoch": 79.43,
+      "learning_rate": 0.00010798076923076923,
+      "loss": 0.1422,
+      "step": 8658
+    },
+    {
+      "epoch": 79.44,
+      "learning_rate": 0.00010793269230769232,
+      "loss": 0.1954,
+      "step": 8659
+    },
+    {
+      "epoch": 79.45,
+      "learning_rate": 0.00010788461538461539,
+      "loss": 0.1716,
+      "step": 8660
+    },
+    {
+      "epoch": 79.46,
+      "learning_rate": 0.00010783653846153846,
+      "loss": 0.2132,
+      "step": 8661
+    },
+    {
+      "epoch": 79.47,
+      "learning_rate": 0.00010778846153846155,
+      "loss": 0.2117,
+      "step": 8662
+    },
+    {
+      "epoch": 79.48,
+      "learning_rate": 0.00010774038461538462,
+      "loss": 0.1417,
+      "step": 8663
+    },
+    {
+      "epoch": 79.49,
+      "learning_rate": 0.00010769230769230771,
+      "loss": 0.1074,
+      "step": 8664
+    },
+    {
+      "epoch": 79.5,
+      "learning_rate": 0.00010764423076923077,
+      "loss": 0.0978,
+      "step": 8665
+    },
+    {
+      "epoch": 79.5,
+      "learning_rate": 0.00010759615384615384,
+      "loss": 0.2645,
+      "step": 8666
+    },
+    {
+      "epoch": 79.51,
+      "learning_rate": 0.00010754807692307692,
+      "loss": 0.1312,
+      "step": 8667
+    },
+    {
+      "epoch": 79.52,
+      "learning_rate": 0.0001075,
+      "loss": 0.2133,
+      "step": 8668
+    },
+    {
+      "epoch": 79.53,
+      "learning_rate": 0.00010745192307692308,
+      "loss": 0.1986,
+      "step": 8669
+    },
+    {
+      "epoch": 79.54,
+      "learning_rate": 0.00010740384615384616,
+      "loss": 0.2138,
+      "step": 8670
+    },
+    {
+      "epoch": 79.55,
+      "learning_rate": 0.00010735576923076923,
+      "loss": 0.1713,
+      "step": 8671
+    },
+    {
+      "epoch": 79.56,
+      "learning_rate": 0.00010730769230769231,
+      "loss": 0.2172,
+      "step": 8672
+    },
+    {
+      "epoch": 79.57,
+      "learning_rate": 0.00010725961538461539,
+      "loss": 0.1822,
+      "step": 8673
+    },
+    {
+      "epoch": 79.58,
+      "learning_rate": 0.00010721153846153846,
+      "loss": 0.1317,
+      "step": 8674
+    },
+    {
+      "epoch": 79.59,
+      "learning_rate": 0.00010716346153846155,
+      "loss": 0.1349,
+      "step": 8675
+    },
+    {
+      "epoch": 79.6,
+      "learning_rate": 0.00010711538461538462,
+      "loss": 0.2122,
+      "step": 8676
+    },
+    {
+      "epoch": 79.61,
+      "learning_rate": 0.0001070673076923077,
+      "loss": 0.124,
+      "step": 8677
+    },
+    {
+      "epoch": 79.61,
+      "learning_rate": 0.00010701923076923078,
+      "loss": 0.2137,
+      "step": 8678
+    },
+    {
+      "epoch": 79.62,
+      "learning_rate": 0.00010697115384615384,
+      "loss": 0.2209,
+      "step": 8679
+    },
+    {
+      "epoch": 79.63,
+      "learning_rate": 0.00010692307692307692,
+      "loss": 0.1189,
+      "step": 8680
+    },
+    {
+      "epoch": 79.64,
+      "learning_rate": 0.000106875,
+      "loss": 0.1452,
+      "step": 8681
+    },
+    {
+      "epoch": 79.65,
+      "learning_rate": 0.00010682692307692308,
+      "loss": 0.1561,
+      "step": 8682
+    },
+    {
+      "epoch": 79.66,
+      "learning_rate": 0.00010677884615384615,
+      "loss": 0.2,
+      "step": 8683
+    },
+    {
+      "epoch": 79.67,
+      "learning_rate": 0.00010673076923076923,
+      "loss": 0.1513,
+      "step": 8684
+    },
+    {
+      "epoch": 79.68,
+      "learning_rate": 0.00010668269230769231,
+      "loss": 0.1302,
+      "step": 8685
+    },
+    {
+      "epoch": 79.69,
+      "learning_rate": 0.00010663461538461539,
+      "loss": 0.127,
+      "step": 8686
+    },
+    {
+      "epoch": 79.7,
+      "learning_rate": 0.00010658653846153846,
+      "loss": 0.1963,
+      "step": 8687
+    },
+    {
+      "epoch": 79.71,
+      "learning_rate": 0.00010653846153846154,
+      "loss": 0.087,
+      "step": 8688
+    },
+    {
+      "epoch": 79.72,
+      "learning_rate": 0.00010649038461538462,
+      "loss": 0.129,
+      "step": 8689
+    },
+    {
+      "epoch": 79.72,
+      "learning_rate": 0.0001064423076923077,
+      "loss": 0.1074,
+      "step": 8690
+    },
+    {
+      "epoch": 79.73,
+      "learning_rate": 0.00010639423076923078,
+      "loss": 0.1191,
+      "step": 8691
+    },
+    {
+      "epoch": 79.74,
+      "learning_rate": 0.00010634615384615385,
+      "loss": 0.145,
+      "step": 8692
+    },
+    {
+      "epoch": 79.75,
+      "learning_rate": 0.00010629807692307692,
+      "loss": 0.1997,
+      "step": 8693
+    },
+    {
+      "epoch": 79.76,
+      "learning_rate": 0.00010625,
+      "loss": 0.2774,
+      "step": 8694
+    },
+    {
+      "epoch": 79.77,
+      "learning_rate": 0.00010620192307692308,
+      "loss": 0.1957,
+      "step": 8695
+    },
+    {
+      "epoch": 79.78,
+      "learning_rate": 0.00010615384615384615,
+      "loss": 0.2317,
+      "step": 8696
+    },
+    {
+      "epoch": 79.79,
+      "learning_rate": 0.00010610576923076923,
+      "loss": 0.1945,
+      "step": 8697
+    },
+    {
+      "epoch": 79.8,
+      "learning_rate": 0.00010605769230769231,
+      "loss": 0.1657,
+      "step": 8698
+    },
+    {
+      "epoch": 79.81,
+      "learning_rate": 0.00010600961538461538,
+      "loss": 0.1921,
+      "step": 8699
+    },
+    {
+      "epoch": 79.82,
+      "learning_rate": 0.00010596153846153847,
+      "loss": 0.1476,
+      "step": 8700
+    },
+    {
+      "epoch": 79.83,
+      "learning_rate": 0.00010591346153846154,
+      "loss": 0.1412,
+      "step": 8701
+    },
+    {
+      "epoch": 79.83,
+      "learning_rate": 0.00010586538461538462,
+      "loss": 0.1831,
+      "step": 8702
+    },
+    {
+      "epoch": 79.84,
+      "learning_rate": 0.0001058173076923077,
+      "loss": 0.1863,
+      "step": 8703
+    },
+    {
+      "epoch": 79.85,
+      "learning_rate": 0.00010576923076923077,
+      "loss": 0.1423,
+      "step": 8704
+    },
+    {
+      "epoch": 79.86,
+      "learning_rate": 0.00010572115384615385,
+      "loss": 0.2027,
+      "step": 8705
+    },
+    {
+      "epoch": 79.87,
+      "learning_rate": 0.00010567307692307693,
+      "loss": 0.1865,
+      "step": 8706
+    },
+    {
+      "epoch": 79.88,
+      "learning_rate": 0.00010562499999999999,
+      "loss": 0.1564,
+      "step": 8707
+    },
+    {
+      "epoch": 79.89,
+      "learning_rate": 0.00010557692307692308,
+      "loss": 0.1514,
+      "step": 8708
+    },
+    {
+      "epoch": 79.9,
+      "learning_rate": 0.00010552884615384615,
+      "loss": 0.1524,
+      "step": 8709
+    },
+    {
+      "epoch": 79.91,
+      "learning_rate": 0.00010548076923076922,
+      "loss": 0.1366,
+      "step": 8710
+    },
+    {
+      "epoch": 79.92,
+      "learning_rate": 0.00010543269230769231,
+      "loss": 0.2073,
+      "step": 8711
+    },
+    {
+      "epoch": 79.93,
+      "learning_rate": 0.00010538461538461538,
+      "loss": 0.1836,
+      "step": 8712
+    },
+    {
+      "epoch": 79.94,
+      "learning_rate": 0.00010533653846153847,
+      "loss": 0.1298,
+      "step": 8713
+    },
+    {
+      "epoch": 79.94,
+      "learning_rate": 0.00010528846153846154,
+      "loss": 0.2621,
+      "step": 8714
+    },
+    {
+      "epoch": 79.95,
+      "learning_rate": 0.00010524038461538461,
+      "loss": 0.1341,
+      "step": 8715
+    },
+    {
+      "epoch": 79.96,
+      "learning_rate": 0.0001051923076923077,
+      "loss": 0.1121,
+      "step": 8716
+    },
+    {
+      "epoch": 79.97,
+      "learning_rate": 0.00010514423076923077,
+      "loss": 0.1739,
+      "step": 8717
+    },
+    {
+      "epoch": 79.98,
+      "learning_rate": 0.00010509615384615386,
+      "loss": 0.1547,
+      "step": 8718
+    },
+    {
+      "epoch": 79.99,
+      "learning_rate": 0.00010504807692307693,
+      "loss": 0.098,
+      "step": 8719
+    },
+    {
+      "epoch": 80.0,
+      "learning_rate": 0.000105,
+      "loss": 0.2179,
+      "step": 8720
+    },
+    {
+      "epoch": 80.01,
+      "learning_rate": 0.00010495192307692308,
+      "loss": 0.2375,
+      "step": 8721
+    },
+    {
+      "epoch": 80.02,
+      "learning_rate": 0.00010490384615384615,
+      "loss": 0.2106,
+      "step": 8722
+    },
+    {
+      "epoch": 80.03,
+      "learning_rate": 0.00010485576923076922,
+      "loss": 0.1564,
+      "step": 8723
+    },
+    {
+      "epoch": 80.04,
+      "learning_rate": 0.00010480769230769231,
+      "loss": 0.1861,
+      "step": 8724
+    },
+    {
+      "epoch": 80.05,
+      "learning_rate": 0.00010475961538461538,
+      "loss": 0.164,
+      "step": 8725
+    },
+    {
+      "epoch": 80.06,
+      "learning_rate": 0.00010471153846153847,
+      "loss": 0.1795,
+      "step": 8726
+    },
+    {
+      "epoch": 80.06,
+      "learning_rate": 0.00010466346153846154,
+      "loss": 0.1668,
+      "step": 8727
+    },
+    {
+      "epoch": 80.07,
+      "learning_rate": 0.00010461538461538461,
+      "loss": 0.2195,
+      "step": 8728
+    },
+    {
+      "epoch": 80.08,
+      "learning_rate": 0.0001045673076923077,
+      "loss": 0.1923,
+      "step": 8729
+    },
+    {
+      "epoch": 80.09,
+      "learning_rate": 0.00010451923076923077,
+      "loss": 0.2336,
+      "step": 8730
+    },
+    {
+      "epoch": 80.1,
+      "learning_rate": 0.00010447115384615386,
+      "loss": 0.1481,
+      "step": 8731
+    },
+    {
+      "epoch": 80.11,
+      "learning_rate": 0.00010442307692307693,
+      "loss": 0.1792,
+      "step": 8732
+    },
+    {
+      "epoch": 80.12,
+      "learning_rate": 0.000104375,
+      "loss": 0.1582,
+      "step": 8733
+    },
+    {
+      "epoch": 80.13,
+      "learning_rate": 0.00010432692307692309,
+      "loss": 0.1367,
+      "step": 8734
+    },
+    {
+      "epoch": 80.14,
+      "learning_rate": 0.00010427884615384615,
+      "loss": 0.1213,
+      "step": 8735
+    },
+    {
+      "epoch": 80.15,
+      "learning_rate": 0.00010423076923076922,
+      "loss": 0.1485,
+      "step": 8736
+    },
+    {
+      "epoch": 80.16,
+      "learning_rate": 0.0001041826923076923,
+      "loss": 0.2547,
+      "step": 8737
+    },
+    {
+      "epoch": 80.17,
+      "learning_rate": 0.00010413461538461538,
+      "loss": 0.1691,
+      "step": 8738
+    },
+    {
+      "epoch": 80.17,
+      "learning_rate": 0.00010408653846153847,
+      "loss": 0.1528,
+      "step": 8739
+    },
+    {
+      "epoch": 80.18,
+      "learning_rate": 0.00010403846153846154,
+      "loss": 0.1775,
+      "step": 8740
+    },
+    {
+      "epoch": 80.19,
+      "learning_rate": 0.00010399038461538461,
+      "loss": 0.1287,
+      "step": 8741
+    },
+    {
+      "epoch": 80.2,
+      "learning_rate": 0.0001039423076923077,
+      "loss": 0.1575,
+      "step": 8742
+    },
+    {
+      "epoch": 80.21,
+      "learning_rate": 0.00010389423076923077,
+      "loss": 0.1351,
+      "step": 8743
+    },
+    {
+      "epoch": 80.22,
+      "learning_rate": 0.00010384615384615386,
+      "loss": 0.1204,
+      "step": 8744
+    },
+    {
+      "epoch": 80.23,
+      "learning_rate": 0.00010379807692307693,
+      "loss": 0.1495,
+      "step": 8745
+    },
+    {
+      "epoch": 80.24,
+      "learning_rate": 0.00010375,
+      "loss": 0.1985,
+      "step": 8746
+    },
+    {
+      "epoch": 80.25,
+      "learning_rate": 0.00010370192307692309,
+      "loss": 0.0747,
+      "step": 8747
+    },
+    {
+      "epoch": 80.26,
+      "learning_rate": 0.00010365384615384616,
+      "loss": 0.3115,
+      "step": 8748
+    },
+    {
+      "epoch": 80.27,
+      "learning_rate": 0.00010360576923076925,
+      "loss": 0.2185,
+      "step": 8749
+    },
+    {
+      "epoch": 80.28,
+      "learning_rate": 0.0001035576923076923,
+      "loss": 0.1598,
+      "step": 8750
+    },
+    {
+      "epoch": 80.28,
+      "learning_rate": 0.00010350961538461538,
+      "loss": 0.1348,
+      "step": 8751
+    },
+    {
+      "epoch": 80.29,
+      "learning_rate": 0.00010346153846153846,
+      "loss": 0.1659,
+      "step": 8752
+    },
+    {
+      "epoch": 80.3,
+      "learning_rate": 0.00010341346153846154,
+      "loss": 0.1715,
+      "step": 8753
+    },
+    {
+      "epoch": 80.31,
+      "learning_rate": 0.00010336538461538461,
+      "loss": 0.1739,
+      "step": 8754
+    },
+    {
+      "epoch": 80.32,
+      "learning_rate": 0.0001033173076923077,
+      "loss": 0.1327,
+      "step": 8755
+    },
+    {
+      "epoch": 80.33,
+      "learning_rate": 0.00010326923076923077,
+      "loss": 0.2103,
+      "step": 8756
+    },
+    {
+      "epoch": 80.34,
+      "learning_rate": 0.00010322115384615385,
+      "loss": 0.1603,
+      "step": 8757
+    },
+    {
+      "epoch": 80.35,
+      "learning_rate": 0.00010317307692307693,
+      "loss": 0.1632,
+      "step": 8758
+    },
+    {
+      "epoch": 80.36,
+      "learning_rate": 0.000103125,
+      "loss": 0.1673,
+      "step": 8759
+    },
+    {
+      "epoch": 80.37,
+      "learning_rate": 0.00010307692307692309,
+      "loss": 0.1893,
+      "step": 8760
+    },
+    {
+      "epoch": 80.38,
+      "learning_rate": 0.00010302884615384616,
+      "loss": 0.1134,
+      "step": 8761
+    },
+    {
+      "epoch": 80.39,
+      "learning_rate": 0.00010298076923076924,
+      "loss": 0.197,
+      "step": 8762
+    },
+    {
+      "epoch": 80.39,
+      "learning_rate": 0.00010293269230769232,
+      "loss": 0.2088,
+      "step": 8763
+    },
+    {
+      "epoch": 80.4,
+      "learning_rate": 0.00010288461538461538,
+      "loss": 0.1723,
+      "step": 8764
+    },
+    {
+      "epoch": 80.41,
+      "learning_rate": 0.00010283653846153846,
+      "loss": 0.2287,
+      "step": 8765
+    },
+    {
+      "epoch": 80.42,
+      "learning_rate": 0.00010278846153846153,
+      "loss": 0.3551,
+      "step": 8766
+    },
+    {
+      "epoch": 80.43,
+      "learning_rate": 0.00010274038461538462,
+      "loss": 0.1957,
+      "step": 8767
+    },
+    {
+      "epoch": 80.44,
+      "learning_rate": 0.0001026923076923077,
+      "loss": 0.1047,
+      "step": 8768
+    },
+    {
+      "epoch": 80.45,
+      "learning_rate": 0.00010264423076923077,
+      "loss": 0.2223,
+      "step": 8769
+    },
+    {
+      "epoch": 80.46,
+      "learning_rate": 0.00010259615384615385,
+      "loss": 0.159,
+      "step": 8770
+    },
+    {
+      "epoch": 80.47,
+      "learning_rate": 0.00010254807692307693,
+      "loss": 0.0886,
+      "step": 8771
+    },
+    {
+      "epoch": 80.48,
+      "learning_rate": 0.0001025,
+      "loss": 0.1547,
+      "step": 8772
+    },
+    {
+      "epoch": 80.49,
+      "learning_rate": 0.00010245192307692308,
+      "loss": 0.1902,
+      "step": 8773
+    },
+    {
+      "epoch": 80.5,
+      "learning_rate": 0.00010240384615384616,
+      "loss": 0.1211,
+      "step": 8774
+    },
+    {
+      "epoch": 80.5,
+      "learning_rate": 0.00010235576923076924,
+      "loss": 0.2338,
+      "step": 8775
+    },
+    {
+      "epoch": 80.51,
+      "learning_rate": 0.00010230769230769232,
+      "loss": 0.1758,
+      "step": 8776
+    },
+    {
+      "epoch": 80.52,
+      "learning_rate": 0.00010225961538461539,
+      "loss": 0.2171,
+      "step": 8777
+    },
+    {
+      "epoch": 80.53,
+      "learning_rate": 0.00010221153846153846,
+      "loss": 0.1725,
+      "step": 8778
+    },
+    {
+      "epoch": 80.54,
+      "learning_rate": 0.00010216346153846153,
+      "loss": 0.2064,
+      "step": 8779
+    },
+    {
+      "epoch": 80.55,
+      "learning_rate": 0.00010211538461538462,
+      "loss": 0.135,
+      "step": 8780
+    },
+    {
+      "epoch": 80.56,
+      "learning_rate": 0.00010206730769230769,
+      "loss": 0.2384,
+      "step": 8781
+    },
+    {
+      "epoch": 80.57,
+      "learning_rate": 0.00010201923076923076,
+      "loss": 0.1691,
+      "step": 8782
+    },
+    {
+      "epoch": 80.58,
+      "learning_rate": 0.00010197115384615385,
+      "loss": 0.2001,
+      "step": 8783
+    },
+    {
+      "epoch": 80.59,
+      "learning_rate": 0.00010192307692307692,
+      "loss": 0.228,
+      "step": 8784
+    },
+    {
+      "epoch": 80.6,
+      "learning_rate": 0.000101875,
+      "loss": 0.1574,
+      "step": 8785
+    },
+    {
+      "epoch": 80.61,
+      "learning_rate": 0.00010182692307692308,
+      "loss": 0.2079,
+      "step": 8786
+    },
+    {
+      "epoch": 80.61,
+      "learning_rate": 0.00010177884615384615,
+      "loss": 0.114,
+      "step": 8787
+    },
+    {
+      "epoch": 80.62,
+      "learning_rate": 0.00010173076923076924,
+      "loss": 0.1913,
+      "step": 8788
+    },
+    {
+      "epoch": 80.63,
+      "learning_rate": 0.00010168269230769231,
+      "loss": 0.1796,
+      "step": 8789
+    },
+    {
+      "epoch": 80.64,
+      "learning_rate": 0.00010163461538461539,
+      "loss": 0.231,
+      "step": 8790
+    },
+    {
+      "epoch": 80.65,
+      "learning_rate": 0.00010158653846153847,
+      "loss": 0.1473,
+      "step": 8791
+    },
+    {
+      "epoch": 80.66,
+      "learning_rate": 0.00010153846153846153,
+      "loss": 0.2056,
+      "step": 8792
+    },
+    {
+      "epoch": 80.67,
+      "learning_rate": 0.00010149038461538462,
+      "loss": 0.1682,
+      "step": 8793
+    },
+    {
+      "epoch": 80.68,
+      "learning_rate": 0.00010144230769230769,
+      "loss": 0.1213,
+      "step": 8794
+    },
+    {
+      "epoch": 80.69,
+      "learning_rate": 0.00010139423076923076,
+      "loss": 0.1772,
+      "step": 8795
+    },
+    {
+      "epoch": 80.7,
+      "learning_rate": 0.00010134615384615385,
+      "loss": 0.1219,
+      "step": 8796
+    },
+    {
+      "epoch": 80.71,
+      "learning_rate": 0.00010129807692307692,
+      "loss": 0.0901,
+      "step": 8797
+    },
+    {
+      "epoch": 80.72,
+      "learning_rate": 0.00010125000000000001,
+      "loss": 0.239,
+      "step": 8798
+    },
+    {
+      "epoch": 80.72,
+      "learning_rate": 0.00010120192307692308,
+      "loss": 0.1445,
+      "step": 8799
+    },
+    {
+      "epoch": 80.73,
+      "learning_rate": 0.00010115384615384615,
+      "loss": 0.1223,
+      "step": 8800
+    },
+    {
+      "epoch": 80.74,
+      "learning_rate": 0.00010110576923076924,
+      "loss": 0.1207,
+      "step": 8801
+    },
+    {
+      "epoch": 80.75,
+      "learning_rate": 0.00010105769230769231,
+      "loss": 0.273,
+      "step": 8802
+    },
+    {
+      "epoch": 80.76,
+      "learning_rate": 0.00010100961538461538,
+      "loss": 0.1871,
+      "step": 8803
+    },
+    {
+      "epoch": 80.77,
+      "learning_rate": 0.00010096153846153847,
+      "loss": 0.2365,
+      "step": 8804
+    },
+    {
+      "epoch": 80.78,
+      "learning_rate": 0.00010091346153846154,
+      "loss": 0.2009,
+      "step": 8805
+    },
+    {
+      "epoch": 80.79,
+      "learning_rate": 0.00010086538461538462,
+      "loss": 0.167,
+      "step": 8806
+    },
+    {
+      "epoch": 80.8,
+      "learning_rate": 0.00010081730769230769,
+      "loss": 0.2015,
+      "step": 8807
+    },
+    {
+      "epoch": 80.81,
+      "learning_rate": 0.00010076923076923076,
+      "loss": 0.2423,
+      "step": 8808
+    },
+    {
+      "epoch": 80.82,
+      "learning_rate": 0.00010072115384615385,
+      "loss": 0.1915,
+      "step": 8809
+    },
+    {
+      "epoch": 80.83,
+      "learning_rate": 0.00010067307692307692,
+      "loss": 0.1434,
+      "step": 8810
+    },
+    {
+      "epoch": 80.83,
+      "learning_rate": 0.000100625,
+      "loss": 0.1765,
+      "step": 8811
+    },
+    {
+      "epoch": 80.84,
+      "learning_rate": 0.00010057692307692308,
+      "loss": 0.1057,
+      "step": 8812
+    },
+    {
+      "epoch": 80.85,
+      "learning_rate": 0.00010052884615384615,
+      "loss": 0.1705,
+      "step": 8813
+    },
+    {
+      "epoch": 80.86,
+      "learning_rate": 0.00010048076923076924,
+      "loss": 0.1929,
+      "step": 8814
+    },
+    {
+      "epoch": 80.87,
+      "learning_rate": 0.00010043269230769231,
+      "loss": 0.2014,
+      "step": 8815
+    },
+    {
+      "epoch": 80.88,
+      "learning_rate": 0.00010038461538461538,
+      "loss": 0.1552,
+      "step": 8816
+    },
+    {
+      "epoch": 80.89,
+      "learning_rate": 0.00010033653846153847,
+      "loss": 0.2327,
+      "step": 8817
+    },
+    {
+      "epoch": 80.9,
+      "learning_rate": 0.00010028846153846154,
+      "loss": 0.1951,
+      "step": 8818
+    },
+    {
+      "epoch": 80.91,
+      "learning_rate": 0.00010024038461538463,
+      "loss": 0.0997,
+      "step": 8819
+    },
+    {
+      "epoch": 80.92,
+      "learning_rate": 0.00010019230769230769,
+      "loss": 0.1733,
+      "step": 8820
+    },
+    {
+      "epoch": 80.93,
+      "learning_rate": 0.00010014423076923076,
+      "loss": 0.0944,
+      "step": 8821
+    },
+    {
+      "epoch": 80.94,
+      "learning_rate": 0.00010009615384615385,
+      "loss": 0.2342,
+      "step": 8822
+    },
+    {
+      "epoch": 80.94,
+      "learning_rate": 0.00010004807692307692,
+      "loss": 0.1735,
+      "step": 8823
+    },
+    {
+      "epoch": 80.95,
+      "learning_rate": 0.0001,
+      "loss": 0.1933,
+      "step": 8824
+    },
+    {
+      "epoch": 80.96,
+      "learning_rate": 9.995192307692308e-05,
+      "loss": 0.1457,
+      "step": 8825
+    },
+    {
+      "epoch": 80.97,
+      "learning_rate": 9.990384615384615e-05,
+      "loss": 0.157,
+      "step": 8826
+    },
+    {
+      "epoch": 80.98,
+      "learning_rate": 9.985576923076924e-05,
+      "loss": 0.1052,
+      "step": 8827
+    },
+    {
+      "epoch": 80.99,
+      "learning_rate": 9.980769230769231e-05,
+      "loss": 0.1254,
+      "step": 8828
+    },
+    {
+      "epoch": 81.0,
+      "learning_rate": 9.97596153846154e-05,
+      "loss": 0.1873,
+      "step": 8829
+    },
+    {
+      "epoch": 81.01,
+      "learning_rate": 9.971153846153847e-05,
+      "loss": 0.2202,
+      "step": 8830
+    },
+    {
+      "epoch": 81.02,
+      "learning_rate": 9.966346153846154e-05,
+      "loss": 0.2045,
+      "step": 8831
+    },
+    {
+      "epoch": 81.03,
+      "learning_rate": 9.961538461538463e-05,
+      "loss": 0.2409,
+      "step": 8832
+    },
+    {
+      "epoch": 81.04,
+      "learning_rate": 9.95673076923077e-05,
+      "loss": 0.1622,
+      "step": 8833
+    },
+    {
+      "epoch": 81.05,
+      "learning_rate": 9.951923076923076e-05,
+      "loss": 0.1881,
+      "step": 8834
+    },
+    {
+      "epoch": 81.06,
+      "learning_rate": 9.947115384615384e-05,
+      "loss": 0.2024,
+      "step": 8835
+    },
+    {
+      "epoch": 81.06,
+      "learning_rate": 9.942307692307692e-05,
+      "loss": 0.2007,
+      "step": 8836
+    },
+    {
+      "epoch": 81.07,
+      "learning_rate": 9.9375e-05,
+      "loss": 0.193,
+      "step": 8837
+    },
+    {
+      "epoch": 81.08,
+      "learning_rate": 9.932692307692308e-05,
+      "loss": 0.1825,
+      "step": 8838
+    },
+    {
+      "epoch": 81.09,
+      "learning_rate": 9.927884615384615e-05,
+      "loss": 0.1058,
+      "step": 8839
+    },
+    {
+      "epoch": 81.1,
+      "learning_rate": 9.923076923076923e-05,
+      "loss": 0.2234,
+      "step": 8840
+    },
+    {
+      "epoch": 81.11,
+      "learning_rate": 9.918269230769231e-05,
+      "loss": 0.1705,
+      "step": 8841
+    },
+    {
+      "epoch": 81.12,
+      "learning_rate": 9.91346153846154e-05,
+      "loss": 0.1159,
+      "step": 8842
+    },
+    {
+      "epoch": 81.13,
+      "learning_rate": 9.908653846153847e-05,
+      "loss": 0.1542,
+      "step": 8843
+    },
+    {
+      "epoch": 81.14,
+      "learning_rate": 9.903846153846154e-05,
+      "loss": 0.1438,
+      "step": 8844
+    },
+    {
+      "epoch": 81.15,
+      "learning_rate": 9.899038461538462e-05,
+      "loss": 0.1384,
+      "step": 8845
+    },
+    {
+      "epoch": 81.16,
+      "learning_rate": 9.89423076923077e-05,
+      "loss": 0.1077,
+      "step": 8846
+    },
+    {
+      "epoch": 81.17,
+      "learning_rate": 9.889423076923078e-05,
+      "loss": 0.1044,
+      "step": 8847
+    },
+    {
+      "epoch": 81.17,
+      "learning_rate": 9.884615384615386e-05,
+      "loss": 0.1732,
+      "step": 8848
+    },
+    {
+      "epoch": 81.18,
+      "learning_rate": 9.879807692307692e-05,
+      "loss": 0.1723,
+      "step": 8849
+    },
+    {
+      "epoch": 81.19,
+      "learning_rate": 9.875e-05,
+      "loss": 0.1374,
+      "step": 8850
+    },
+    {
+      "epoch": 81.2,
+      "learning_rate": 9.870192307692307e-05,
+      "loss": 0.1515,
+      "step": 8851
+    },
+    {
+      "epoch": 81.21,
+      "learning_rate": 9.865384615384615e-05,
+      "loss": 0.1963,
+      "step": 8852
+    },
+    {
+      "epoch": 81.22,
+      "learning_rate": 9.860576923076923e-05,
+      "loss": 0.1741,
+      "step": 8853
+    },
+    {
+      "epoch": 81.23,
+      "learning_rate": 9.85576923076923e-05,
+      "loss": 0.1142,
+      "step": 8854
+    },
+    {
+      "epoch": 81.24,
+      "learning_rate": 9.850961538461539e-05,
+      "loss": 0.139,
+      "step": 8855
+    },
+    {
+      "epoch": 81.25,
+      "learning_rate": 9.846153846153846e-05,
+      "loss": 0.1003,
+      "step": 8856
+    },
+    {
+      "epoch": 81.26,
+      "learning_rate": 9.841346153846154e-05,
+      "loss": 0.3042,
+      "step": 8857
+    },
+    {
+      "epoch": 81.27,
+      "learning_rate": 9.836538461538462e-05,
+      "loss": 0.1662,
+      "step": 8858
+    },
+    {
+      "epoch": 81.28,
+      "learning_rate": 9.83173076923077e-05,
+      "loss": 0.1546,
+      "step": 8859
+    },
+    {
+      "epoch": 81.28,
+      "learning_rate": 9.826923076923078e-05,
+      "loss": 0.1483,
+      "step": 8860
+    },
+    {
+      "epoch": 81.29,
+      "learning_rate": 9.822115384615385e-05,
+      "loss": 0.1927,
+      "step": 8861
+    },
+    {
+      "epoch": 81.3,
+      "learning_rate": 9.817307692307693e-05,
+      "loss": 0.1437,
+      "step": 8862
+    },
+    {
+      "epoch": 81.31,
+      "learning_rate": 9.8125e-05,
+      "loss": 0.1636,
+      "step": 8863
+    },
+    {
+      "epoch": 81.32,
+      "learning_rate": 9.807692307692307e-05,
+      "loss": 0.206,
+      "step": 8864
+    },
+    {
+      "epoch": 81.33,
+      "learning_rate": 9.802884615384615e-05,
+      "loss": 0.2844,
+      "step": 8865
+    },
+    {
+      "epoch": 81.34,
+      "learning_rate": 9.798076923076923e-05,
+      "loss": 0.1835,
+      "step": 8866
+    },
+    {
+      "epoch": 81.35,
+      "learning_rate": 9.79326923076923e-05,
+      "loss": 0.1753,
+      "step": 8867
+    },
+    {
+      "epoch": 81.36,
+      "learning_rate": 9.788461538461539e-05,
+      "loss": 0.1718,
+      "step": 8868
+    },
+    {
+      "epoch": 81.37,
+      "learning_rate": 9.783653846153846e-05,
+      "loss": 0.2207,
+      "step": 8869
+    },
+    {
+      "epoch": 81.38,
+      "learning_rate": 9.778846153846154e-05,
+      "loss": 0.1967,
+      "step": 8870
+    },
+    {
+      "epoch": 81.39,
+      "learning_rate": 9.774038461538462e-05,
+      "loss": 0.2389,
+      "step": 8871
+    },
+    {
+      "epoch": 81.39,
+      "learning_rate": 9.76923076923077e-05,
+      "loss": 0.1278,
+      "step": 8872
+    },
+    {
+      "epoch": 81.4,
+      "learning_rate": 9.764423076923078e-05,
+      "loss": 0.0897,
+      "step": 8873
+    },
+    {
+      "epoch": 81.41,
+      "learning_rate": 9.759615384615385e-05,
+      "loss": 0.1395,
+      "step": 8874
+    },
+    {
+      "epoch": 81.42,
+      "learning_rate": 9.754807692307693e-05,
+      "loss": 0.1636,
+      "step": 8875
+    },
+    {
+      "epoch": 81.43,
+      "learning_rate": 9.750000000000001e-05,
+      "loss": 0.154,
+      "step": 8876
+    },
+    {
+      "epoch": 81.44,
+      "learning_rate": 9.745192307692307e-05,
+      "loss": 0.1978,
+      "step": 8877
+    },
+    {
+      "epoch": 81.45,
+      "learning_rate": 9.740384615384616e-05,
+      "loss": 0.1423,
+      "step": 8878
+    },
+    {
+      "epoch": 81.46,
+      "learning_rate": 9.735576923076923e-05,
+      "loss": 0.1681,
+      "step": 8879
+    },
+    {
+      "epoch": 81.47,
+      "learning_rate": 9.73076923076923e-05,
+      "loss": 0.097,
+      "step": 8880
+    },
+    {
+      "epoch": 81.48,
+      "learning_rate": 9.725961538461539e-05,
+      "loss": 0.255,
+      "step": 8881
+    },
+    {
+      "epoch": 81.49,
+      "learning_rate": 9.721153846153846e-05,
+      "loss": 0.0634,
+      "step": 8882
+    },
+    {
+      "epoch": 81.5,
+      "learning_rate": 9.716346153846153e-05,
+      "loss": 0.0535,
+      "step": 8883
+    },
+    {
+      "epoch": 81.5,
+      "learning_rate": 9.711538461538462e-05,
+      "loss": 0.2108,
+      "step": 8884
+    },
+    {
+      "epoch": 81.51,
+      "learning_rate": 9.706730769230769e-05,
+      "loss": 0.1821,
+      "step": 8885
+    },
+    {
+      "epoch": 81.52,
+      "learning_rate": 9.701923076923078e-05,
+      "loss": 0.1164,
+      "step": 8886
+    },
+    {
+      "epoch": 81.53,
+      "learning_rate": 9.697115384615385e-05,
+      "loss": 0.1521,
+      "step": 8887
+    },
+    {
+      "epoch": 81.54,
+      "learning_rate": 9.692307692307692e-05,
+      "loss": 0.2102,
+      "step": 8888
+    },
+    {
+      "epoch": 81.55,
+      "learning_rate": 9.687500000000001e-05,
+      "loss": 0.1987,
+      "step": 8889
+    },
+    {
+      "epoch": 81.56,
+      "learning_rate": 9.682692307692308e-05,
+      "loss": 0.1712,
+      "step": 8890
+    },
+    {
+      "epoch": 81.57,
+      "learning_rate": 9.677884615384616e-05,
+      "loss": 0.1985,
+      "step": 8891
+    },
+    {
+      "epoch": 81.58,
+      "learning_rate": 9.673076923076923e-05,
+      "loss": 0.2145,
+      "step": 8892
+    },
+    {
+      "epoch": 81.59,
+      "learning_rate": 9.66826923076923e-05,
+      "loss": 0.1715,
+      "step": 8893
+    },
+    {
+      "epoch": 81.6,
+      "learning_rate": 9.663461538461539e-05,
+      "loss": 0.2151,
+      "step": 8894
+    },
+    {
+      "epoch": 81.61,
+      "learning_rate": 9.658653846153846e-05,
+      "loss": 0.1592,
+      "step": 8895
+    },
+    {
+      "epoch": 81.61,
+      "learning_rate": 9.653846153846155e-05,
+      "loss": 0.2112,
+      "step": 8896
+    },
+    {
+      "epoch": 81.62,
+      "learning_rate": 9.649038461538462e-05,
+      "loss": 0.1473,
+      "step": 8897
+    },
+    {
+      "epoch": 81.63,
+      "learning_rate": 9.644230769230769e-05,
+      "loss": 0.1448,
+      "step": 8898
+    },
+    {
+      "epoch": 81.64,
+      "learning_rate": 9.639423076923078e-05,
+      "loss": 0.2321,
+      "step": 8899
+    },
+    {
+      "epoch": 81.65,
+      "learning_rate": 9.634615384615385e-05,
+      "loss": 0.229,
+      "step": 8900
+    },
+    {
+      "epoch": 81.66,
+      "learning_rate": 9.629807692307692e-05,
+      "loss": 0.1707,
+      "step": 8901
+    },
+    {
+      "epoch": 81.67,
+      "learning_rate": 9.625000000000001e-05,
+      "loss": 0.1468,
+      "step": 8902
+    },
+    {
+      "epoch": 81.68,
+      "learning_rate": 9.620192307692308e-05,
+      "loss": 0.1355,
+      "step": 8903
+    },
+    {
+      "epoch": 81.69,
+      "learning_rate": 9.615384615384617e-05,
+      "loss": 0.2036,
+      "step": 8904
+    },
+    {
+      "epoch": 81.7,
+      "learning_rate": 9.610576923076923e-05,
+      "loss": 0.195,
+      "step": 8905
+    },
+    {
+      "epoch": 81.71,
+      "learning_rate": 9.60576923076923e-05,
+      "loss": 0.171,
+      "step": 8906
+    },
+    {
+      "epoch": 81.72,
+      "learning_rate": 9.600961538461539e-05,
+      "loss": 0.1492,
+      "step": 8907
+    },
+    {
+      "epoch": 81.72,
+      "learning_rate": 9.596153846153846e-05,
+      "loss": 0.1278,
+      "step": 8908
+    },
+    {
+      "epoch": 81.73,
+      "learning_rate": 9.591346153846154e-05,
+      "loss": 0.1754,
+      "step": 8909
+    },
+    {
+      "epoch": 81.74,
+      "learning_rate": 9.586538461538462e-05,
+      "loss": 0.143,
+      "step": 8910
+    },
+    {
+      "epoch": 81.75,
+      "learning_rate": 9.581730769230769e-05,
+      "loss": 0.2,
+      "step": 8911
+    },
+    {
+      "epoch": 81.76,
+      "learning_rate": 9.576923076923078e-05,
+      "loss": 0.1738,
+      "step": 8912
+    },
+    {
+      "epoch": 81.77,
+      "learning_rate": 9.572115384615385e-05,
+      "loss": 0.2109,
+      "step": 8913
+    },
+    {
+      "epoch": 81.78,
+      "learning_rate": 9.567307692307692e-05,
+      "loss": 0.1676,
+      "step": 8914
+    },
+    {
+      "epoch": 81.79,
+      "learning_rate": 9.562500000000001e-05,
+      "loss": 0.1776,
+      "step": 8915
+    },
+    {
+      "epoch": 81.8,
+      "learning_rate": 9.557692307692308e-05,
+      "loss": 0.1332,
+      "step": 8916
+    },
+    {
+      "epoch": 81.81,
+      "learning_rate": 9.552884615384617e-05,
+      "loss": 0.1564,
+      "step": 8917
+    },
+    {
+      "epoch": 81.82,
+      "learning_rate": 9.548076923076924e-05,
+      "loss": 0.1586,
+      "step": 8918
+    },
+    {
+      "epoch": 81.83,
+      "learning_rate": 9.54326923076923e-05,
+      "loss": 0.1512,
+      "step": 8919
+    },
+    {
+      "epoch": 81.83,
+      "learning_rate": 9.538461538461538e-05,
+      "loss": 0.1286,
+      "step": 8920
+    },
+    {
+      "epoch": 81.84,
+      "learning_rate": 9.533653846153846e-05,
+      "loss": 0.1448,
+      "step": 8921
+    },
+    {
+      "epoch": 81.85,
+      "learning_rate": 9.528846153846154e-05,
+      "loss": 0.1284,
+      "step": 8922
+    },
+    {
+      "epoch": 81.86,
+      "learning_rate": 9.524038461538461e-05,
+      "loss": 0.1553,
+      "step": 8923
+    },
+    {
+      "epoch": 81.87,
+      "learning_rate": 9.519230769230769e-05,
+      "loss": 0.1308,
+      "step": 8924
+    },
+    {
+      "epoch": 81.88,
+      "learning_rate": 9.514423076923077e-05,
+      "loss": 0.1008,
+      "step": 8925
+    },
+    {
+      "epoch": 81.89,
+      "learning_rate": 9.509615384615385e-05,
+      "loss": 0.1576,
+      "step": 8926
+    },
+    {
+      "epoch": 81.9,
+      "learning_rate": 9.504807692307693e-05,
+      "loss": 0.1304,
+      "step": 8927
+    },
+    {
+      "epoch": 81.91,
+      "learning_rate": 9.5e-05,
+      "loss": 0.1831,
+      "step": 8928
+    },
+    {
+      "epoch": 81.92,
+      "learning_rate": 9.495192307692308e-05,
+      "loss": 0.1907,
+      "step": 8929
+    },
+    {
+      "epoch": 81.93,
+      "learning_rate": 9.490384615384616e-05,
+      "loss": 0.1377,
+      "step": 8930
+    },
+    {
+      "epoch": 81.94,
+      "learning_rate": 9.485576923076924e-05,
+      "loss": 0.174,
+      "step": 8931
+    },
+    {
+      "epoch": 81.94,
+      "learning_rate": 9.480769230769231e-05,
+      "loss": 0.1341,
+      "step": 8932
+    },
+    {
+      "epoch": 81.95,
+      "learning_rate": 9.475961538461538e-05,
+      "loss": 0.2077,
+      "step": 8933
+    },
+    {
+      "epoch": 81.96,
+      "learning_rate": 9.471153846153845e-05,
+      "loss": 0.1159,
+      "step": 8934
+    },
+    {
+      "epoch": 81.97,
+      "learning_rate": 9.466346153846154e-05,
+      "loss": 0.1112,
+      "step": 8935
+    },
+    {
+      "epoch": 81.98,
+      "learning_rate": 9.461538461538461e-05,
+      "loss": 0.095,
+      "step": 8936
+    },
+    {
+      "epoch": 81.99,
+      "learning_rate": 9.456730769230769e-05,
+      "loss": 0.1087,
+      "step": 8937
+    },
+    {
+      "epoch": 82.0,
+      "learning_rate": 9.451923076923077e-05,
+      "loss": 0.1969,
+      "step": 8938
+    },
+    {
+      "epoch": 82.01,
+      "learning_rate": 9.447115384615384e-05,
+      "loss": 0.2326,
+      "step": 8939
+    },
+    {
+      "epoch": 82.02,
+      "learning_rate": 9.442307692307693e-05,
+      "loss": 0.1962,
+      "step": 8940
+    },
+    {
+      "epoch": 82.03,
+      "learning_rate": 9.4375e-05,
+      "loss": 0.2396,
+      "step": 8941
+    },
+    {
+      "epoch": 82.04,
+      "learning_rate": 9.432692307692308e-05,
+      "loss": 0.1181,
+      "step": 8942
+    },
+    {
+      "epoch": 82.05,
+      "learning_rate": 9.427884615384616e-05,
+      "loss": 0.1833,
+      "step": 8943
+    },
+    {
+      "epoch": 82.06,
+      "learning_rate": 9.423076923076924e-05,
+      "loss": 0.1965,
+      "step": 8944
+    },
+    {
+      "epoch": 82.06,
+      "learning_rate": 9.418269230769232e-05,
+      "loss": 0.1387,
+      "step": 8945
+    },
+    {
+      "epoch": 82.07,
+      "learning_rate": 9.41346153846154e-05,
+      "loss": 0.173,
+      "step": 8946
+    },
+    {
+      "epoch": 82.08,
+      "learning_rate": 9.408653846153845e-05,
+      "loss": 0.1952,
+      "step": 8947
+    },
+    {
+      "epoch": 82.09,
+      "learning_rate": 9.403846153846154e-05,
+      "loss": 0.1242,
+      "step": 8948
+    },
+    {
+      "epoch": 82.1,
+      "learning_rate": 9.399038461538461e-05,
+      "loss": 0.1505,
+      "step": 8949
+    },
+    {
+      "epoch": 82.11,
+      "learning_rate": 9.394230769230768e-05,
+      "loss": 0.1579,
+      "step": 8950
+    },
+    {
+      "epoch": 82.12,
+      "learning_rate": 9.389423076923077e-05,
+      "loss": 0.155,
+      "step": 8951
+    },
+    {
+      "epoch": 82.13,
+      "learning_rate": 9.384615384615384e-05,
+      "loss": 0.2281,
+      "step": 8952
+    },
+    {
+      "epoch": 82.14,
+      "learning_rate": 9.379807692307693e-05,
+      "loss": 0.1866,
+      "step": 8953
+    },
+    {
+      "epoch": 82.15,
+      "learning_rate": 9.375e-05,
+      "loss": 0.1177,
+      "step": 8954
+    },
+    {
+      "epoch": 82.16,
+      "learning_rate": 9.370192307692307e-05,
+      "loss": 0.1574,
+      "step": 8955
+    },
+    {
+      "epoch": 82.17,
+      "learning_rate": 9.365384615384616e-05,
+      "loss": 0.1272,
+      "step": 8956
+    },
+    {
+      "epoch": 82.17,
+      "learning_rate": 9.360576923076923e-05,
+      "loss": 0.1516,
+      "step": 8957
+    },
+    {
+      "epoch": 82.18,
+      "learning_rate": 9.355769230769232e-05,
+      "loss": 0.1613,
+      "step": 8958
+    },
+    {
+      "epoch": 82.19,
+      "learning_rate": 9.350961538461539e-05,
+      "loss": 0.1199,
+      "step": 8959
+    },
+    {
+      "epoch": 82.2,
+      "learning_rate": 9.346153846153846e-05,
+      "loss": 0.1317,
+      "step": 8960
+    },
+    {
+      "epoch": 82.21,
+      "learning_rate": 9.341346153846155e-05,
+      "loss": 0.1455,
+      "step": 8961
+    },
+    {
+      "epoch": 82.22,
+      "learning_rate": 9.336538461538461e-05,
+      "loss": 0.1576,
+      "step": 8962
+    },
+    {
+      "epoch": 82.23,
+      "learning_rate": 9.331730769230768e-05,
+      "loss": 0.1535,
+      "step": 8963
+    },
+    {
+      "epoch": 82.24,
+      "learning_rate": 9.326923076923077e-05,
+      "loss": 0.1525,
+      "step": 8964
+    },
+    {
+      "epoch": 82.25,
+      "learning_rate": 9.322115384615384e-05,
+      "loss": 0.1139,
+      "step": 8965
+    },
+    {
+      "epoch": 82.26,
+      "learning_rate": 9.317307692307693e-05,
+      "loss": 0.2594,
+      "step": 8966
+    },
+    {
+      "epoch": 82.27,
+      "learning_rate": 9.3125e-05,
+      "loss": 0.1752,
+      "step": 8967
+    },
+    {
+      "epoch": 82.28,
+      "learning_rate": 9.307692307692307e-05,
+      "loss": 0.2531,
+      "step": 8968
+    },
+    {
+      "epoch": 82.28,
+      "learning_rate": 9.302884615384616e-05,
+      "loss": 0.155,
+      "step": 8969
+    },
+    {
+      "epoch": 82.29,
+      "learning_rate": 9.298076923076923e-05,
+      "loss": 0.1101,
+      "step": 8970
+    },
+    {
+      "epoch": 82.3,
+      "learning_rate": 9.293269230769232e-05,
+      "loss": 0.1273,
+      "step": 8971
+    },
+    {
+      "epoch": 82.31,
+      "learning_rate": 9.288461538461539e-05,
+      "loss": 0.1913,
+      "step": 8972
+    },
+    {
+      "epoch": 82.32,
+      "learning_rate": 9.283653846153846e-05,
+      "loss": 0.1089,
+      "step": 8973
+    },
+    {
+      "epoch": 82.33,
+      "learning_rate": 9.278846153846155e-05,
+      "loss": 0.1623,
+      "step": 8974
+    },
+    {
+      "epoch": 82.34,
+      "learning_rate": 9.274038461538462e-05,
+      "loss": 0.1425,
+      "step": 8975
+    },
+    {
+      "epoch": 82.35,
+      "learning_rate": 9.26923076923077e-05,
+      "loss": 0.1888,
+      "step": 8976
+    },
+    {
+      "epoch": 82.36,
+      "learning_rate": 9.264423076923077e-05,
+      "loss": 0.1358,
+      "step": 8977
+    },
+    {
+      "epoch": 82.37,
+      "learning_rate": 9.259615384615384e-05,
+      "loss": 0.1379,
+      "step": 8978
+    },
+    {
+      "epoch": 82.38,
+      "learning_rate": 9.254807692307693e-05,
+      "loss": 0.1365,
+      "step": 8979
+    },
+    {
+      "epoch": 82.39,
+      "learning_rate": 9.25e-05,
+      "loss": 0.1398,
+      "step": 8980
+    },
+    {
+      "epoch": 82.39,
+      "learning_rate": 9.245192307692307e-05,
+      "loss": 0.104,
+      "step": 8981
+    },
+    {
+      "epoch": 82.4,
+      "learning_rate": 9.240384615384616e-05,
+      "loss": 0.1755,
+      "step": 8982
+    },
+    {
+      "epoch": 82.41,
+      "learning_rate": 9.235576923076923e-05,
+      "loss": 0.0836,
+      "step": 8983
+    },
+    {
+      "epoch": 82.42,
+      "learning_rate": 9.230769230769232e-05,
+      "loss": 0.1617,
+      "step": 8984
+    },
+    {
+      "epoch": 82.43,
+      "learning_rate": 9.225961538461539e-05,
+      "loss": 0.1419,
+      "step": 8985
+    },
+    {
+      "epoch": 82.44,
+      "learning_rate": 9.221153846153846e-05,
+      "loss": 0.1154,
+      "step": 8986
+    },
+    {
+      "epoch": 82.45,
+      "learning_rate": 9.216346153846155e-05,
+      "loss": 0.1655,
+      "step": 8987
+    },
+    {
+      "epoch": 82.46,
+      "learning_rate": 9.211538461538462e-05,
+      "loss": 0.2006,
+      "step": 8988
+    },
+    {
+      "epoch": 82.47,
+      "learning_rate": 9.20673076923077e-05,
+      "loss": 0.1207,
+      "step": 8989
+    },
+    {
+      "epoch": 82.48,
+      "learning_rate": 9.201923076923077e-05,
+      "loss": 0.189,
+      "step": 8990
+    },
+    {
+      "epoch": 82.49,
+      "learning_rate": 9.197115384615384e-05,
+      "loss": 0.1609,
+      "step": 8991
+    },
+    {
+      "epoch": 82.5,
+      "learning_rate": 9.192307692307692e-05,
+      "loss": 0.1997,
+      "step": 8992
+    },
+    {
+      "epoch": 82.5,
+      "learning_rate": 9.1875e-05,
+      "loss": 0.2666,
+      "step": 8993
+    },
+    {
+      "epoch": 82.51,
+      "learning_rate": 9.182692307692308e-05,
+      "loss": 0.1912,
+      "step": 8994
+    },
+    {
+      "epoch": 82.52,
+      "learning_rate": 9.177884615384616e-05,
+      "loss": 0.1681,
+      "step": 8995
+    },
+    {
+      "epoch": 82.53,
+      "learning_rate": 9.173076923076923e-05,
+      "loss": 0.1693,
+      "step": 8996
+    },
+    {
+      "epoch": 82.54,
+      "learning_rate": 9.168269230769231e-05,
+      "loss": 0.2184,
+      "step": 8997
+    },
+    {
+      "epoch": 82.55,
+      "learning_rate": 9.163461538461539e-05,
+      "loss": 0.1632,
+      "step": 8998
+    },
+    {
+      "epoch": 82.56,
+      "learning_rate": 9.158653846153846e-05,
+      "loss": 0.1693,
+      "step": 8999
+    },
+    {
+      "epoch": 82.57,
+      "learning_rate": 9.153846153846155e-05,
+      "loss": 0.2005,
+      "step": 9000
+    },
+    {
+      "epoch": 82.57,
+      "eval_cer": 0.14295566650174665,
+      "eval_loss": 0.7228291034698486,
+      "eval_runtime": 87.0833,
+      "eval_samples_per_second": 18.913,
+      "eval_steps_per_second": 2.366,
+      "eval_wer": 0.5045449902972117,
+      "step": 9000
+    },
+    {
+      "epoch": 82.58,
+      "learning_rate": 9.149038461538462e-05,
+      "loss": 0.1661,
+      "step": 9001
+    },
+    {
+      "epoch": 82.59,
+      "learning_rate": 9.14423076923077e-05,
+      "loss": 0.1566,
+      "step": 9002
+    },
+    {
+      "epoch": 82.6,
+      "learning_rate": 9.139423076923078e-05,
+      "loss": 0.1465,
+      "step": 9003
+    },
+    {
+      "epoch": 82.61,
+      "learning_rate": 9.134615384615384e-05,
+      "loss": 0.2009,
+      "step": 9004
+    },
+    {
+      "epoch": 82.61,
+      "learning_rate": 9.129807692307692e-05,
+      "loss": 0.119,
+      "step": 9005
+    },
+    {
+      "epoch": 82.62,
+      "learning_rate": 9.125e-05,
+      "loss": 0.141,
+      "step": 9006
+    },
+    {
+      "epoch": 82.63,
+      "learning_rate": 9.120192307692308e-05,
+      "loss": 0.1953,
+      "step": 9007
+    },
+    {
+      "epoch": 82.64,
+      "learning_rate": 9.115384615384615e-05,
+      "loss": 0.1823,
+      "step": 9008
+    },
+    {
+      "epoch": 82.65,
+      "learning_rate": 9.110576923076923e-05,
+      "loss": 0.16,
+      "step": 9009
+    },
+    {
+      "epoch": 82.66,
+      "learning_rate": 9.105769230769231e-05,
+      "loss": 0.2436,
+      "step": 9010
+    },
+    {
+      "epoch": 82.67,
+      "learning_rate": 9.100961538461539e-05,
+      "loss": 0.1125,
+      "step": 9011
+    },
+    {
+      "epoch": 82.68,
+      "learning_rate": 9.096153846153846e-05,
+      "loss": 0.1324,
+      "step": 9012
+    },
+    {
+      "epoch": 82.69,
+      "learning_rate": 9.091346153846154e-05,
+      "loss": 0.1596,
+      "step": 9013
+    },
+    {
+      "epoch": 82.7,
+      "learning_rate": 9.086538461538462e-05,
+      "loss": 0.1917,
+      "step": 9014
+    },
+    {
+      "epoch": 82.71,
+      "learning_rate": 9.08173076923077e-05,
+      "loss": 0.1063,
+      "step": 9015
+    },
+    {
+      "epoch": 82.72,
+      "learning_rate": 9.076923076923078e-05,
+      "loss": 0.1923,
+      "step": 9016
+    },
+    {
+      "epoch": 82.72,
+      "learning_rate": 9.072115384615385e-05,
+      "loss": 0.1209,
+      "step": 9017
+    },
+    {
+      "epoch": 82.73,
+      "learning_rate": 9.067307692307692e-05,
+      "loss": 0.1106,
+      "step": 9018
+    },
+    {
+      "epoch": 82.74,
+      "learning_rate": 9.0625e-05,
+      "loss": 0.0696,
+      "step": 9019
+    },
+    {
+      "epoch": 82.75,
+      "learning_rate": 9.057692307692308e-05,
+      "loss": 0.1834,
+      "step": 9020
+    },
+    {
+      "epoch": 82.76,
+      "learning_rate": 9.052884615384615e-05,
+      "loss": 0.175,
+      "step": 9021
+    },
+    {
+      "epoch": 82.77,
+      "learning_rate": 9.048076923076923e-05,
+      "loss": 0.1561,
+      "step": 9022
+    },
+    {
+      "epoch": 82.78,
+      "learning_rate": 9.043269230769231e-05,
+      "loss": 0.1708,
+      "step": 9023
+    },
+    {
+      "epoch": 82.79,
+      "learning_rate": 9.038461538461538e-05,
+      "loss": 0.1665,
+      "step": 9024
+    },
+    {
+      "epoch": 82.8,
+      "learning_rate": 9.033653846153847e-05,
+      "loss": 0.1893,
+      "step": 9025
+    },
+    {
+      "epoch": 82.81,
+      "learning_rate": 9.028846153846154e-05,
+      "loss": 0.1292,
+      "step": 9026
+    },
+    {
+      "epoch": 82.82,
+      "learning_rate": 9.024038461538462e-05,
+      "loss": 0.1463,
+      "step": 9027
+    },
+    {
+      "epoch": 82.83,
+      "learning_rate": 9.01923076923077e-05,
+      "loss": 0.1797,
+      "step": 9028
+    },
+    {
+      "epoch": 82.83,
+      "learning_rate": 9.014423076923077e-05,
+      "loss": 0.1373,
+      "step": 9029
+    },
+    {
+      "epoch": 82.84,
+      "learning_rate": 9.009615384615385e-05,
+      "loss": 0.1947,
+      "step": 9030
+    },
+    {
+      "epoch": 82.85,
+      "learning_rate": 9.004807692307693e-05,
+      "loss": 0.1187,
+      "step": 9031
+    },
+    {
+      "epoch": 82.86,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 0.1239,
+      "step": 9032
+    },
+    {
+      "epoch": 82.87,
+      "learning_rate": 8.995192307692308e-05,
+      "loss": 0.1558,
+      "step": 9033
+    },
+    {
+      "epoch": 82.88,
+      "learning_rate": 8.990384615384615e-05,
+      "loss": 0.1347,
+      "step": 9034
+    },
+    {
+      "epoch": 82.89,
+      "learning_rate": 8.985576923076922e-05,
+      "loss": 0.1541,
+      "step": 9035
+    },
+    {
+      "epoch": 82.9,
+      "learning_rate": 8.980769230769231e-05,
+      "loss": 0.1844,
+      "step": 9036
+    },
+    {
+      "epoch": 82.91,
+      "learning_rate": 8.975961538461538e-05,
+      "loss": 0.1066,
+      "step": 9037
+    },
+    {
+      "epoch": 82.92,
+      "learning_rate": 8.971153846153847e-05,
+      "loss": 0.141,
+      "step": 9038
+    },
+    {
+      "epoch": 82.93,
+      "learning_rate": 8.966346153846154e-05,
+      "loss": 0.1392,
+      "step": 9039
+    },
+    {
+      "epoch": 82.94,
+      "learning_rate": 8.961538461538461e-05,
+      "loss": 0.1582,
+      "step": 9040
+    },
+    {
+      "epoch": 82.94,
+      "learning_rate": 8.95673076923077e-05,
+      "loss": 0.1439,
+      "step": 9041
+    },
+    {
+      "epoch": 82.95,
+      "learning_rate": 8.951923076923077e-05,
+      "loss": 0.1687,
+      "step": 9042
+    },
+    {
+      "epoch": 82.96,
+      "learning_rate": 8.947115384615386e-05,
+      "loss": 0.1141,
+      "step": 9043
+    },
+    {
+      "epoch": 82.97,
+      "learning_rate": 8.942307692307693e-05,
+      "loss": 0.1378,
+      "step": 9044
+    },
+    {
+      "epoch": 82.98,
+      "learning_rate": 8.9375e-05,
+      "loss": 0.112,
+      "step": 9045
+    },
+    {
+      "epoch": 82.99,
+      "learning_rate": 8.932692307692308e-05,
+      "loss": 0.0541,
+      "step": 9046
+    },
+    {
+      "epoch": 83.0,
+      "learning_rate": 8.927884615384615e-05,
+      "loss": 0.1518,
+      "step": 9047
+    },
+    {
+      "epoch": 83.01,
+      "learning_rate": 8.923076923076922e-05,
+      "loss": 0.2067,
+      "step": 9048
+    },
+    {
+      "epoch": 83.02,
+      "learning_rate": 8.918269230769231e-05,
+      "loss": 0.2098,
+      "step": 9049
+    },
+    {
+      "epoch": 83.03,
+      "learning_rate": 8.913461538461538e-05,
+      "loss": 0.1892,
+      "step": 9050
+    },
+    {
+      "epoch": 83.04,
+      "learning_rate": 8.908653846153847e-05,
+      "loss": 0.1833,
+      "step": 9051
+    },
+    {
+      "epoch": 83.05,
+      "learning_rate": 8.903846153846154e-05,
+      "loss": 0.1529,
+      "step": 9052
+    },
+    {
+      "epoch": 83.06,
+      "learning_rate": 8.899038461538461e-05,
+      "loss": 0.1365,
+      "step": 9053
+    },
+    {
+      "epoch": 83.06,
+      "learning_rate": 8.89423076923077e-05,
+      "loss": 0.1001,
+      "step": 9054
+    },
+    {
+      "epoch": 83.07,
+      "learning_rate": 8.889423076923077e-05,
+      "loss": 0.1502,
+      "step": 9055
+    },
+    {
+      "epoch": 83.08,
+      "learning_rate": 8.884615384615386e-05,
+      "loss": 0.1435,
+      "step": 9056
+    },
+    {
+      "epoch": 83.09,
+      "learning_rate": 8.879807692307693e-05,
+      "loss": 0.1372,
+      "step": 9057
+    },
+    {
+      "epoch": 83.1,
+      "learning_rate": 8.875e-05,
+      "loss": 0.1292,
+      "step": 9058
+    },
+    {
+      "epoch": 83.11,
+      "learning_rate": 8.870192307692309e-05,
+      "loss": 0.124,
+      "step": 9059
+    },
+    {
+      "epoch": 83.12,
+      "learning_rate": 8.865384615384615e-05,
+      "loss": 0.1547,
+      "step": 9060
+    },
+    {
+      "epoch": 83.13,
+      "learning_rate": 8.860576923076922e-05,
+      "loss": 0.1357,
+      "step": 9061
+    },
+    {
+      "epoch": 83.14,
+      "learning_rate": 8.85576923076923e-05,
+      "loss": 0.1867,
+      "step": 9062
+    },
+    {
+      "epoch": 83.15,
+      "learning_rate": 8.850961538461538e-05,
+      "loss": 0.209,
+      "step": 9063
+    },
+    {
+      "epoch": 83.16,
+      "learning_rate": 8.846153846153847e-05,
+      "loss": 0.2189,
+      "step": 9064
+    },
+    {
+      "epoch": 83.17,
+      "learning_rate": 8.841346153846154e-05,
+      "loss": 0.2481,
+      "step": 9065
+    },
+    {
+      "epoch": 83.17,
+      "learning_rate": 8.836538461538461e-05,
+      "loss": 0.1449,
+      "step": 9066
+    },
+    {
+      "epoch": 83.18,
+      "learning_rate": 8.83173076923077e-05,
+      "loss": 0.1686,
+      "step": 9067
+    },
+    {
+      "epoch": 83.19,
+      "learning_rate": 8.826923076923077e-05,
+      "loss": 0.1263,
+      "step": 9068
+    },
+    {
+      "epoch": 83.2,
+      "learning_rate": 8.822115384615386e-05,
+      "loss": 0.0718,
+      "step": 9069
+    },
+    {
+      "epoch": 83.21,
+      "learning_rate": 8.817307692307693e-05,
+      "loss": 0.1346,
+      "step": 9070
+    },
+    {
+      "epoch": 83.22,
+      "learning_rate": 8.8125e-05,
+      "loss": 0.146,
+      "step": 9071
+    },
+    {
+      "epoch": 83.23,
+      "learning_rate": 8.807692307692309e-05,
+      "loss": 0.1808,
+      "step": 9072
+    },
+    {
+      "epoch": 83.24,
+      "learning_rate": 8.802884615384616e-05,
+      "loss": 0.1284,
+      "step": 9073
+    },
+    {
+      "epoch": 83.25,
+      "learning_rate": 8.798076923076925e-05,
+      "loss": 0.1146,
+      "step": 9074
+    },
+    {
+      "epoch": 83.26,
+      "learning_rate": 8.79326923076923e-05,
+      "loss": 0.2614,
+      "step": 9075
+    },
+    {
+      "epoch": 83.27,
+      "learning_rate": 8.788461538461538e-05,
+      "loss": 0.1866,
+      "step": 9076
+    },
+    {
+      "epoch": 83.28,
+      "learning_rate": 8.783653846153846e-05,
+      "loss": 0.2096,
+      "step": 9077
+    },
+    {
+      "epoch": 83.28,
+      "learning_rate": 8.778846153846154e-05,
+      "loss": 0.1495,
+      "step": 9078
+    },
+    {
+      "epoch": 83.29,
+      "learning_rate": 8.774038461538461e-05,
+      "loss": 0.1853,
+      "step": 9079
+    },
+    {
+      "epoch": 83.3,
+      "learning_rate": 8.76923076923077e-05,
+      "loss": 0.16,
+      "step": 9080
+    },
+    {
+      "epoch": 83.31,
+      "learning_rate": 8.764423076923077e-05,
+      "loss": 0.1437,
+      "step": 9081
+    },
+    {
+      "epoch": 83.32,
+      "learning_rate": 8.759615384615385e-05,
+      "loss": 0.2371,
+      "step": 9082
+    },
+    {
+      "epoch": 83.33,
+      "learning_rate": 8.754807692307693e-05,
+      "loss": 0.1283,
+      "step": 9083
+    },
+    {
+      "epoch": 83.34,
+      "learning_rate": 8.75e-05,
+      "loss": 0.0895,
+      "step": 9084
+    },
+    {
+      "epoch": 83.35,
+      "learning_rate": 8.745192307692309e-05,
+      "loss": 0.1238,
+      "step": 9085
+    },
+    {
+      "epoch": 83.36,
+      "learning_rate": 8.740384615384616e-05,
+      "loss": 0.1462,
+      "step": 9086
+    },
+    {
+      "epoch": 83.37,
+      "learning_rate": 8.735576923076924e-05,
+      "loss": 0.1208,
+      "step": 9087
+    },
+    {
+      "epoch": 83.38,
+      "learning_rate": 8.730769230769232e-05,
+      "loss": 0.1521,
+      "step": 9088
+    },
+    {
+      "epoch": 83.39,
+      "learning_rate": 8.725961538461538e-05,
+      "loss": 0.1193,
+      "step": 9089
+    },
+    {
+      "epoch": 83.39,
+      "learning_rate": 8.721153846153846e-05,
+      "loss": 0.1834,
+      "step": 9090
+    },
+    {
+      "epoch": 83.4,
+      "learning_rate": 8.716346153846153e-05,
+      "loss": 0.1732,
+      "step": 9091
+    },
+    {
+      "epoch": 83.41,
+      "learning_rate": 8.711538461538462e-05,
+      "loss": 0.2351,
+      "step": 9092
+    },
+    {
+      "epoch": 83.42,
+      "learning_rate": 8.70673076923077e-05,
+      "loss": 0.1156,
+      "step": 9093
+    },
+    {
+      "epoch": 83.43,
+      "learning_rate": 8.701923076923077e-05,
+      "loss": 0.1442,
+      "step": 9094
+    },
+    {
+      "epoch": 83.44,
+      "learning_rate": 8.697115384615385e-05,
+      "loss": 0.0711,
+      "step": 9095
+    },
+    {
+      "epoch": 83.45,
+      "learning_rate": 8.692307692307692e-05,
+      "loss": 0.1759,
+      "step": 9096
+    },
+    {
+      "epoch": 83.46,
+      "learning_rate": 8.6875e-05,
+      "loss": 0.1069,
+      "step": 9097
+    },
+    {
+      "epoch": 83.47,
+      "learning_rate": 8.682692307692308e-05,
+      "loss": 0.1256,
+      "step": 9098
+    },
+    {
+      "epoch": 83.48,
+      "learning_rate": 8.677884615384616e-05,
+      "loss": 0.1227,
+      "step": 9099
+    },
+    {
+      "epoch": 83.49,
+      "learning_rate": 8.673076923076924e-05,
+      "loss": 0.108,
+      "step": 9100
+    },
+    {
+      "epoch": 83.5,
+      "learning_rate": 8.668269230769232e-05,
+      "loss": 0.1526,
+      "step": 9101
+    },
+    {
+      "epoch": 83.5,
+      "learning_rate": 8.663461538461539e-05,
+      "loss": 0.258,
+      "step": 9102
+    },
+    {
+      "epoch": 83.51,
+      "learning_rate": 8.658653846153846e-05,
+      "loss": 0.1261,
+      "step": 9103
+    },
+    {
+      "epoch": 83.52,
+      "learning_rate": 8.653846153846153e-05,
+      "loss": 0.1502,
+      "step": 9104
+    },
+    {
+      "epoch": 83.53,
+      "learning_rate": 8.649038461538462e-05,
+      "loss": 0.1455,
+      "step": 9105
+    },
+    {
+      "epoch": 83.54,
+      "learning_rate": 8.644230769230769e-05,
+      "loss": 0.2085,
+      "step": 9106
+    },
+    {
+      "epoch": 83.55,
+      "learning_rate": 8.639423076923076e-05,
+      "loss": 0.1665,
+      "step": 9107
+    },
+    {
+      "epoch": 83.56,
+      "learning_rate": 8.634615384615385e-05,
+      "loss": 0.1267,
+      "step": 9108
+    },
+    {
+      "epoch": 83.57,
+      "learning_rate": 8.629807692307692e-05,
+      "loss": 0.0905,
+      "step": 9109
+    },
+    {
+      "epoch": 83.58,
+      "learning_rate": 8.625e-05,
+      "loss": 0.1288,
+      "step": 9110
+    },
+    {
+      "epoch": 83.59,
+      "learning_rate": 8.620192307692308e-05,
+      "loss": 0.2228,
+      "step": 9111
+    },
+    {
+      "epoch": 83.6,
+      "learning_rate": 8.615384615384615e-05,
+      "loss": 0.1618,
+      "step": 9112
+    },
+    {
+      "epoch": 83.61,
+      "learning_rate": 8.610576923076924e-05,
+      "loss": 0.1353,
+      "step": 9113
+    },
+    {
+      "epoch": 83.61,
+      "learning_rate": 8.605769230769231e-05,
+      "loss": 0.1354,
+      "step": 9114
+    },
+    {
+      "epoch": 83.62,
+      "learning_rate": 8.600961538461539e-05,
+      "loss": 0.1469,
+      "step": 9115
+    },
+    {
+      "epoch": 83.63,
+      "learning_rate": 8.596153846153847e-05,
+      "loss": 0.1813,
+      "step": 9116
+    },
+    {
+      "epoch": 83.64,
+      "learning_rate": 8.591346153846153e-05,
+      "loss": 0.1663,
+      "step": 9117
+    },
+    {
+      "epoch": 83.65,
+      "learning_rate": 8.586538461538462e-05,
+      "loss": 0.2271,
+      "step": 9118
+    },
+    {
+      "epoch": 83.66,
+      "learning_rate": 8.581730769230769e-05,
+      "loss": 0.1215,
+      "step": 9119
+    },
+    {
+      "epoch": 83.67,
+      "learning_rate": 8.576923076923076e-05,
+      "loss": 0.2306,
+      "step": 9120
+    },
+    {
+      "epoch": 83.68,
+      "learning_rate": 8.572115384615385e-05,
+      "loss": 0.1724,
+      "step": 9121
+    },
+    {
+      "epoch": 83.69,
+      "learning_rate": 8.567307692307692e-05,
+      "loss": 0.2067,
+      "step": 9122
+    },
+    {
+      "epoch": 83.7,
+      "learning_rate": 8.562500000000001e-05,
+      "loss": 0.1113,
+      "step": 9123
+    },
+    {
+      "epoch": 83.71,
+      "learning_rate": 8.557692307692308e-05,
+      "loss": 0.1394,
+      "step": 9124
+    },
+    {
+      "epoch": 83.72,
+      "learning_rate": 8.552884615384615e-05,
+      "loss": 0.1637,
+      "step": 9125
+    },
+    {
+      "epoch": 83.72,
+      "learning_rate": 8.548076923076924e-05,
+      "loss": 0.2263,
+      "step": 9126
+    },
+    {
+      "epoch": 83.73,
+      "learning_rate": 8.543269230769231e-05,
+      "loss": 0.1086,
+      "step": 9127
+    },
+    {
+      "epoch": 83.74,
+      "learning_rate": 8.538461538461538e-05,
+      "loss": 0.0853,
+      "step": 9128
+    },
+    {
+      "epoch": 83.75,
+      "learning_rate": 8.533653846153847e-05,
+      "loss": 0.2066,
+      "step": 9129
+    },
+    {
+      "epoch": 83.76,
+      "learning_rate": 8.528846153846154e-05,
+      "loss": 0.1847,
+      "step": 9130
+    },
+    {
+      "epoch": 83.77,
+      "learning_rate": 8.524038461538462e-05,
+      "loss": 0.1446,
+      "step": 9131
+    },
+    {
+      "epoch": 83.78,
+      "learning_rate": 8.519230769230769e-05,
+      "loss": 0.1507,
+      "step": 9132
+    },
+    {
+      "epoch": 83.79,
+      "learning_rate": 8.514423076923076e-05,
+      "loss": 0.2219,
+      "step": 9133
+    },
+    {
+      "epoch": 83.8,
+      "learning_rate": 8.509615384615385e-05,
+      "loss": 0.1408,
+      "step": 9134
+    },
+    {
+      "epoch": 83.81,
+      "learning_rate": 8.504807692307692e-05,
+      "loss": 0.1606,
+      "step": 9135
+    },
+    {
+      "epoch": 83.82,
+      "learning_rate": 8.5e-05,
+      "loss": 0.1151,
+      "step": 9136
+    },
+    {
+      "epoch": 83.83,
+      "learning_rate": 8.495192307692308e-05,
+      "loss": 0.1845,
+      "step": 9137
+    },
+    {
+      "epoch": 83.83,
+      "learning_rate": 8.490384615384615e-05,
+      "loss": 0.1579,
+      "step": 9138
+    },
+    {
+      "epoch": 83.84,
+      "learning_rate": 8.485576923076924e-05,
+      "loss": 0.0981,
+      "step": 9139
+    },
+    {
+      "epoch": 83.85,
+      "learning_rate": 8.480769230769231e-05,
+      "loss": 0.1744,
+      "step": 9140
+    },
+    {
+      "epoch": 83.86,
+      "learning_rate": 8.475961538461538e-05,
+      "loss": 0.1429,
+      "step": 9141
+    },
+    {
+      "epoch": 83.87,
+      "learning_rate": 8.471153846153847e-05,
+      "loss": 0.138,
+      "step": 9142
+    },
+    {
+      "epoch": 83.88,
+      "learning_rate": 8.466346153846154e-05,
+      "loss": 0.1523,
+      "step": 9143
+    },
+    {
+      "epoch": 83.89,
+      "learning_rate": 8.461538461538463e-05,
+      "loss": 0.163,
+      "step": 9144
+    },
+    {
+      "epoch": 83.9,
+      "learning_rate": 8.456730769230769e-05,
+      "loss": 0.1205,
+      "step": 9145
+    },
+    {
+      "epoch": 83.91,
+      "learning_rate": 8.451923076923076e-05,
+      "loss": 0.0902,
+      "step": 9146
+    },
+    {
+      "epoch": 83.92,
+      "learning_rate": 8.447115384615385e-05,
+      "loss": 0.1273,
+      "step": 9147
+    },
+    {
+      "epoch": 83.93,
+      "learning_rate": 8.442307692307692e-05,
+      "loss": 0.183,
+      "step": 9148
+    },
+    {
+      "epoch": 83.94,
+      "learning_rate": 8.4375e-05,
+      "loss": 0.1771,
+      "step": 9149
+    },
+    {
+      "epoch": 83.94,
+      "learning_rate": 8.432692307692308e-05,
+      "loss": 0.1551,
+      "step": 9150
+    },
+    {
+      "epoch": 83.95,
+      "learning_rate": 8.427884615384615e-05,
+      "loss": 0.2116,
+      "step": 9151
+    },
+    {
+      "epoch": 83.96,
+      "learning_rate": 8.423076923076924e-05,
+      "loss": 0.1206,
+      "step": 9152
+    },
+    {
+      "epoch": 83.97,
+      "learning_rate": 8.418269230769231e-05,
+      "loss": 0.1717,
+      "step": 9153
+    },
+    {
+      "epoch": 83.98,
+      "learning_rate": 8.41346153846154e-05,
+      "loss": 0.2124,
+      "step": 9154
+    },
+    {
+      "epoch": 83.99,
+      "learning_rate": 8.408653846153847e-05,
+      "loss": 0.1112,
+      "step": 9155
+    },
+    {
+      "epoch": 84.0,
+      "learning_rate": 8.403846153846154e-05,
+      "loss": 0.1664,
+      "step": 9156
+    },
+    {
+      "epoch": 84.01,
+      "learning_rate": 8.399038461538463e-05,
+      "loss": 0.199,
+      "step": 9157
+    },
+    {
+      "epoch": 84.02,
+      "learning_rate": 8.39423076923077e-05,
+      "loss": 0.1856,
+      "step": 9158
+    },
+    {
+      "epoch": 84.03,
+      "learning_rate": 8.389423076923076e-05,
+      "loss": 0.1455,
+      "step": 9159
+    },
+    {
+      "epoch": 84.04,
+      "learning_rate": 8.384615384615384e-05,
+      "loss": 0.1465,
+      "step": 9160
+    },
+    {
+      "epoch": 84.05,
+      "learning_rate": 8.379807692307692e-05,
+      "loss": 0.1524,
+      "step": 9161
+    },
+    {
+      "epoch": 84.06,
+      "learning_rate": 8.375e-05,
+      "loss": 0.1535,
+      "step": 9162
+    },
+    {
+      "epoch": 84.06,
+      "learning_rate": 8.370192307692308e-05,
+      "loss": 0.1345,
+      "step": 9163
+    },
+    {
+      "epoch": 84.07,
+      "learning_rate": 8.365384615384615e-05,
+      "loss": 0.1337,
+      "step": 9164
+    },
+    {
+      "epoch": 84.08,
+      "learning_rate": 8.360576923076923e-05,
+      "loss": 0.1204,
+      "step": 9165
+    },
+    {
+      "epoch": 84.09,
+      "learning_rate": 8.355769230769231e-05,
+      "loss": 0.1178,
+      "step": 9166
+    },
+    {
+      "epoch": 84.1,
+      "learning_rate": 8.350961538461539e-05,
+      "loss": 0.2097,
+      "step": 9167
+    },
+    {
+      "epoch": 84.11,
+      "learning_rate": 8.346153846153847e-05,
+      "loss": 0.1811,
+      "step": 9168
+    },
+    {
+      "epoch": 84.12,
+      "learning_rate": 8.341346153846154e-05,
+      "loss": 0.1924,
+      "step": 9169
+    },
+    {
+      "epoch": 84.13,
+      "learning_rate": 8.336538461538462e-05,
+      "loss": 0.1827,
+      "step": 9170
+    },
+    {
+      "epoch": 84.14,
+      "learning_rate": 8.33173076923077e-05,
+      "loss": 0.1013,
+      "step": 9171
+    },
+    {
+      "epoch": 84.15,
+      "learning_rate": 8.326923076923078e-05,
+      "loss": 0.1771,
+      "step": 9172
+    },
+    {
+      "epoch": 84.16,
+      "learning_rate": 8.322115384615386e-05,
+      "loss": 0.1763,
+      "step": 9173
+    },
+    {
+      "epoch": 84.17,
+      "learning_rate": 8.317307692307692e-05,
+      "loss": 0.1146,
+      "step": 9174
+    },
+    {
+      "epoch": 84.17,
+      "learning_rate": 8.3125e-05,
+      "loss": 0.144,
+      "step": 9175
+    },
+    {
+      "epoch": 84.18,
+      "learning_rate": 8.307692307692307e-05,
+      "loss": 0.1288,
+      "step": 9176
+    },
+    {
+      "epoch": 84.19,
+      "learning_rate": 8.302884615384615e-05,
+      "loss": 0.123,
+      "step": 9177
+    },
+    {
+      "epoch": 84.2,
+      "learning_rate": 8.298076923076923e-05,
+      "loss": 0.1166,
+      "step": 9178
+    },
+    {
+      "epoch": 84.21,
+      "learning_rate": 8.29326923076923e-05,
+      "loss": 0.0903,
+      "step": 9179
+    },
+    {
+      "epoch": 84.22,
+      "learning_rate": 8.288461538461539e-05,
+      "loss": 0.0948,
+      "step": 9180
+    },
+    {
+      "epoch": 84.23,
+      "learning_rate": 8.283653846153846e-05,
+      "loss": 0.1042,
+      "step": 9181
+    },
+    {
+      "epoch": 84.24,
+      "learning_rate": 8.278846153846154e-05,
+      "loss": 0.1363,
+      "step": 9182
+    },
+    {
+      "epoch": 84.25,
+      "learning_rate": 8.274038461538462e-05,
+      "loss": 0.1263,
+      "step": 9183
+    },
+    {
+      "epoch": 84.26,
+      "learning_rate": 8.26923076923077e-05,
+      "loss": 0.1698,
+      "step": 9184
+    },
+    {
+      "epoch": 84.27,
+      "learning_rate": 8.264423076923078e-05,
+      "loss": 0.177,
+      "step": 9185
+    },
+    {
+      "epoch": 84.28,
+      "learning_rate": 8.259615384615385e-05,
+      "loss": 0.2084,
+      "step": 9186
+    },
+    {
+      "epoch": 84.28,
+      "learning_rate": 8.254807692307693e-05,
+      "loss": 0.1276,
+      "step": 9187
+    },
+    {
+      "epoch": 84.29,
+      "learning_rate": 8.25e-05,
+      "loss": 0.1697,
+      "step": 9188
+    },
+    {
+      "epoch": 84.3,
+      "learning_rate": 8.245192307692307e-05,
+      "loss": 0.1682,
+      "step": 9189
+    },
+    {
+      "epoch": 84.31,
+      "learning_rate": 8.240384615384614e-05,
+      "loss": 0.2074,
+      "step": 9190
+    },
+    {
+      "epoch": 84.32,
+      "learning_rate": 8.235576923076923e-05,
+      "loss": 0.1154,
+      "step": 9191
+    },
+    {
+      "epoch": 84.33,
+      "learning_rate": 8.23076923076923e-05,
+      "loss": 0.1618,
+      "step": 9192
+    },
+    {
+      "epoch": 84.34,
+      "learning_rate": 8.225961538461539e-05,
+      "loss": 0.1707,
+      "step": 9193
+    },
+    {
+      "epoch": 84.35,
+      "learning_rate": 8.221153846153846e-05,
+      "loss": 0.2293,
+      "step": 9194
+    },
+    {
+      "epoch": 84.36,
+      "learning_rate": 8.216346153846154e-05,
+      "loss": 0.1229,
+      "step": 9195
+    },
+    {
+      "epoch": 84.37,
+      "learning_rate": 8.211538461538462e-05,
+      "loss": 0.1695,
+      "step": 9196
+    },
+    {
+      "epoch": 84.38,
+      "learning_rate": 8.20673076923077e-05,
+      "loss": 0.1938,
+      "step": 9197
+    },
+    {
+      "epoch": 84.39,
+      "learning_rate": 8.201923076923078e-05,
+      "loss": 0.1925,
+      "step": 9198
+    },
+    {
+      "epoch": 84.39,
+      "learning_rate": 8.197115384615385e-05,
+      "loss": 0.1246,
+      "step": 9199
+    },
+    {
+      "epoch": 84.4,
+      "learning_rate": 8.192307692307693e-05,
+      "loss": 0.1722,
+      "step": 9200
+    },
+    {
+      "epoch": 84.41,
+      "learning_rate": 8.187500000000001e-05,
+      "loss": 0.0917,
+      "step": 9201
+    },
+    {
+      "epoch": 84.42,
+      "learning_rate": 8.182692307692307e-05,
+      "loss": 0.1673,
+      "step": 9202
+    },
+    {
+      "epoch": 84.43,
+      "learning_rate": 8.177884615384616e-05,
+      "loss": 0.1783,
+      "step": 9203
+    },
+    {
+      "epoch": 84.44,
+      "learning_rate": 8.173076923076923e-05,
+      "loss": 0.1931,
+      "step": 9204
+    },
+    {
+      "epoch": 84.45,
+      "learning_rate": 8.16826923076923e-05,
+      "loss": 0.2035,
+      "step": 9205
+    },
+    {
+      "epoch": 84.46,
+      "learning_rate": 8.163461538461539e-05,
+      "loss": 0.2259,
+      "step": 9206
+    },
+    {
+      "epoch": 84.47,
+      "learning_rate": 8.158653846153846e-05,
+      "loss": 0.1957,
+      "step": 9207
+    },
+    {
+      "epoch": 84.48,
+      "learning_rate": 8.153846153846153e-05,
+      "loss": 0.1858,
+      "step": 9208
+    },
+    {
+      "epoch": 84.49,
+      "learning_rate": 8.149038461538462e-05,
+      "loss": 0.0791,
+      "step": 9209
+    },
+    {
+      "epoch": 84.5,
+      "learning_rate": 8.144230769230769e-05,
+      "loss": 0.1363,
+      "step": 9210
+    },
+    {
+      "epoch": 84.5,
+      "learning_rate": 8.139423076923078e-05,
+      "loss": 0.2191,
+      "step": 9211
+    },
+    {
+      "epoch": 84.51,
+      "learning_rate": 8.134615384615385e-05,
+      "loss": 0.1547,
+      "step": 9212
+    },
+    {
+      "epoch": 84.52,
+      "learning_rate": 8.129807692307692e-05,
+      "loss": 0.1654,
+      "step": 9213
+    },
+    {
+      "epoch": 84.53,
+      "learning_rate": 8.125000000000001e-05,
+      "loss": 0.148,
+      "step": 9214
+    },
+    {
+      "epoch": 84.54,
+      "learning_rate": 8.120192307692308e-05,
+      "loss": 0.1654,
+      "step": 9215
+    },
+    {
+      "epoch": 84.55,
+      "learning_rate": 8.115384615384616e-05,
+      "loss": 0.1649,
+      "step": 9216
+    },
+    {
+      "epoch": 84.56,
+      "learning_rate": 8.110576923076923e-05,
+      "loss": 0.1616,
+      "step": 9217
+    },
+    {
+      "epoch": 84.57,
+      "learning_rate": 8.10576923076923e-05,
+      "loss": 0.1401,
+      "step": 9218
+    },
+    {
+      "epoch": 84.58,
+      "learning_rate": 8.100961538461539e-05,
+      "loss": 0.1063,
+      "step": 9219
+    },
+    {
+      "epoch": 84.59,
+      "learning_rate": 8.096153846153846e-05,
+      "loss": 0.1932,
+      "step": 9220
+    },
+    {
+      "epoch": 84.6,
+      "learning_rate": 8.091346153846155e-05,
+      "loss": 0.1853,
+      "step": 9221
+    },
+    {
+      "epoch": 84.61,
+      "learning_rate": 8.086538461538462e-05,
+      "loss": 0.1919,
+      "step": 9222
+    },
+    {
+      "epoch": 84.61,
+      "learning_rate": 8.081730769230769e-05,
+      "loss": 0.125,
+      "step": 9223
+    },
+    {
+      "epoch": 84.62,
+      "learning_rate": 8.076923076923078e-05,
+      "loss": 0.1274,
+      "step": 9224
+    },
+    {
+      "epoch": 84.63,
+      "learning_rate": 8.072115384615385e-05,
+      "loss": 0.1577,
+      "step": 9225
+    },
+    {
+      "epoch": 84.64,
+      "learning_rate": 8.067307692307692e-05,
+      "loss": 0.1621,
+      "step": 9226
+    },
+    {
+      "epoch": 84.65,
+      "learning_rate": 8.062500000000001e-05,
+      "loss": 0.1379,
+      "step": 9227
+    },
+    {
+      "epoch": 84.66,
+      "learning_rate": 8.057692307692308e-05,
+      "loss": 0.1286,
+      "step": 9228
+    },
+    {
+      "epoch": 84.67,
+      "learning_rate": 8.052884615384617e-05,
+      "loss": 0.145,
+      "step": 9229
+    },
+    {
+      "epoch": 84.68,
+      "learning_rate": 8.048076923076923e-05,
+      "loss": 0.1559,
+      "step": 9230
+    },
+    {
+      "epoch": 84.69,
+      "learning_rate": 8.04326923076923e-05,
+      "loss": 0.1382,
+      "step": 9231
+    },
+    {
+      "epoch": 84.7,
+      "learning_rate": 8.038461538461538e-05,
+      "loss": 0.1175,
+      "step": 9232
+    },
+    {
+      "epoch": 84.71,
+      "learning_rate": 8.033653846153846e-05,
+      "loss": 0.0949,
+      "step": 9233
+    },
+    {
+      "epoch": 84.72,
+      "learning_rate": 8.028846153846154e-05,
+      "loss": 0.106,
+      "step": 9234
+    },
+    {
+      "epoch": 84.72,
+      "learning_rate": 8.024038461538462e-05,
+      "loss": 0.2506,
+      "step": 9235
+    },
+    {
+      "epoch": 84.73,
+      "learning_rate": 8.019230769230769e-05,
+      "loss": 0.1278,
+      "step": 9236
+    },
+    {
+      "epoch": 84.74,
+      "learning_rate": 8.014423076923078e-05,
+      "loss": 0.1495,
+      "step": 9237
+    },
+    {
+      "epoch": 84.75,
+      "learning_rate": 8.009615384615385e-05,
+      "loss": 0.173,
+      "step": 9238
+    },
+    {
+      "epoch": 84.76,
+      "learning_rate": 8.004807692307692e-05,
+      "loss": 0.1266,
+      "step": 9239
+    },
+    {
+      "epoch": 84.77,
+      "learning_rate": 8e-05,
+      "loss": 0.1752,
+      "step": 9240
+    },
+    {
+      "epoch": 84.78,
+      "learning_rate": 7.995192307692308e-05,
+      "loss": 0.1961,
+      "step": 9241
+    },
+    {
+      "epoch": 84.79,
+      "learning_rate": 7.990384615384617e-05,
+      "loss": 0.2241,
+      "step": 9242
+    },
+    {
+      "epoch": 84.8,
+      "learning_rate": 7.985576923076924e-05,
+      "loss": 0.167,
+      "step": 9243
+    },
+    {
+      "epoch": 84.81,
+      "learning_rate": 7.98076923076923e-05,
+      "loss": 0.141,
+      "step": 9244
+    },
+    {
+      "epoch": 84.82,
+      "learning_rate": 7.975961538461538e-05,
+      "loss": 0.0856,
+      "step": 9245
+    },
+    {
+      "epoch": 84.83,
+      "learning_rate": 7.971153846153846e-05,
+      "loss": 0.1087,
+      "step": 9246
+    },
+    {
+      "epoch": 84.83,
+      "learning_rate": 7.966346153846154e-05,
+      "loss": 0.2209,
+      "step": 9247
+    },
+    {
+      "epoch": 84.84,
+      "learning_rate": 7.961538461538461e-05,
+      "loss": 0.1841,
+      "step": 9248
+    },
+    {
+      "epoch": 84.85,
+      "learning_rate": 7.956730769230769e-05,
+      "loss": 0.1508,
+      "step": 9249
+    },
+    {
+      "epoch": 84.86,
+      "learning_rate": 7.951923076923077e-05,
+      "loss": 0.1364,
+      "step": 9250
+    },
+    {
+      "epoch": 84.87,
+      "learning_rate": 7.947115384615385e-05,
+      "loss": 0.1669,
+      "step": 9251
+    },
+    {
+      "epoch": 84.88,
+      "learning_rate": 7.942307692307693e-05,
+      "loss": 0.1532,
+      "step": 9252
+    },
+    {
+      "epoch": 84.89,
+      "learning_rate": 7.9375e-05,
+      "loss": 0.1869,
+      "step": 9253
+    },
+    {
+      "epoch": 84.9,
+      "learning_rate": 7.932692307692308e-05,
+      "loss": 0.1043,
+      "step": 9254
+    },
+    {
+      "epoch": 84.91,
+      "learning_rate": 7.927884615384616e-05,
+      "loss": 0.0989,
+      "step": 9255
+    },
+    {
+      "epoch": 84.92,
+      "learning_rate": 7.923076923076924e-05,
+      "loss": 0.0916,
+      "step": 9256
+    },
+    {
+      "epoch": 84.93,
+      "learning_rate": 7.918269230769231e-05,
+      "loss": 0.1552,
+      "step": 9257
+    },
+    {
+      "epoch": 84.94,
+      "learning_rate": 7.913461538461538e-05,
+      "loss": 0.1399,
+      "step": 9258
+    },
+    {
+      "epoch": 84.94,
+      "learning_rate": 7.908653846153845e-05,
+      "loss": 0.1748,
+      "step": 9259
+    },
+    {
+      "epoch": 84.95,
+      "learning_rate": 7.903846153846154e-05,
+      "loss": 0.1362,
+      "step": 9260
+    },
+    {
+      "epoch": 84.96,
+      "learning_rate": 7.899038461538461e-05,
+      "loss": 0.0911,
+      "step": 9261
+    },
+    {
+      "epoch": 84.97,
+      "learning_rate": 7.894230769230769e-05,
+      "loss": 0.0819,
+      "step": 9262
+    },
+    {
+      "epoch": 84.98,
+      "learning_rate": 7.889423076923077e-05,
+      "loss": 0.1332,
+      "step": 9263
+    },
+    {
+      "epoch": 84.99,
+      "learning_rate": 7.884615384615384e-05,
+      "loss": 0.0796,
+      "step": 9264
+    },
+    {
+      "epoch": 85.0,
+      "learning_rate": 7.879807692307693e-05,
+      "loss": 0.1213,
+      "step": 9265
+    },
+    {
+      "epoch": 85.01,
+      "learning_rate": 7.875e-05,
+      "loss": 0.2334,
+      "step": 9266
+    },
+    {
+      "epoch": 85.02,
+      "learning_rate": 7.870192307692308e-05,
+      "loss": 0.1957,
+      "step": 9267
+    },
+    {
+      "epoch": 85.03,
+      "learning_rate": 7.865384615384616e-05,
+      "loss": 0.1461,
+      "step": 9268
+    },
+    {
+      "epoch": 85.04,
+      "learning_rate": 7.860576923076923e-05,
+      "loss": 0.144,
+      "step": 9269
+    },
+    {
+      "epoch": 85.05,
+      "learning_rate": 7.855769230769232e-05,
+      "loss": 0.1397,
+      "step": 9270
+    },
+    {
+      "epoch": 85.06,
+      "learning_rate": 7.85096153846154e-05,
+      "loss": 0.2071,
+      "step": 9271
+    },
+    {
+      "epoch": 85.06,
+      "learning_rate": 7.846153846153845e-05,
+      "loss": 0.1428,
+      "step": 9272
+    },
+    {
+      "epoch": 85.07,
+      "learning_rate": 7.841346153846154e-05,
+      "loss": 0.1217,
+      "step": 9273
+    },
+    {
+      "epoch": 85.08,
+      "learning_rate": 7.836538461538461e-05,
+      "loss": 0.1312,
+      "step": 9274
+    },
+    {
+      "epoch": 85.09,
+      "learning_rate": 7.831730769230768e-05,
+      "loss": 0.1433,
+      "step": 9275
+    },
+    {
+      "epoch": 85.1,
+      "learning_rate": 7.826923076923077e-05,
+      "loss": 0.1639,
+      "step": 9276
+    },
+    {
+      "epoch": 85.11,
+      "learning_rate": 7.822115384615384e-05,
+      "loss": 0.1626,
+      "step": 9277
+    },
+    {
+      "epoch": 85.12,
+      "learning_rate": 7.817307692307693e-05,
+      "loss": 0.1798,
+      "step": 9278
+    },
+    {
+      "epoch": 85.13,
+      "learning_rate": 7.8125e-05,
+      "loss": 0.1301,
+      "step": 9279
+    },
+    {
+      "epoch": 85.14,
+      "learning_rate": 7.807692307692307e-05,
+      "loss": 0.1047,
+      "step": 9280
+    },
+    {
+      "epoch": 85.15,
+      "learning_rate": 7.802884615384616e-05,
+      "loss": 0.121,
+      "step": 9281
+    },
+    {
+      "epoch": 85.16,
+      "learning_rate": 7.798076923076923e-05,
+      "loss": 0.1215,
+      "step": 9282
+    },
+    {
+      "epoch": 85.17,
+      "learning_rate": 7.793269230769232e-05,
+      "loss": 0.1321,
+      "step": 9283
+    },
+    {
+      "epoch": 85.17,
+      "learning_rate": 7.788461538461539e-05,
+      "loss": 0.0994,
+      "step": 9284
+    },
+    {
+      "epoch": 85.18,
+      "learning_rate": 7.783653846153846e-05,
+      "loss": 0.1201,
+      "step": 9285
+    },
+    {
+      "epoch": 85.19,
+      "learning_rate": 7.778846153846155e-05,
+      "loss": 0.1249,
+      "step": 9286
+    },
+    {
+      "epoch": 85.2,
+      "learning_rate": 7.774038461538461e-05,
+      "loss": 0.1323,
+      "step": 9287
+    },
+    {
+      "epoch": 85.21,
+      "learning_rate": 7.769230769230768e-05,
+      "loss": 0.1447,
+      "step": 9288
+    },
+    {
+      "epoch": 85.22,
+      "learning_rate": 7.764423076923077e-05,
+      "loss": 0.1441,
+      "step": 9289
+    },
+    {
+      "epoch": 85.23,
+      "learning_rate": 7.759615384615384e-05,
+      "loss": 0.1069,
+      "step": 9290
+    },
+    {
+      "epoch": 85.24,
+      "learning_rate": 7.754807692307693e-05,
+      "loss": 0.0919,
+      "step": 9291
+    },
+    {
+      "epoch": 85.25,
+      "learning_rate": 7.75e-05,
+      "loss": 0.0987,
+      "step": 9292
+    },
+    {
+      "epoch": 85.26,
+      "learning_rate": 7.745192307692307e-05,
+      "loss": 0.2999,
+      "step": 9293
+    },
+    {
+      "epoch": 85.27,
+      "learning_rate": 7.740384615384616e-05,
+      "loss": 0.1462,
+      "step": 9294
+    },
+    {
+      "epoch": 85.28,
+      "learning_rate": 7.735576923076923e-05,
+      "loss": 0.1638,
+      "step": 9295
+    },
+    {
+      "epoch": 85.28,
+      "learning_rate": 7.730769230769232e-05,
+      "loss": 0.1576,
+      "step": 9296
+    },
+    {
+      "epoch": 85.29,
+      "learning_rate": 7.725961538461539e-05,
+      "loss": 0.1212,
+      "step": 9297
+    },
+    {
+      "epoch": 85.3,
+      "learning_rate": 7.721153846153846e-05,
+      "loss": 0.1395,
+      "step": 9298
+    },
+    {
+      "epoch": 85.31,
+      "learning_rate": 7.716346153846155e-05,
+      "loss": 0.1754,
+      "step": 9299
+    },
+    {
+      "epoch": 85.32,
+      "learning_rate": 7.711538461538462e-05,
+      "loss": 0.1137,
+      "step": 9300
+    },
+    {
+      "epoch": 85.33,
+      "learning_rate": 7.70673076923077e-05,
+      "loss": 0.0955,
+      "step": 9301
+    },
+    {
+      "epoch": 85.34,
+      "learning_rate": 7.701923076923077e-05,
+      "loss": 0.1341,
+      "step": 9302
+    },
+    {
+      "epoch": 85.35,
+      "learning_rate": 7.697115384615384e-05,
+      "loss": 0.154,
+      "step": 9303
+    },
+    {
+      "epoch": 85.36,
+      "learning_rate": 7.692307692307693e-05,
+      "loss": 0.1513,
+      "step": 9304
+    },
+    {
+      "epoch": 85.37,
+      "learning_rate": 7.6875e-05,
+      "loss": 0.1012,
+      "step": 9305
+    },
+    {
+      "epoch": 85.38,
+      "learning_rate": 7.682692307692307e-05,
+      "loss": 0.1706,
+      "step": 9306
+    },
+    {
+      "epoch": 85.39,
+      "learning_rate": 7.677884615384616e-05,
+      "loss": 0.1861,
+      "step": 9307
+    },
+    {
+      "epoch": 85.39,
+      "learning_rate": 7.673076923076923e-05,
+      "loss": 0.15,
+      "step": 9308
+    },
+    {
+      "epoch": 85.4,
+      "learning_rate": 7.668269230769232e-05,
+      "loss": 0.0647,
+      "step": 9309
+    },
+    {
+      "epoch": 85.41,
+      "learning_rate": 7.663461538461539e-05,
+      "loss": 0.1116,
+      "step": 9310
+    },
+    {
+      "epoch": 85.42,
+      "learning_rate": 7.658653846153846e-05,
+      "loss": 0.1701,
+      "step": 9311
+    },
+    {
+      "epoch": 85.43,
+      "learning_rate": 7.653846153846155e-05,
+      "loss": 0.1461,
+      "step": 9312
+    },
+    {
+      "epoch": 85.44,
+      "learning_rate": 7.649038461538462e-05,
+      "loss": 0.1529,
+      "step": 9313
+    },
+    {
+      "epoch": 85.45,
+      "learning_rate": 7.64423076923077e-05,
+      "loss": 0.1362,
+      "step": 9314
+    },
+    {
+      "epoch": 85.46,
+      "learning_rate": 7.639423076923077e-05,
+      "loss": 0.2455,
+      "step": 9315
+    },
+    {
+      "epoch": 85.47,
+      "learning_rate": 7.634615384615384e-05,
+      "loss": 0.1365,
+      "step": 9316
+    },
+    {
+      "epoch": 85.48,
+      "learning_rate": 7.629807692307692e-05,
+      "loss": 0.0964,
+      "step": 9317
+    },
+    {
+      "epoch": 85.49,
+      "learning_rate": 7.625e-05,
+      "loss": 0.0862,
+      "step": 9318
+    },
+    {
+      "epoch": 85.5,
+      "learning_rate": 7.620192307692308e-05,
+      "loss": 0.0817,
+      "step": 9319
+    },
+    {
+      "epoch": 85.5,
+      "learning_rate": 7.615384615384616e-05,
+      "loss": 0.2189,
+      "step": 9320
+    },
+    {
+      "epoch": 85.51,
+      "learning_rate": 7.610576923076923e-05,
+      "loss": 0.1551,
+      "step": 9321
+    },
+    {
+      "epoch": 85.52,
+      "learning_rate": 7.605769230769231e-05,
+      "loss": 0.194,
+      "step": 9322
+    },
+    {
+      "epoch": 85.53,
+      "learning_rate": 7.600961538461539e-05,
+      "loss": 0.1547,
+      "step": 9323
+    },
+    {
+      "epoch": 85.54,
+      "learning_rate": 7.596153846153846e-05,
+      "loss": 0.1556,
+      "step": 9324
+    },
+    {
+      "epoch": 85.55,
+      "learning_rate": 7.591346153846155e-05,
+      "loss": 0.1529,
+      "step": 9325
+    },
+    {
+      "epoch": 85.56,
+      "learning_rate": 7.586538461538462e-05,
+      "loss": 0.1683,
+      "step": 9326
+    },
+    {
+      "epoch": 85.57,
+      "learning_rate": 7.58173076923077e-05,
+      "loss": 0.1833,
+      "step": 9327
+    },
+    {
+      "epoch": 85.58,
+      "learning_rate": 7.576923076923078e-05,
+      "loss": 0.1286,
+      "step": 9328
+    },
+    {
+      "epoch": 85.59,
+      "learning_rate": 7.572115384615384e-05,
+      "loss": 0.1196,
+      "step": 9329
+    },
+    {
+      "epoch": 85.6,
+      "learning_rate": 7.567307692307692e-05,
+      "loss": 0.2412,
+      "step": 9330
+    },
+    {
+      "epoch": 85.61,
+      "learning_rate": 7.5625e-05,
+      "loss": 0.1774,
+      "step": 9331
+    },
+    {
+      "epoch": 85.61,
+      "learning_rate": 7.557692307692308e-05,
+      "loss": 0.1266,
+      "step": 9332
+    },
+    {
+      "epoch": 85.62,
+      "learning_rate": 7.552884615384615e-05,
+      "loss": 0.1289,
+      "step": 9333
+    },
+    {
+      "epoch": 85.63,
+      "learning_rate": 7.548076923076923e-05,
+      "loss": 0.1557,
+      "step": 9334
+    },
+    {
+      "epoch": 85.64,
+      "learning_rate": 7.543269230769231e-05,
+      "loss": 0.2274,
+      "step": 9335
+    },
+    {
+      "epoch": 85.65,
+      "learning_rate": 7.538461538461539e-05,
+      "loss": 0.1473,
+      "step": 9336
+    },
+    {
+      "epoch": 85.66,
+      "learning_rate": 7.533653846153846e-05,
+      "loss": 0.0956,
+      "step": 9337
+    },
+    {
+      "epoch": 85.67,
+      "learning_rate": 7.528846153846154e-05,
+      "loss": 0.1052,
+      "step": 9338
+    },
+    {
+      "epoch": 85.68,
+      "learning_rate": 7.524038461538462e-05,
+      "loss": 0.1142,
+      "step": 9339
+    },
+    {
+      "epoch": 85.69,
+      "learning_rate": 7.51923076923077e-05,
+      "loss": 0.1302,
+      "step": 9340
+    },
+    {
+      "epoch": 85.7,
+      "learning_rate": 7.514423076923078e-05,
+      "loss": 0.1671,
+      "step": 9341
+    },
+    {
+      "epoch": 85.71,
+      "learning_rate": 7.509615384615385e-05,
+      "loss": 0.0898,
+      "step": 9342
+    },
+    {
+      "epoch": 85.72,
+      "learning_rate": 7.504807692307692e-05,
+      "loss": 0.1501,
+      "step": 9343
+    },
+    {
+      "epoch": 85.72,
+      "learning_rate": 7.5e-05,
+      "loss": 0.087,
+      "step": 9344
+    },
+    {
+      "epoch": 85.73,
+      "learning_rate": 7.495192307692308e-05,
+      "loss": 0.1777,
+      "step": 9345
+    },
+    {
+      "epoch": 85.74,
+      "learning_rate": 7.490384615384615e-05,
+      "loss": 0.1025,
+      "step": 9346
+    },
+    {
+      "epoch": 85.75,
+      "learning_rate": 7.485576923076922e-05,
+      "loss": 0.2567,
+      "step": 9347
+    },
+    {
+      "epoch": 85.76,
+      "learning_rate": 7.480769230769231e-05,
+      "loss": 0.2158,
+      "step": 9348
+    },
+    {
+      "epoch": 85.77,
+      "learning_rate": 7.475961538461538e-05,
+      "loss": 0.1935,
+      "step": 9349
+    },
+    {
+      "epoch": 85.78,
+      "learning_rate": 7.471153846153847e-05,
+      "loss": 0.1509,
+      "step": 9350
+    },
+    {
+      "epoch": 85.79,
+      "learning_rate": 7.466346153846154e-05,
+      "loss": 0.1426,
+      "step": 9351
+    },
+    {
+      "epoch": 85.8,
+      "learning_rate": 7.461538461538462e-05,
+      "loss": 0.1838,
+      "step": 9352
+    },
+    {
+      "epoch": 85.81,
+      "learning_rate": 7.45673076923077e-05,
+      "loss": 0.1188,
+      "step": 9353
+    },
+    {
+      "epoch": 85.82,
+      "learning_rate": 7.451923076923077e-05,
+      "loss": 0.1139,
+      "step": 9354
+    },
+    {
+      "epoch": 85.83,
+      "learning_rate": 7.447115384615385e-05,
+      "loss": 0.181,
+      "step": 9355
+    },
+    {
+      "epoch": 85.83,
+      "learning_rate": 7.442307692307693e-05,
+      "loss": 0.1155,
+      "step": 9356
+    },
+    {
+      "epoch": 85.84,
+      "learning_rate": 7.437499999999999e-05,
+      "loss": 0.1205,
+      "step": 9357
+    },
+    {
+      "epoch": 85.85,
+      "learning_rate": 7.432692307692308e-05,
+      "loss": 0.2628,
+      "step": 9358
+    },
+    {
+      "epoch": 85.86,
+      "learning_rate": 7.427884615384615e-05,
+      "loss": 0.1377,
+      "step": 9359
+    },
+    {
+      "epoch": 85.87,
+      "learning_rate": 7.423076923076922e-05,
+      "loss": 0.0976,
+      "step": 9360
+    },
+    {
+      "epoch": 85.88,
+      "learning_rate": 7.418269230769231e-05,
+      "loss": 0.1331,
+      "step": 9361
+    },
+    {
+      "epoch": 85.89,
+      "learning_rate": 7.413461538461538e-05,
+      "loss": 0.1691,
+      "step": 9362
+    },
+    {
+      "epoch": 85.9,
+      "learning_rate": 7.408653846153847e-05,
+      "loss": 0.1622,
+      "step": 9363
+    },
+    {
+      "epoch": 85.91,
+      "learning_rate": 7.403846153846154e-05,
+      "loss": 0.144,
+      "step": 9364
+    },
+    {
+      "epoch": 85.92,
+      "learning_rate": 7.399038461538461e-05,
+      "loss": 0.1144,
+      "step": 9365
+    },
+    {
+      "epoch": 85.93,
+      "learning_rate": 7.39423076923077e-05,
+      "loss": 0.155,
+      "step": 9366
+    },
+    {
+      "epoch": 85.94,
+      "learning_rate": 7.389423076923077e-05,
+      "loss": 0.175,
+      "step": 9367
+    },
+    {
+      "epoch": 85.94,
+      "learning_rate": 7.384615384615386e-05,
+      "loss": 0.1496,
+      "step": 9368
+    },
+    {
+      "epoch": 85.95,
+      "learning_rate": 7.379807692307693e-05,
+      "loss": 0.0624,
+      "step": 9369
+    },
+    {
+      "epoch": 85.96,
+      "learning_rate": 7.375e-05,
+      "loss": 0.1243,
+      "step": 9370
+    },
+    {
+      "epoch": 85.97,
+      "learning_rate": 7.370192307692308e-05,
+      "loss": 0.1526,
+      "step": 9371
+    },
+    {
+      "epoch": 85.98,
+      "learning_rate": 7.365384615384615e-05,
+      "loss": 0.1305,
+      "step": 9372
+    },
+    {
+      "epoch": 85.99,
+      "learning_rate": 7.360576923076922e-05,
+      "loss": 0.1113,
+      "step": 9373
+    },
+    {
+      "epoch": 86.0,
+      "learning_rate": 7.355769230769231e-05,
+      "loss": 0.1474,
+      "step": 9374
+    },
+    {
+      "epoch": 86.01,
+      "learning_rate": 7.350961538461538e-05,
+      "loss": 0.1679,
+      "step": 9375
+    },
+    {
+      "epoch": 86.02,
+      "learning_rate": 7.346153846153847e-05,
+      "loss": 0.1834,
+      "step": 9376
+    },
+    {
+      "epoch": 86.03,
+      "learning_rate": 7.341346153846154e-05,
+      "loss": 0.1287,
+      "step": 9377
+    },
+    {
+      "epoch": 86.04,
+      "learning_rate": 7.336538461538461e-05,
+      "loss": 0.1311,
+      "step": 9378
+    },
+    {
+      "epoch": 86.05,
+      "learning_rate": 7.33173076923077e-05,
+      "loss": 0.1295,
+      "step": 9379
+    },
+    {
+      "epoch": 86.06,
+      "learning_rate": 7.326923076923077e-05,
+      "loss": 0.1,
+      "step": 9380
+    },
+    {
+      "epoch": 86.06,
+      "learning_rate": 7.322115384615386e-05,
+      "loss": 0.1227,
+      "step": 9381
+    },
+    {
+      "epoch": 86.07,
+      "learning_rate": 7.317307692307693e-05,
+      "loss": 0.1811,
+      "step": 9382
+    },
+    {
+      "epoch": 86.08,
+      "learning_rate": 7.3125e-05,
+      "loss": 0.1125,
+      "step": 9383
+    },
+    {
+      "epoch": 86.09,
+      "learning_rate": 7.307692307692309e-05,
+      "loss": 0.1584,
+      "step": 9384
+    },
+    {
+      "epoch": 86.1,
+      "learning_rate": 7.302884615384615e-05,
+      "loss": 0.0998,
+      "step": 9385
+    },
+    {
+      "epoch": 86.11,
+      "learning_rate": 7.298076923076922e-05,
+      "loss": 0.1956,
+      "step": 9386
+    },
+    {
+      "epoch": 86.12,
+      "learning_rate": 7.29326923076923e-05,
+      "loss": 0.1393,
+      "step": 9387
+    },
+    {
+      "epoch": 86.13,
+      "learning_rate": 7.288461538461538e-05,
+      "loss": 0.077,
+      "step": 9388
+    },
+    {
+      "epoch": 86.14,
+      "learning_rate": 7.283653846153846e-05,
+      "loss": 0.0816,
+      "step": 9389
+    },
+    {
+      "epoch": 86.15,
+      "learning_rate": 7.278846153846154e-05,
+      "loss": 0.1006,
+      "step": 9390
+    },
+    {
+      "epoch": 86.16,
+      "learning_rate": 7.274038461538461e-05,
+      "loss": 0.0997,
+      "step": 9391
+    },
+    {
+      "epoch": 86.17,
+      "learning_rate": 7.26923076923077e-05,
+      "loss": 0.0961,
+      "step": 9392
+    },
+    {
+      "epoch": 86.17,
+      "learning_rate": 7.264423076923077e-05,
+      "loss": 0.1484,
+      "step": 9393
+    },
+    {
+      "epoch": 86.18,
+      "learning_rate": 7.259615384615386e-05,
+      "loss": 0.0862,
+      "step": 9394
+    },
+    {
+      "epoch": 86.19,
+      "learning_rate": 7.254807692307693e-05,
+      "loss": 0.1668,
+      "step": 9395
+    },
+    {
+      "epoch": 86.2,
+      "learning_rate": 7.25e-05,
+      "loss": 0.1111,
+      "step": 9396
+    },
+    {
+      "epoch": 86.21,
+      "learning_rate": 7.245192307692309e-05,
+      "loss": 0.208,
+      "step": 9397
+    },
+    {
+      "epoch": 86.22,
+      "learning_rate": 7.240384615384616e-05,
+      "loss": 0.1218,
+      "step": 9398
+    },
+    {
+      "epoch": 86.23,
+      "learning_rate": 7.235576923076925e-05,
+      "loss": 0.108,
+      "step": 9399
+    },
+    {
+      "epoch": 86.24,
+      "learning_rate": 7.23076923076923e-05,
+      "loss": 0.1372,
+      "step": 9400
+    },
+    {
+      "epoch": 86.25,
+      "learning_rate": 7.225961538461538e-05,
+      "loss": 0.0859,
+      "step": 9401
+    },
+    {
+      "epoch": 86.26,
+      "learning_rate": 7.221153846153846e-05,
+      "loss": 0.1804,
+      "step": 9402
+    },
+    {
+      "epoch": 86.27,
+      "learning_rate": 7.216346153846154e-05,
+      "loss": 0.2142,
+      "step": 9403
+    },
+    {
+      "epoch": 86.28,
+      "learning_rate": 7.211538461538461e-05,
+      "loss": 0.1402,
+      "step": 9404
+    },
+    {
+      "epoch": 86.28,
+      "learning_rate": 7.20673076923077e-05,
+      "loss": 0.1638,
+      "step": 9405
+    },
+    {
+      "epoch": 86.29,
+      "learning_rate": 7.201923076923077e-05,
+      "loss": 0.1403,
+      "step": 9406
+    },
+    {
+      "epoch": 86.3,
+      "learning_rate": 7.197115384615385e-05,
+      "loss": 0.1532,
+      "step": 9407
+    },
+    {
+      "epoch": 86.31,
+      "learning_rate": 7.192307692307693e-05,
+      "loss": 0.1371,
+      "step": 9408
+    },
+    {
+      "epoch": 86.32,
+      "learning_rate": 7.1875e-05,
+      "loss": 0.0713,
+      "step": 9409
+    },
+    {
+      "epoch": 86.33,
+      "learning_rate": 7.182692307692308e-05,
+      "loss": 0.1227,
+      "step": 9410
+    },
+    {
+      "epoch": 86.34,
+      "learning_rate": 7.177884615384616e-05,
+      "loss": 0.1559,
+      "step": 9411
+    },
+    {
+      "epoch": 86.35,
+      "learning_rate": 7.173076923076924e-05,
+      "loss": 0.1341,
+      "step": 9412
+    },
+    {
+      "epoch": 86.36,
+      "learning_rate": 7.168269230769232e-05,
+      "loss": 0.1139,
+      "step": 9413
+    },
+    {
+      "epoch": 86.37,
+      "learning_rate": 7.163461538461538e-05,
+      "loss": 0.1335,
+      "step": 9414
+    },
+    {
+      "epoch": 86.38,
+      "learning_rate": 7.158653846153846e-05,
+      "loss": 0.1238,
+      "step": 9415
+    },
+    {
+      "epoch": 86.39,
+      "learning_rate": 7.153846153846153e-05,
+      "loss": 0.1318,
+      "step": 9416
+    },
+    {
+      "epoch": 86.39,
+      "learning_rate": 7.149038461538462e-05,
+      "loss": 0.2198,
+      "step": 9417
+    },
+    {
+      "epoch": 86.4,
+      "learning_rate": 7.144230769230769e-05,
+      "loss": 0.1471,
+      "step": 9418
+    },
+    {
+      "epoch": 86.41,
+      "learning_rate": 7.139423076923077e-05,
+      "loss": 0.1189,
+      "step": 9419
+    },
+    {
+      "epoch": 86.42,
+      "learning_rate": 7.134615384615385e-05,
+      "loss": 0.1109,
+      "step": 9420
+    },
+    {
+      "epoch": 86.43,
+      "learning_rate": 7.129807692307692e-05,
+      "loss": 0.1004,
+      "step": 9421
+    },
+    {
+      "epoch": 86.44,
+      "learning_rate": 7.125e-05,
+      "loss": 0.1034,
+      "step": 9422
+    },
+    {
+      "epoch": 86.45,
+      "learning_rate": 7.120192307692308e-05,
+      "loss": 0.1587,
+      "step": 9423
+    },
+    {
+      "epoch": 86.46,
+      "learning_rate": 7.115384615384616e-05,
+      "loss": 0.2005,
+      "step": 9424
+    },
+    {
+      "epoch": 86.47,
+      "learning_rate": 7.110576923076924e-05,
+      "loss": 0.1199,
+      "step": 9425
+    },
+    {
+      "epoch": 86.48,
+      "learning_rate": 7.105769230769231e-05,
+      "loss": 0.158,
+      "step": 9426
+    },
+    {
+      "epoch": 86.49,
+      "learning_rate": 7.100961538461539e-05,
+      "loss": 0.0995,
+      "step": 9427
+    },
+    {
+      "epoch": 86.5,
+      "learning_rate": 7.096153846153846e-05,
+      "loss": 0.0528,
+      "step": 9428
+    },
+    {
+      "epoch": 86.5,
+      "learning_rate": 7.091346153846153e-05,
+      "loss": 0.2135,
+      "step": 9429
+    },
+    {
+      "epoch": 86.51,
+      "learning_rate": 7.086538461538462e-05,
+      "loss": 0.1509,
+      "step": 9430
+    },
+    {
+      "epoch": 86.52,
+      "learning_rate": 7.081730769230769e-05,
+      "loss": 0.2105,
+      "step": 9431
+    },
+    {
+      "epoch": 86.53,
+      "learning_rate": 7.076923076923076e-05,
+      "loss": 0.1546,
+      "step": 9432
+    },
+    {
+      "epoch": 86.54,
+      "learning_rate": 7.072115384615385e-05,
+      "loss": 0.1398,
+      "step": 9433
+    },
+    {
+      "epoch": 86.55,
+      "learning_rate": 7.067307692307692e-05,
+      "loss": 0.1662,
+      "step": 9434
+    },
+    {
+      "epoch": 86.56,
+      "learning_rate": 7.0625e-05,
+      "loss": 0.1573,
+      "step": 9435
+    },
+    {
+      "epoch": 86.57,
+      "learning_rate": 7.057692307692308e-05,
+      "loss": 0.143,
+      "step": 9436
+    },
+    {
+      "epoch": 86.58,
+      "learning_rate": 7.052884615384615e-05,
+      "loss": 0.1698,
+      "step": 9437
+    },
+    {
+      "epoch": 86.59,
+      "learning_rate": 7.048076923076924e-05,
+      "loss": 0.1589,
+      "step": 9438
+    },
+    {
+      "epoch": 86.6,
+      "learning_rate": 7.043269230769231e-05,
+      "loss": 0.1378,
+      "step": 9439
+    },
+    {
+      "epoch": 86.61,
+      "learning_rate": 7.038461538461539e-05,
+      "loss": 0.1418,
+      "step": 9440
+    },
+    {
+      "epoch": 86.61,
+      "learning_rate": 7.033653846153847e-05,
+      "loss": 0.2019,
+      "step": 9441
+    },
+    {
+      "epoch": 86.62,
+      "learning_rate": 7.028846153846153e-05,
+      "loss": 0.1008,
+      "step": 9442
+    },
+    {
+      "epoch": 86.63,
+      "learning_rate": 7.024038461538462e-05,
+      "loss": 0.1587,
+      "step": 9443
+    },
+    {
+      "epoch": 86.64,
+      "learning_rate": 7.019230769230769e-05,
+      "loss": 0.2315,
+      "step": 9444
+    },
+    {
+      "epoch": 86.65,
+      "learning_rate": 7.014423076923076e-05,
+      "loss": 0.0928,
+      "step": 9445
+    },
+    {
+      "epoch": 86.66,
+      "learning_rate": 7.009615384615385e-05,
+      "loss": 0.1177,
+      "step": 9446
+    },
+    {
+      "epoch": 86.67,
+      "learning_rate": 7.004807692307692e-05,
+      "loss": 0.1186,
+      "step": 9447
+    },
+    {
+      "epoch": 86.68,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 0.1442,
+      "step": 9448
+    },
+    {
+      "epoch": 86.69,
+      "learning_rate": 6.995192307692308e-05,
+      "loss": 0.0902,
+      "step": 9449
+    },
+    {
+      "epoch": 86.7,
+      "learning_rate": 6.990384615384615e-05,
+      "loss": 0.1755,
+      "step": 9450
+    },
+    {
+      "epoch": 86.71,
+      "learning_rate": 6.985576923076924e-05,
+      "loss": 0.1848,
+      "step": 9451
+    },
+    {
+      "epoch": 86.72,
+      "learning_rate": 6.980769230769231e-05,
+      "loss": 0.1121,
+      "step": 9452
+    },
+    {
+      "epoch": 86.72,
+      "learning_rate": 6.975961538461538e-05,
+      "loss": 0.1095,
+      "step": 9453
+    },
+    {
+      "epoch": 86.73,
+      "learning_rate": 6.971153846153847e-05,
+      "loss": 0.0726,
+      "step": 9454
+    },
+    {
+      "epoch": 86.74,
+      "learning_rate": 6.966346153846154e-05,
+      "loss": 0.0664,
+      "step": 9455
+    },
+    {
+      "epoch": 86.75,
+      "learning_rate": 6.961538461538462e-05,
+      "loss": 0.1714,
+      "step": 9456
+    },
+    {
+      "epoch": 86.76,
+      "learning_rate": 6.956730769230769e-05,
+      "loss": 0.1205,
+      "step": 9457
+    },
+    {
+      "epoch": 86.77,
+      "learning_rate": 6.951923076923076e-05,
+      "loss": 0.1663,
+      "step": 9458
+    },
+    {
+      "epoch": 86.78,
+      "learning_rate": 6.947115384615385e-05,
+      "loss": 0.1493,
+      "step": 9459
+    },
+    {
+      "epoch": 86.79,
+      "learning_rate": 6.942307692307692e-05,
+      "loss": 0.1718,
+      "step": 9460
+    },
+    {
+      "epoch": 86.8,
+      "learning_rate": 6.9375e-05,
+      "loss": 0.0988,
+      "step": 9461
+    },
+    {
+      "epoch": 86.81,
+      "learning_rate": 6.932692307692308e-05,
+      "loss": 0.1518,
+      "step": 9462
+    },
+    {
+      "epoch": 86.82,
+      "learning_rate": 6.927884615384615e-05,
+      "loss": 0.0988,
+      "step": 9463
+    },
+    {
+      "epoch": 86.83,
+      "learning_rate": 6.923076923076924e-05,
+      "loss": 0.1658,
+      "step": 9464
+    },
+    {
+      "epoch": 86.83,
+      "learning_rate": 6.918269230769231e-05,
+      "loss": 0.1929,
+      "step": 9465
+    },
+    {
+      "epoch": 86.84,
+      "learning_rate": 6.913461538461538e-05,
+      "loss": 0.0848,
+      "step": 9466
+    },
+    {
+      "epoch": 86.85,
+      "learning_rate": 6.908653846153847e-05,
+      "loss": 0.1664,
+      "step": 9467
+    },
+    {
+      "epoch": 86.86,
+      "learning_rate": 6.903846153846154e-05,
+      "loss": 0.1532,
+      "step": 9468
+    },
+    {
+      "epoch": 86.87,
+      "learning_rate": 6.899038461538463e-05,
+      "loss": 0.1214,
+      "step": 9469
+    },
+    {
+      "epoch": 86.88,
+      "learning_rate": 6.894230769230769e-05,
+      "loss": 0.1908,
+      "step": 9470
+    },
+    {
+      "epoch": 86.89,
+      "learning_rate": 6.889423076923076e-05,
+      "loss": 0.1849,
+      "step": 9471
+    },
+    {
+      "epoch": 86.9,
+      "learning_rate": 6.884615384615385e-05,
+      "loss": 0.1593,
+      "step": 9472
+    },
+    {
+      "epoch": 86.91,
+      "learning_rate": 6.879807692307692e-05,
+      "loss": 0.1327,
+      "step": 9473
+    },
+    {
+      "epoch": 86.92,
+      "learning_rate": 6.875e-05,
+      "loss": 0.1156,
+      "step": 9474
+    },
+    {
+      "epoch": 86.93,
+      "learning_rate": 6.870192307692308e-05,
+      "loss": 0.0964,
+      "step": 9475
+    },
+    {
+      "epoch": 86.94,
+      "learning_rate": 6.865384615384615e-05,
+      "loss": 0.1479,
+      "step": 9476
+    },
+    {
+      "epoch": 86.94,
+      "learning_rate": 6.860576923076924e-05,
+      "loss": 0.1444,
+      "step": 9477
+    },
+    {
+      "epoch": 86.95,
+      "learning_rate": 6.855769230769231e-05,
+      "loss": 0.0962,
+      "step": 9478
+    },
+    {
+      "epoch": 86.96,
+      "learning_rate": 6.85096153846154e-05,
+      "loss": 0.2024,
+      "step": 9479
+    },
+    {
+      "epoch": 86.97,
+      "learning_rate": 6.846153846153847e-05,
+      "loss": 0.1814,
+      "step": 9480
+    },
+    {
+      "epoch": 86.98,
+      "learning_rate": 6.841346153846154e-05,
+      "loss": 0.1375,
+      "step": 9481
+    },
+    {
+      "epoch": 86.99,
+      "learning_rate": 6.836538461538463e-05,
+      "loss": 0.0426,
+      "step": 9482
+    },
+    {
+      "epoch": 87.0,
+      "learning_rate": 6.83173076923077e-05,
+      "loss": 0.1461,
+      "step": 9483
+    },
+    {
+      "epoch": 87.01,
+      "learning_rate": 6.826923076923076e-05,
+      "loss": 0.1826,
+      "step": 9484
+    },
+    {
+      "epoch": 87.02,
+      "learning_rate": 6.822115384615384e-05,
+      "loss": 0.1389,
+      "step": 9485
+    },
+    {
+      "epoch": 87.03,
+      "learning_rate": 6.817307692307692e-05,
+      "loss": 0.1218,
+      "step": 9486
+    },
+    {
+      "epoch": 87.04,
+      "learning_rate": 6.8125e-05,
+      "loss": 0.138,
+      "step": 9487
+    },
+    {
+      "epoch": 87.05,
+      "learning_rate": 6.807692307692308e-05,
+      "loss": 0.1296,
+      "step": 9488
+    },
+    {
+      "epoch": 87.06,
+      "learning_rate": 6.802884615384615e-05,
+      "loss": 0.158,
+      "step": 9489
+    },
+    {
+      "epoch": 87.06,
+      "learning_rate": 6.798076923076923e-05,
+      "loss": 0.1789,
+      "step": 9490
+    },
+    {
+      "epoch": 87.07,
+      "learning_rate": 6.79326923076923e-05,
+      "loss": 0.1482,
+      "step": 9491
+    },
+    {
+      "epoch": 87.08,
+      "learning_rate": 6.788461538461539e-05,
+      "loss": 0.1955,
+      "step": 9492
+    },
+    {
+      "epoch": 87.09,
+      "learning_rate": 6.783653846153847e-05,
+      "loss": 0.1637,
+      "step": 9493
+    },
+    {
+      "epoch": 87.1,
+      "learning_rate": 6.778846153846154e-05,
+      "loss": 0.1061,
+      "step": 9494
+    },
+    {
+      "epoch": 87.11,
+      "learning_rate": 6.774038461538462e-05,
+      "loss": 0.1266,
+      "step": 9495
+    },
+    {
+      "epoch": 87.12,
+      "learning_rate": 6.76923076923077e-05,
+      "loss": 0.1789,
+      "step": 9496
+    },
+    {
+      "epoch": 87.13,
+      "learning_rate": 6.764423076923078e-05,
+      "loss": 0.1897,
+      "step": 9497
+    },
+    {
+      "epoch": 87.14,
+      "learning_rate": 6.759615384615386e-05,
+      "loss": 0.1709,
+      "step": 9498
+    },
+    {
+      "epoch": 87.15,
+      "learning_rate": 6.754807692307691e-05,
+      "loss": 0.1275,
+      "step": 9499
+    },
+    {
+      "epoch": 87.16,
+      "learning_rate": 6.75e-05,
+      "loss": 0.1355,
+      "step": 9500
+    },
+    {
+      "epoch": 87.16,
+      "eval_cer": 0.13751330604656742,
+      "eval_loss": 0.7231259942054749,
+      "eval_runtime": 87.6102,
+      "eval_samples_per_second": 18.799,
+      "eval_steps_per_second": 2.351,
+      "eval_wer": 0.4868757021754673,
+      "step": 9500
+    },
+    {
+      "epoch": 87.17,
+      "learning_rate": 6.745192307692307e-05,
+      "loss": 0.1996,
+      "step": 9501
+    },
+    {
+      "epoch": 87.17,
+      "learning_rate": 6.740384615384615e-05,
+      "loss": 0.1264,
+      "step": 9502
+    },
+    {
+      "epoch": 87.18,
+      "learning_rate": 6.735576923076923e-05,
+      "loss": 0.113,
+      "step": 9503
+    },
+    {
+      "epoch": 87.19,
+      "learning_rate": 6.73076923076923e-05,
+      "loss": 0.1021,
+      "step": 9504
+    },
+    {
+      "epoch": 87.2,
+      "learning_rate": 6.725961538461539e-05,
+      "loss": 0.0832,
+      "step": 9505
+    },
+    {
+      "epoch": 87.21,
+      "learning_rate": 6.721153846153846e-05,
+      "loss": 0.142,
+      "step": 9506
+    },
+    {
+      "epoch": 87.22,
+      "learning_rate": 6.716346153846154e-05,
+      "loss": 0.0934,
+      "step": 9507
+    },
+    {
+      "epoch": 87.23,
+      "learning_rate": 6.711538461538462e-05,
+      "loss": 0.0711,
+      "step": 9508
+    },
+    {
+      "epoch": 87.24,
+      "learning_rate": 6.70673076923077e-05,
+      "loss": 0.1208,
+      "step": 9509
+    },
+    {
+      "epoch": 87.25,
+      "learning_rate": 6.701923076923078e-05,
+      "loss": 0.0663,
+      "step": 9510
+    },
+    {
+      "epoch": 87.26,
+      "learning_rate": 6.697115384615385e-05,
+      "loss": 0.1678,
+      "step": 9511
+    },
+    {
+      "epoch": 87.27,
+      "learning_rate": 6.692307692307693e-05,
+      "loss": 0.1582,
+      "step": 9512
+    },
+    {
+      "epoch": 87.28,
+      "learning_rate": 6.6875e-05,
+      "loss": 0.1865,
+      "step": 9513
+    },
+    {
+      "epoch": 87.28,
+      "learning_rate": 6.682692307692307e-05,
+      "loss": 0.1658,
+      "step": 9514
+    },
+    {
+      "epoch": 87.29,
+      "learning_rate": 6.677884615384614e-05,
+      "loss": 0.2006,
+      "step": 9515
+    },
+    {
+      "epoch": 87.3,
+      "learning_rate": 6.673076923076923e-05,
+      "loss": 0.1505,
+      "step": 9516
+    },
+    {
+      "epoch": 87.31,
+      "learning_rate": 6.66826923076923e-05,
+      "loss": 0.1607,
+      "step": 9517
+    },
+    {
+      "epoch": 87.32,
+      "learning_rate": 6.663461538461539e-05,
+      "loss": 0.0957,
+      "step": 9518
+    },
+    {
+      "epoch": 87.33,
+      "learning_rate": 6.658653846153846e-05,
+      "loss": 0.0856,
+      "step": 9519
+    },
+    {
+      "epoch": 87.34,
+      "learning_rate": 6.653846153846153e-05,
+      "loss": 0.1901,
+      "step": 9520
+    },
+    {
+      "epoch": 87.35,
+      "learning_rate": 6.649038461538462e-05,
+      "loss": 0.1357,
+      "step": 9521
+    },
+    {
+      "epoch": 87.36,
+      "learning_rate": 6.64423076923077e-05,
+      "loss": 0.1106,
+      "step": 9522
+    },
+    {
+      "epoch": 87.37,
+      "learning_rate": 6.639423076923078e-05,
+      "loss": 0.112,
+      "step": 9523
+    },
+    {
+      "epoch": 87.38,
+      "learning_rate": 6.634615384615385e-05,
+      "loss": 0.1785,
+      "step": 9524
+    },
+    {
+      "epoch": 87.39,
+      "learning_rate": 6.629807692307692e-05,
+      "loss": 0.1949,
+      "step": 9525
+    },
+    {
+      "epoch": 87.39,
+      "learning_rate": 6.625000000000001e-05,
+      "loss": 0.2036,
+      "step": 9526
+    },
+    {
+      "epoch": 87.4,
+      "learning_rate": 6.620192307692307e-05,
+      "loss": 0.107,
+      "step": 9527
+    },
+    {
+      "epoch": 87.41,
+      "learning_rate": 6.615384615384616e-05,
+      "loss": 0.0958,
+      "step": 9528
+    },
+    {
+      "epoch": 87.42,
+      "learning_rate": 6.610576923076923e-05,
+      "loss": 0.1589,
+      "step": 9529
+    },
+    {
+      "epoch": 87.43,
+      "learning_rate": 6.60576923076923e-05,
+      "loss": 0.1254,
+      "step": 9530
+    },
+    {
+      "epoch": 87.44,
+      "learning_rate": 6.600961538461539e-05,
+      "loss": 0.1216,
+      "step": 9531
+    },
+    {
+      "epoch": 87.45,
+      "learning_rate": 6.596153846153846e-05,
+      "loss": 0.1428,
+      "step": 9532
+    },
+    {
+      "epoch": 87.46,
+      "learning_rate": 6.591346153846153e-05,
+      "loss": 0.1391,
+      "step": 9533
+    },
+    {
+      "epoch": 87.47,
+      "learning_rate": 6.586538461538462e-05,
+      "loss": 0.1575,
+      "step": 9534
+    },
+    {
+      "epoch": 87.48,
+      "learning_rate": 6.581730769230769e-05,
+      "loss": 0.0904,
+      "step": 9535
+    },
+    {
+      "epoch": 87.49,
+      "learning_rate": 6.576923076923078e-05,
+      "loss": 0.1451,
+      "step": 9536
+    },
+    {
+      "epoch": 87.5,
+      "learning_rate": 6.572115384615385e-05,
+      "loss": 0.1107,
+      "step": 9537
+    },
+    {
+      "epoch": 87.5,
+      "learning_rate": 6.567307692307692e-05,
+      "loss": 0.1919,
+      "step": 9538
+    },
+    {
+      "epoch": 87.51,
+      "learning_rate": 6.562500000000001e-05,
+      "loss": 0.1193,
+      "step": 9539
+    },
+    {
+      "epoch": 87.52,
+      "learning_rate": 6.557692307692308e-05,
+      "loss": 0.1324,
+      "step": 9540
+    },
+    {
+      "epoch": 87.53,
+      "learning_rate": 6.552884615384615e-05,
+      "loss": 0.1574,
+      "step": 9541
+    },
+    {
+      "epoch": 87.54,
+      "learning_rate": 6.548076923076923e-05,
+      "loss": 0.1442,
+      "step": 9542
+    },
+    {
+      "epoch": 87.55,
+      "learning_rate": 6.54326923076923e-05,
+      "loss": 0.1341,
+      "step": 9543
+    },
+    {
+      "epoch": 87.56,
+      "learning_rate": 6.538461538461539e-05,
+      "loss": 0.1532,
+      "step": 9544
+    },
+    {
+      "epoch": 87.57,
+      "learning_rate": 6.533653846153846e-05,
+      "loss": 0.1194,
+      "step": 9545
+    },
+    {
+      "epoch": 87.58,
+      "learning_rate": 6.528846153846154e-05,
+      "loss": 0.1266,
+      "step": 9546
+    },
+    {
+      "epoch": 87.59,
+      "learning_rate": 6.524038461538462e-05,
+      "loss": 0.1222,
+      "step": 9547
+    },
+    {
+      "epoch": 87.6,
+      "learning_rate": 6.519230769230769e-05,
+      "loss": 0.147,
+      "step": 9548
+    },
+    {
+      "epoch": 87.61,
+      "learning_rate": 6.514423076923078e-05,
+      "loss": 0.1699,
+      "step": 9549
+    },
+    {
+      "epoch": 87.61,
+      "learning_rate": 6.509615384615385e-05,
+      "loss": 0.1203,
+      "step": 9550
+    },
+    {
+      "epoch": 87.62,
+      "learning_rate": 6.504807692307692e-05,
+      "loss": 0.1695,
+      "step": 9551
+    },
+    {
+      "epoch": 87.63,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 0.1181,
+      "step": 9552
+    },
+    {
+      "epoch": 87.64,
+      "learning_rate": 6.495192307692308e-05,
+      "loss": 0.1796,
+      "step": 9553
+    },
+    {
+      "epoch": 87.65,
+      "learning_rate": 6.490384615384617e-05,
+      "loss": 0.1312,
+      "step": 9554
+    },
+    {
+      "epoch": 87.66,
+      "learning_rate": 6.485576923076923e-05,
+      "loss": 0.0787,
+      "step": 9555
+    },
+    {
+      "epoch": 87.67,
+      "learning_rate": 6.48076923076923e-05,
+      "loss": 0.1266,
+      "step": 9556
+    },
+    {
+      "epoch": 87.68,
+      "learning_rate": 6.475961538461538e-05,
+      "loss": 0.1416,
+      "step": 9557
+    },
+    {
+      "epoch": 87.69,
+      "learning_rate": 6.471153846153846e-05,
+      "loss": 0.1461,
+      "step": 9558
+    },
+    {
+      "epoch": 87.7,
+      "learning_rate": 6.466346153846154e-05,
+      "loss": 0.0855,
+      "step": 9559
+    },
+    {
+      "epoch": 87.71,
+      "learning_rate": 6.461538461538462e-05,
+      "loss": 0.1158,
+      "step": 9560
+    },
+    {
+      "epoch": 87.72,
+      "learning_rate": 6.456730769230769e-05,
+      "loss": 0.1457,
+      "step": 9561
+    },
+    {
+      "epoch": 87.72,
+      "learning_rate": 6.451923076923077e-05,
+      "loss": 0.1088,
+      "step": 9562
+    },
+    {
+      "epoch": 87.73,
+      "learning_rate": 6.447115384615385e-05,
+      "loss": 0.138,
+      "step": 9563
+    },
+    {
+      "epoch": 87.74,
+      "learning_rate": 6.442307692307692e-05,
+      "loss": 0.0832,
+      "step": 9564
+    },
+    {
+      "epoch": 87.75,
+      "learning_rate": 6.4375e-05,
+      "loss": 0.2148,
+      "step": 9565
+    },
+    {
+      "epoch": 87.76,
+      "learning_rate": 6.432692307692308e-05,
+      "loss": 0.1498,
+      "step": 9566
+    },
+    {
+      "epoch": 87.77,
+      "learning_rate": 6.427884615384617e-05,
+      "loss": 0.0962,
+      "step": 9567
+    },
+    {
+      "epoch": 87.78,
+      "learning_rate": 6.423076923076924e-05,
+      "loss": 0.1845,
+      "step": 9568
+    },
+    {
+      "epoch": 87.79,
+      "learning_rate": 6.41826923076923e-05,
+      "loss": 0.1193,
+      "step": 9569
+    },
+    {
+      "epoch": 87.8,
+      "learning_rate": 6.413461538461538e-05,
+      "loss": 0.1714,
+      "step": 9570
+    },
+    {
+      "epoch": 87.81,
+      "learning_rate": 6.408653846153846e-05,
+      "loss": 0.1226,
+      "step": 9571
+    },
+    {
+      "epoch": 87.82,
+      "learning_rate": 6.403846153846154e-05,
+      "loss": 0.1721,
+      "step": 9572
+    },
+    {
+      "epoch": 87.83,
+      "learning_rate": 6.399038461538461e-05,
+      "loss": 0.1375,
+      "step": 9573
+    },
+    {
+      "epoch": 87.83,
+      "learning_rate": 6.394230769230769e-05,
+      "loss": 0.1527,
+      "step": 9574
+    },
+    {
+      "epoch": 87.84,
+      "learning_rate": 6.389423076923077e-05,
+      "loss": 0.1312,
+      "step": 9575
+    },
+    {
+      "epoch": 87.85,
+      "learning_rate": 6.384615384615385e-05,
+      "loss": 0.073,
+      "step": 9576
+    },
+    {
+      "epoch": 87.86,
+      "learning_rate": 6.379807692307693e-05,
+      "loss": 0.1198,
+      "step": 9577
+    },
+    {
+      "epoch": 87.87,
+      "learning_rate": 6.375e-05,
+      "loss": 0.1454,
+      "step": 9578
+    },
+    {
+      "epoch": 87.88,
+      "learning_rate": 6.370192307692308e-05,
+      "loss": 0.1873,
+      "step": 9579
+    },
+    {
+      "epoch": 87.89,
+      "learning_rate": 6.365384615384616e-05,
+      "loss": 0.0827,
+      "step": 9580
+    },
+    {
+      "epoch": 87.9,
+      "learning_rate": 6.360576923076924e-05,
+      "loss": 0.1549,
+      "step": 9581
+    },
+    {
+      "epoch": 87.91,
+      "learning_rate": 6.355769230769231e-05,
+      "loss": 0.139,
+      "step": 9582
+    },
+    {
+      "epoch": 87.92,
+      "learning_rate": 6.350961538461538e-05,
+      "loss": 0.1254,
+      "step": 9583
+    },
+    {
+      "epoch": 87.93,
+      "learning_rate": 6.346153846153845e-05,
+      "loss": 0.1223,
+      "step": 9584
+    },
+    {
+      "epoch": 87.94,
+      "learning_rate": 6.341346153846154e-05,
+      "loss": 0.1137,
+      "step": 9585
+    },
+    {
+      "epoch": 87.94,
+      "learning_rate": 6.336538461538461e-05,
+      "loss": 0.1299,
+      "step": 9586
+    },
+    {
+      "epoch": 87.95,
+      "learning_rate": 6.331730769230769e-05,
+      "loss": 0.1344,
+      "step": 9587
+    },
+    {
+      "epoch": 87.96,
+      "learning_rate": 6.326923076923077e-05,
+      "loss": 0.1739,
+      "step": 9588
+    },
+    {
+      "epoch": 87.97,
+      "learning_rate": 6.322115384615384e-05,
+      "loss": 0.0571,
+      "step": 9589
+    },
+    {
+      "epoch": 87.98,
+      "learning_rate": 6.317307692307693e-05,
+      "loss": 0.1085,
+      "step": 9590
+    },
+    {
+      "epoch": 87.99,
+      "learning_rate": 6.3125e-05,
+      "loss": 0.0683,
+      "step": 9591
+    },
+    {
+      "epoch": 88.0,
+      "learning_rate": 6.307692307692308e-05,
+      "loss": 0.1292,
+      "step": 9592
+    },
+    {
+      "epoch": 88.01,
+      "learning_rate": 6.302884615384616e-05,
+      "loss": 0.2118,
+      "step": 9593
+    },
+    {
+      "epoch": 88.02,
+      "learning_rate": 6.298076923076923e-05,
+      "loss": 0.1485,
+      "step": 9594
+    },
+    {
+      "epoch": 88.03,
+      "learning_rate": 6.293269230769232e-05,
+      "loss": 0.216,
+      "step": 9595
+    },
+    {
+      "epoch": 88.04,
+      "learning_rate": 6.288461538461539e-05,
+      "loss": 0.13,
+      "step": 9596
+    },
+    {
+      "epoch": 88.05,
+      "learning_rate": 6.283653846153845e-05,
+      "loss": 0.1422,
+      "step": 9597
+    },
+    {
+      "epoch": 88.06,
+      "learning_rate": 6.278846153846154e-05,
+      "loss": 0.1442,
+      "step": 9598
+    },
+    {
+      "epoch": 88.06,
+      "learning_rate": 6.274038461538461e-05,
+      "loss": 0.1331,
+      "step": 9599
+    },
+    {
+      "epoch": 88.07,
+      "learning_rate": 6.269230769230768e-05,
+      "loss": 0.1218,
+      "step": 9600
+    },
+    {
+      "epoch": 88.08,
+      "learning_rate": 6.264423076923077e-05,
+      "loss": 0.0791,
+      "step": 9601
+    },
+    {
+      "epoch": 88.09,
+      "learning_rate": 6.259615384615384e-05,
+      "loss": 0.1788,
+      "step": 9602
+    },
+    {
+      "epoch": 88.1,
+      "learning_rate": 6.254807692307693e-05,
+      "loss": 0.1207,
+      "step": 9603
+    },
+    {
+      "epoch": 88.11,
+      "learning_rate": 6.25e-05,
+      "loss": 0.1348,
+      "step": 9604
+    },
+    {
+      "epoch": 88.12,
+      "learning_rate": 6.245192307692307e-05,
+      "loss": 0.0937,
+      "step": 9605
+    },
+    {
+      "epoch": 88.13,
+      "learning_rate": 6.240384615384616e-05,
+      "loss": 0.1446,
+      "step": 9606
+    },
+    {
+      "epoch": 88.14,
+      "learning_rate": 6.235576923076923e-05,
+      "loss": 0.1008,
+      "step": 9607
+    },
+    {
+      "epoch": 88.15,
+      "learning_rate": 6.23076923076923e-05,
+      "loss": 0.1219,
+      "step": 9608
+    },
+    {
+      "epoch": 88.16,
+      "learning_rate": 6.225961538461539e-05,
+      "loss": 0.174,
+      "step": 9609
+    },
+    {
+      "epoch": 88.17,
+      "learning_rate": 6.221153846153846e-05,
+      "loss": 0.1145,
+      "step": 9610
+    },
+    {
+      "epoch": 88.17,
+      "learning_rate": 6.216346153846154e-05,
+      "loss": 0.1028,
+      "step": 9611
+    },
+    {
+      "epoch": 88.18,
+      "learning_rate": 6.211538461538461e-05,
+      "loss": 0.1331,
+      "step": 9612
+    },
+    {
+      "epoch": 88.19,
+      "learning_rate": 6.20673076923077e-05,
+      "loss": 0.1283,
+      "step": 9613
+    },
+    {
+      "epoch": 88.2,
+      "learning_rate": 6.201923076923077e-05,
+      "loss": 0.1415,
+      "step": 9614
+    },
+    {
+      "epoch": 88.21,
+      "learning_rate": 6.197115384615385e-05,
+      "loss": 0.1471,
+      "step": 9615
+    },
+    {
+      "epoch": 88.22,
+      "learning_rate": 6.192307692307693e-05,
+      "loss": 0.1308,
+      "step": 9616
+    },
+    {
+      "epoch": 88.23,
+      "learning_rate": 6.1875e-05,
+      "loss": 0.1141,
+      "step": 9617
+    },
+    {
+      "epoch": 88.24,
+      "learning_rate": 6.182692307692309e-05,
+      "loss": 0.0765,
+      "step": 9618
+    },
+    {
+      "epoch": 88.25,
+      "learning_rate": 6.177884615384616e-05,
+      "loss": 0.0409,
+      "step": 9619
+    },
+    {
+      "epoch": 88.26,
+      "learning_rate": 6.173076923076923e-05,
+      "loss": 0.1979,
+      "step": 9620
+    },
+    {
+      "epoch": 88.27,
+      "learning_rate": 6.16826923076923e-05,
+      "loss": 0.2122,
+      "step": 9621
+    },
+    {
+      "epoch": 88.28,
+      "learning_rate": 6.163461538461539e-05,
+      "loss": 0.1194,
+      "step": 9622
+    },
+    {
+      "epoch": 88.28,
+      "learning_rate": 6.158653846153846e-05,
+      "loss": 0.1441,
+      "step": 9623
+    },
+    {
+      "epoch": 88.29,
+      "learning_rate": 6.153846153846155e-05,
+      "loss": 0.1344,
+      "step": 9624
+    },
+    {
+      "epoch": 88.3,
+      "learning_rate": 6.149038461538462e-05,
+      "loss": 0.1353,
+      "step": 9625
+    },
+    {
+      "epoch": 88.31,
+      "learning_rate": 6.14423076923077e-05,
+      "loss": 0.1292,
+      "step": 9626
+    },
+    {
+      "epoch": 88.32,
+      "learning_rate": 6.139423076923077e-05,
+      "loss": 0.1471,
+      "step": 9627
+    },
+    {
+      "epoch": 88.33,
+      "learning_rate": 6.134615384615385e-05,
+      "loss": 0.1182,
+      "step": 9628
+    },
+    {
+      "epoch": 88.34,
+      "learning_rate": 6.129807692307693e-05,
+      "loss": 0.0703,
+      "step": 9629
+    },
+    {
+      "epoch": 88.35,
+      "learning_rate": 6.125e-05,
+      "loss": 0.1057,
+      "step": 9630
+    },
+    {
+      "epoch": 88.36,
+      "learning_rate": 6.120192307692308e-05,
+      "loss": 0.1825,
+      "step": 9631
+    },
+    {
+      "epoch": 88.37,
+      "learning_rate": 6.115384615384616e-05,
+      "loss": 0.1523,
+      "step": 9632
+    },
+    {
+      "epoch": 88.38,
+      "learning_rate": 6.110576923076923e-05,
+      "loss": 0.1005,
+      "step": 9633
+    },
+    {
+      "epoch": 88.39,
+      "learning_rate": 6.10576923076923e-05,
+      "loss": 0.1161,
+      "step": 9634
+    },
+    {
+      "epoch": 88.39,
+      "learning_rate": 6.100961538461539e-05,
+      "loss": 0.1246,
+      "step": 9635
+    },
+    {
+      "epoch": 88.4,
+      "learning_rate": 6.096153846153846e-05,
+      "loss": 0.1058,
+      "step": 9636
+    },
+    {
+      "epoch": 88.41,
+      "learning_rate": 6.091346153846154e-05,
+      "loss": 0.1294,
+      "step": 9637
+    },
+    {
+      "epoch": 88.42,
+      "learning_rate": 6.086538461538461e-05,
+      "loss": 0.0687,
+      "step": 9638
+    },
+    {
+      "epoch": 88.43,
+      "learning_rate": 6.081730769230769e-05,
+      "loss": 0.1614,
+      "step": 9639
+    },
+    {
+      "epoch": 88.44,
+      "learning_rate": 6.076923076923077e-05,
+      "loss": 0.1409,
+      "step": 9640
+    },
+    {
+      "epoch": 88.45,
+      "learning_rate": 6.072115384615385e-05,
+      "loss": 0.1481,
+      "step": 9641
+    },
+    {
+      "epoch": 88.46,
+      "learning_rate": 6.0673076923076924e-05,
+      "loss": 0.1433,
+      "step": 9642
+    },
+    {
+      "epoch": 88.47,
+      "learning_rate": 6.0624999999999996e-05,
+      "loss": 0.071,
+      "step": 9643
+    },
+    {
+      "epoch": 88.48,
+      "learning_rate": 6.0576923076923076e-05,
+      "loss": 0.1086,
+      "step": 9644
+    },
+    {
+      "epoch": 88.49,
+      "learning_rate": 6.0528846153846155e-05,
+      "loss": 0.0777,
+      "step": 9645
+    },
+    {
+      "epoch": 88.5,
+      "learning_rate": 6.0480769230769235e-05,
+      "loss": 0.1076,
+      "step": 9646
+    },
+    {
+      "epoch": 88.5,
+      "learning_rate": 6.043269230769231e-05,
+      "loss": 0.2214,
+      "step": 9647
+    },
+    {
+      "epoch": 88.51,
+      "learning_rate": 6.038461538461539e-05,
+      "loss": 0.1561,
+      "step": 9648
+    },
+    {
+      "epoch": 88.52,
+      "learning_rate": 6.0336538461538466e-05,
+      "loss": 0.1659,
+      "step": 9649
+    },
+    {
+      "epoch": 88.53,
+      "learning_rate": 6.028846153846154e-05,
+      "loss": 0.0981,
+      "step": 9650
+    },
+    {
+      "epoch": 88.54,
+      "learning_rate": 6.024038461538461e-05,
+      "loss": 0.1279,
+      "step": 9651
+    },
+    {
+      "epoch": 88.55,
+      "learning_rate": 6.019230769230769e-05,
+      "loss": 0.1563,
+      "step": 9652
+    },
+    {
+      "epoch": 88.56,
+      "learning_rate": 6.014423076923077e-05,
+      "loss": 0.1304,
+      "step": 9653
+    },
+    {
+      "epoch": 88.57,
+      "learning_rate": 6.009615384615385e-05,
+      "loss": 0.2099,
+      "step": 9654
+    },
+    {
+      "epoch": 88.58,
+      "learning_rate": 6.004807692307693e-05,
+      "loss": 0.1883,
+      "step": 9655
+    },
+    {
+      "epoch": 88.59,
+      "learning_rate": 6e-05,
+      "loss": 0.1068,
+      "step": 9656
+    },
+    {
+      "epoch": 88.6,
+      "learning_rate": 5.9951923076923074e-05,
+      "loss": 0.1524,
+      "step": 9657
+    },
+    {
+      "epoch": 88.61,
+      "learning_rate": 5.9903846153846154e-05,
+      "loss": 0.1896,
+      "step": 9658
+    },
+    {
+      "epoch": 88.61,
+      "learning_rate": 5.985576923076923e-05,
+      "loss": 0.1074,
+      "step": 9659
+    },
+    {
+      "epoch": 88.62,
+      "learning_rate": 5.9807692307692306e-05,
+      "loss": 0.1495,
+      "step": 9660
+    },
+    {
+      "epoch": 88.63,
+      "learning_rate": 5.9759615384615385e-05,
+      "loss": 0.1727,
+      "step": 9661
+    },
+    {
+      "epoch": 88.64,
+      "learning_rate": 5.9711538461538464e-05,
+      "loss": 0.0765,
+      "step": 9662
+    },
+    {
+      "epoch": 88.65,
+      "learning_rate": 5.9663461538461544e-05,
+      "loss": 0.1558,
+      "step": 9663
+    },
+    {
+      "epoch": 88.66,
+      "learning_rate": 5.961538461538462e-05,
+      "loss": 0.0986,
+      "step": 9664
+    },
+    {
+      "epoch": 88.67,
+      "learning_rate": 5.956730769230769e-05,
+      "loss": 0.2356,
+      "step": 9665
+    },
+    {
+      "epoch": 88.68,
+      "learning_rate": 5.951923076923077e-05,
+      "loss": 0.1203,
+      "step": 9666
+    },
+    {
+      "epoch": 88.69,
+      "learning_rate": 5.947115384615385e-05,
+      "loss": 0.1462,
+      "step": 9667
+    },
+    {
+      "epoch": 88.7,
+      "learning_rate": 5.942307692307693e-05,
+      "loss": 0.1389,
+      "step": 9668
+    },
+    {
+      "epoch": 88.71,
+      "learning_rate": 5.9375e-05,
+      "loss": 0.0584,
+      "step": 9669
+    },
+    {
+      "epoch": 88.72,
+      "learning_rate": 5.932692307692308e-05,
+      "loss": 0.1013,
+      "step": 9670
+    },
+    {
+      "epoch": 88.72,
+      "learning_rate": 5.927884615384616e-05,
+      "loss": 0.1175,
+      "step": 9671
+    },
+    {
+      "epoch": 88.73,
+      "learning_rate": 5.923076923076923e-05,
+      "loss": 0.1118,
+      "step": 9672
+    },
+    {
+      "epoch": 88.74,
+      "learning_rate": 5.918269230769231e-05,
+      "loss": 0.0841,
+      "step": 9673
+    },
+    {
+      "epoch": 88.75,
+      "learning_rate": 5.913461538461538e-05,
+      "loss": 0.1866,
+      "step": 9674
+    },
+    {
+      "epoch": 88.76,
+      "learning_rate": 5.908653846153846e-05,
+      "loss": 0.1216,
+      "step": 9675
+    },
+    {
+      "epoch": 88.77,
+      "learning_rate": 5.903846153846154e-05,
+      "loss": 0.1678,
+      "step": 9676
+    },
+    {
+      "epoch": 88.78,
+      "learning_rate": 5.899038461538462e-05,
+      "loss": 0.1817,
+      "step": 9677
+    },
+    {
+      "epoch": 88.79,
+      "learning_rate": 5.8942307692307694e-05,
+      "loss": 0.1171,
+      "step": 9678
+    },
+    {
+      "epoch": 88.8,
+      "learning_rate": 5.889423076923077e-05,
+      "loss": 0.1293,
+      "step": 9679
+    },
+    {
+      "epoch": 88.81,
+      "learning_rate": 5.8846153846153846e-05,
+      "loss": 0.1277,
+      "step": 9680
+    },
+    {
+      "epoch": 88.82,
+      "learning_rate": 5.8798076923076926e-05,
+      "loss": 0.1557,
+      "step": 9681
+    },
+    {
+      "epoch": 88.83,
+      "learning_rate": 5.875e-05,
+      "loss": 0.069,
+      "step": 9682
+    },
+    {
+      "epoch": 88.83,
+      "learning_rate": 5.870192307692308e-05,
+      "loss": 0.1769,
+      "step": 9683
+    },
+    {
+      "epoch": 88.84,
+      "learning_rate": 5.865384615384616e-05,
+      "loss": 0.1206,
+      "step": 9684
+    },
+    {
+      "epoch": 88.85,
+      "learning_rate": 5.8605769230769236e-05,
+      "loss": 0.139,
+      "step": 9685
+    },
+    {
+      "epoch": 88.86,
+      "learning_rate": 5.855769230769231e-05,
+      "loss": 0.1397,
+      "step": 9686
+    },
+    {
+      "epoch": 88.87,
+      "learning_rate": 5.850961538461538e-05,
+      "loss": 0.1732,
+      "step": 9687
+    },
+    {
+      "epoch": 88.88,
+      "learning_rate": 5.846153846153846e-05,
+      "loss": 0.1792,
+      "step": 9688
+    },
+    {
+      "epoch": 88.89,
+      "learning_rate": 5.841346153846154e-05,
+      "loss": 0.1733,
+      "step": 9689
+    },
+    {
+      "epoch": 88.9,
+      "learning_rate": 5.836538461538462e-05,
+      "loss": 0.0796,
+      "step": 9690
+    },
+    {
+      "epoch": 88.91,
+      "learning_rate": 5.831730769230769e-05,
+      "loss": 0.0867,
+      "step": 9691
+    },
+    {
+      "epoch": 88.92,
+      "learning_rate": 5.826923076923077e-05,
+      "loss": 0.1352,
+      "step": 9692
+    },
+    {
+      "epoch": 88.93,
+      "learning_rate": 5.8221153846153845e-05,
+      "loss": 0.0587,
+      "step": 9693
+    },
+    {
+      "epoch": 88.94,
+      "learning_rate": 5.8173076923076924e-05,
+      "loss": 0.1228,
+      "step": 9694
+    },
+    {
+      "epoch": 88.94,
+      "learning_rate": 5.8125e-05,
+      "loss": 0.1109,
+      "step": 9695
+    },
+    {
+      "epoch": 88.95,
+      "learning_rate": 5.8076923076923076e-05,
+      "loss": 0.1246,
+      "step": 9696
+    },
+    {
+      "epoch": 88.96,
+      "learning_rate": 5.8028846153846155e-05,
+      "loss": 0.0764,
+      "step": 9697
+    },
+    {
+      "epoch": 88.97,
+      "learning_rate": 5.7980769230769235e-05,
+      "loss": 0.1184,
+      "step": 9698
+    },
+    {
+      "epoch": 88.98,
+      "learning_rate": 5.7932692307692314e-05,
+      "loss": 0.0925,
+      "step": 9699
+    },
+    {
+      "epoch": 88.99,
+      "learning_rate": 5.788461538461538e-05,
+      "loss": 0.0761,
+      "step": 9700
+    },
+    {
+      "epoch": 89.0,
+      "learning_rate": 5.783653846153846e-05,
+      "loss": 0.107,
+      "step": 9701
+    },
+    {
+      "epoch": 89.01,
+      "learning_rate": 5.778846153846154e-05,
+      "loss": 0.197,
+      "step": 9702
+    },
+    {
+      "epoch": 89.02,
+      "learning_rate": 5.774038461538462e-05,
+      "loss": 0.1454,
+      "step": 9703
+    },
+    {
+      "epoch": 89.03,
+      "learning_rate": 5.76923076923077e-05,
+      "loss": 0.1786,
+      "step": 9704
+    },
+    {
+      "epoch": 89.04,
+      "learning_rate": 5.764423076923077e-05,
+      "loss": 0.1343,
+      "step": 9705
+    },
+    {
+      "epoch": 89.05,
+      "learning_rate": 5.759615384615385e-05,
+      "loss": 0.1453,
+      "step": 9706
+    },
+    {
+      "epoch": 89.06,
+      "learning_rate": 5.754807692307692e-05,
+      "loss": 0.1737,
+      "step": 9707
+    },
+    {
+      "epoch": 89.06,
+      "learning_rate": 5.75e-05,
+      "loss": 0.1285,
+      "step": 9708
+    },
+    {
+      "epoch": 89.07,
+      "learning_rate": 5.7451923076923074e-05,
+      "loss": 0.1032,
+      "step": 9709
+    },
+    {
+      "epoch": 89.08,
+      "learning_rate": 5.7403846153846154e-05,
+      "loss": 0.1601,
+      "step": 9710
+    },
+    {
+      "epoch": 89.09,
+      "learning_rate": 5.735576923076923e-05,
+      "loss": 0.2164,
+      "step": 9711
+    },
+    {
+      "epoch": 89.1,
+      "learning_rate": 5.730769230769231e-05,
+      "loss": 0.1312,
+      "step": 9712
+    },
+    {
+      "epoch": 89.11,
+      "learning_rate": 5.725961538461539e-05,
+      "loss": 0.1755,
+      "step": 9713
+    },
+    {
+      "epoch": 89.12,
+      "learning_rate": 5.721153846153846e-05,
+      "loss": 0.1354,
+      "step": 9714
+    },
+    {
+      "epoch": 89.13,
+      "learning_rate": 5.716346153846154e-05,
+      "loss": 0.1687,
+      "step": 9715
+    },
+    {
+      "epoch": 89.14,
+      "learning_rate": 5.711538461538462e-05,
+      "loss": 0.1288,
+      "step": 9716
+    },
+    {
+      "epoch": 89.15,
+      "learning_rate": 5.7067307692307696e-05,
+      "loss": 0.1676,
+      "step": 9717
+    },
+    {
+      "epoch": 89.16,
+      "learning_rate": 5.701923076923077e-05,
+      "loss": 0.1368,
+      "step": 9718
+    },
+    {
+      "epoch": 89.17,
+      "learning_rate": 5.697115384615385e-05,
+      "loss": 0.1339,
+      "step": 9719
+    },
+    {
+      "epoch": 89.17,
+      "learning_rate": 5.692307692307693e-05,
+      "loss": 0.0582,
+      "step": 9720
+    },
+    {
+      "epoch": 89.18,
+      "learning_rate": 5.6875e-05,
+      "loss": 0.1489,
+      "step": 9721
+    },
+    {
+      "epoch": 89.19,
+      "learning_rate": 5.682692307692307e-05,
+      "loss": 0.1139,
+      "step": 9722
+    },
+    {
+      "epoch": 89.2,
+      "learning_rate": 5.677884615384615e-05,
+      "loss": 0.1069,
+      "step": 9723
+    },
+    {
+      "epoch": 89.21,
+      "learning_rate": 5.673076923076923e-05,
+      "loss": 0.0741,
+      "step": 9724
+    },
+    {
+      "epoch": 89.22,
+      "learning_rate": 5.668269230769231e-05,
+      "loss": 0.1072,
+      "step": 9725
+    },
+    {
+      "epoch": 89.23,
+      "learning_rate": 5.663461538461539e-05,
+      "loss": 0.1312,
+      "step": 9726
+    },
+    {
+      "epoch": 89.24,
+      "learning_rate": 5.658653846153846e-05,
+      "loss": 0.088,
+      "step": 9727
+    },
+    {
+      "epoch": 89.25,
+      "learning_rate": 5.6538461538461536e-05,
+      "loss": 0.106,
+      "step": 9728
+    },
+    {
+      "epoch": 89.26,
+      "learning_rate": 5.6490384615384615e-05,
+      "loss": 0.1871,
+      "step": 9729
+    },
+    {
+      "epoch": 89.27,
+      "learning_rate": 5.6442307692307694e-05,
+      "loss": 0.1549,
+      "step": 9730
+    },
+    {
+      "epoch": 89.28,
+      "learning_rate": 5.639423076923077e-05,
+      "loss": 0.1526,
+      "step": 9731
+    },
+    {
+      "epoch": 89.28,
+      "learning_rate": 5.6346153846153846e-05,
+      "loss": 0.1616,
+      "step": 9732
+    },
+    {
+      "epoch": 89.29,
+      "learning_rate": 5.6298076923076926e-05,
+      "loss": 0.143,
+      "step": 9733
+    },
+    {
+      "epoch": 89.3,
+      "learning_rate": 5.6250000000000005e-05,
+      "loss": 0.1825,
+      "step": 9734
+    },
+    {
+      "epoch": 89.31,
+      "learning_rate": 5.620192307692308e-05,
+      "loss": 0.1565,
+      "step": 9735
+    },
+    {
+      "epoch": 89.32,
+      "learning_rate": 5.615384615384615e-05,
+      "loss": 0.1431,
+      "step": 9736
+    },
+    {
+      "epoch": 89.33,
+      "learning_rate": 5.610576923076923e-05,
+      "loss": 0.1273,
+      "step": 9737
+    },
+    {
+      "epoch": 89.34,
+      "learning_rate": 5.605769230769231e-05,
+      "loss": 0.135,
+      "step": 9738
+    },
+    {
+      "epoch": 89.35,
+      "learning_rate": 5.600961538461539e-05,
+      "loss": 0.0929,
+      "step": 9739
+    },
+    {
+      "epoch": 89.36,
+      "learning_rate": 5.596153846153846e-05,
+      "loss": 0.1005,
+      "step": 9740
+    },
+    {
+      "epoch": 89.37,
+      "learning_rate": 5.591346153846154e-05,
+      "loss": 0.1332,
+      "step": 9741
+    },
+    {
+      "epoch": 89.38,
+      "learning_rate": 5.586538461538461e-05,
+      "loss": 0.1559,
+      "step": 9742
+    },
+    {
+      "epoch": 89.39,
+      "learning_rate": 5.581730769230769e-05,
+      "loss": 0.0833,
+      "step": 9743
+    },
+    {
+      "epoch": 89.39,
+      "learning_rate": 5.576923076923077e-05,
+      "loss": 0.0879,
+      "step": 9744
+    },
+    {
+      "epoch": 89.4,
+      "learning_rate": 5.5721153846153845e-05,
+      "loss": 0.2114,
+      "step": 9745
+    },
+    {
+      "epoch": 89.41,
+      "learning_rate": 5.5673076923076924e-05,
+      "loss": 0.1211,
+      "step": 9746
+    },
+    {
+      "epoch": 89.42,
+      "learning_rate": 5.5625000000000004e-05,
+      "loss": 0.1303,
+      "step": 9747
+    },
+    {
+      "epoch": 89.43,
+      "learning_rate": 5.557692307692308e-05,
+      "loss": 0.1355,
+      "step": 9748
+    },
+    {
+      "epoch": 89.44,
+      "learning_rate": 5.552884615384615e-05,
+      "loss": 0.1173,
+      "step": 9749
+    },
+    {
+      "epoch": 89.45,
+      "learning_rate": 5.548076923076923e-05,
+      "loss": 0.089,
+      "step": 9750
+    },
+    {
+      "epoch": 89.46,
+      "learning_rate": 5.543269230769231e-05,
+      "loss": 0.0885,
+      "step": 9751
+    },
+    {
+      "epoch": 89.47,
+      "learning_rate": 5.538461538461539e-05,
+      "loss": 0.0582,
+      "step": 9752
+    },
+    {
+      "epoch": 89.48,
+      "learning_rate": 5.5336538461538466e-05,
+      "loss": 0.1143,
+      "step": 9753
+    },
+    {
+      "epoch": 89.49,
+      "learning_rate": 5.528846153846154e-05,
+      "loss": 0.115,
+      "step": 9754
+    },
+    {
+      "epoch": 89.5,
+      "learning_rate": 5.524038461538462e-05,
+      "loss": 0.0613,
+      "step": 9755
+    },
+    {
+      "epoch": 89.5,
+      "learning_rate": 5.519230769230769e-05,
+      "loss": 0.2199,
+      "step": 9756
+    },
+    {
+      "epoch": 89.51,
+      "learning_rate": 5.514423076923077e-05,
+      "loss": 0.1282,
+      "step": 9757
+    },
+    {
+      "epoch": 89.52,
+      "learning_rate": 5.509615384615384e-05,
+      "loss": 0.1569,
+      "step": 9758
+    },
+    {
+      "epoch": 89.53,
+      "learning_rate": 5.504807692307692e-05,
+      "loss": 0.1601,
+      "step": 9759
+    },
+    {
+      "epoch": 89.54,
+      "learning_rate": 5.5e-05,
+      "loss": 0.1883,
+      "step": 9760
+    },
+    {
+      "epoch": 89.55,
+      "learning_rate": 5.495192307692308e-05,
+      "loss": 0.0927,
+      "step": 9761
+    },
+    {
+      "epoch": 89.56,
+      "learning_rate": 5.490384615384616e-05,
+      "loss": 0.1647,
+      "step": 9762
+    },
+    {
+      "epoch": 89.57,
+      "learning_rate": 5.485576923076923e-05,
+      "loss": 0.0911,
+      "step": 9763
+    },
+    {
+      "epoch": 89.58,
+      "learning_rate": 5.4807692307692306e-05,
+      "loss": 0.1371,
+      "step": 9764
+    },
+    {
+      "epoch": 89.59,
+      "learning_rate": 5.4759615384615385e-05,
+      "loss": 0.1016,
+      "step": 9765
+    },
+    {
+      "epoch": 89.6,
+      "learning_rate": 5.4711538461538465e-05,
+      "loss": 0.1837,
+      "step": 9766
+    },
+    {
+      "epoch": 89.61,
+      "learning_rate": 5.466346153846154e-05,
+      "loss": 0.1072,
+      "step": 9767
+    },
+    {
+      "epoch": 89.61,
+      "learning_rate": 5.461538461538462e-05,
+      "loss": 0.123,
+      "step": 9768
+    },
+    {
+      "epoch": 89.62,
+      "learning_rate": 5.4567307692307696e-05,
+      "loss": 0.1259,
+      "step": 9769
+    },
+    {
+      "epoch": 89.63,
+      "learning_rate": 5.4519230769230776e-05,
+      "loss": 0.0933,
+      "step": 9770
+    },
+    {
+      "epoch": 89.64,
+      "learning_rate": 5.447115384615384e-05,
+      "loss": 0.0928,
+      "step": 9771
+    },
+    {
+      "epoch": 89.65,
+      "learning_rate": 5.442307692307692e-05,
+      "loss": 0.1317,
+      "step": 9772
+    },
+    {
+      "epoch": 89.66,
+      "learning_rate": 5.4375e-05,
+      "loss": 0.1437,
+      "step": 9773
+    },
+    {
+      "epoch": 89.67,
+      "learning_rate": 5.432692307692308e-05,
+      "loss": 0.1234,
+      "step": 9774
+    },
+    {
+      "epoch": 89.68,
+      "learning_rate": 5.427884615384616e-05,
+      "loss": 0.1867,
+      "step": 9775
+    },
+    {
+      "epoch": 89.69,
+      "learning_rate": 5.423076923076923e-05,
+      "loss": 0.0883,
+      "step": 9776
+    },
+    {
+      "epoch": 89.7,
+      "learning_rate": 5.418269230769231e-05,
+      "loss": 0.1397,
+      "step": 9777
+    },
+    {
+      "epoch": 89.71,
+      "learning_rate": 5.4134615384615384e-05,
+      "loss": 0.0901,
+      "step": 9778
+    },
+    {
+      "epoch": 89.72,
+      "learning_rate": 5.408653846153846e-05,
+      "loss": 0.1321,
+      "step": 9779
+    },
+    {
+      "epoch": 89.72,
+      "learning_rate": 5.4038461538461536e-05,
+      "loss": 0.105,
+      "step": 9780
+    },
+    {
+      "epoch": 89.73,
+      "learning_rate": 5.3990384615384615e-05,
+      "loss": 0.1553,
+      "step": 9781
+    },
+    {
+      "epoch": 89.74,
+      "learning_rate": 5.3942307692307695e-05,
+      "loss": 0.0378,
+      "step": 9782
+    },
+    {
+      "epoch": 89.75,
+      "learning_rate": 5.3894230769230774e-05,
+      "loss": 0.1474,
+      "step": 9783
+    },
+    {
+      "epoch": 89.76,
+      "learning_rate": 5.3846153846153853e-05,
+      "loss": 0.125,
+      "step": 9784
+    },
+    {
+      "epoch": 89.77,
+      "learning_rate": 5.379807692307692e-05,
+      "loss": 0.1573,
+      "step": 9785
+    },
+    {
+      "epoch": 89.78,
+      "learning_rate": 5.375e-05,
+      "loss": 0.1226,
+      "step": 9786
+    },
+    {
+      "epoch": 89.79,
+      "learning_rate": 5.370192307692308e-05,
+      "loss": 0.1394,
+      "step": 9787
+    },
+    {
+      "epoch": 89.8,
+      "learning_rate": 5.365384615384616e-05,
+      "loss": 0.1614,
+      "step": 9788
+    },
+    {
+      "epoch": 89.81,
+      "learning_rate": 5.360576923076923e-05,
+      "loss": 0.1592,
+      "step": 9789
+    },
+    {
+      "epoch": 89.82,
+      "learning_rate": 5.355769230769231e-05,
+      "loss": 0.1935,
+      "step": 9790
+    },
+    {
+      "epoch": 89.83,
+      "learning_rate": 5.350961538461539e-05,
+      "loss": 0.0978,
+      "step": 9791
+    },
+    {
+      "epoch": 89.83,
+      "learning_rate": 5.346153846153846e-05,
+      "loss": 0.1983,
+      "step": 9792
+    },
+    {
+      "epoch": 89.84,
+      "learning_rate": 5.341346153846154e-05,
+      "loss": 0.1433,
+      "step": 9793
+    },
+    {
+      "epoch": 89.85,
+      "learning_rate": 5.3365384615384614e-05,
+      "loss": 0.1347,
+      "step": 9794
+    },
+    {
+      "epoch": 89.86,
+      "learning_rate": 5.331730769230769e-05,
+      "loss": 0.1119,
+      "step": 9795
+    },
+    {
+      "epoch": 89.87,
+      "learning_rate": 5.326923076923077e-05,
+      "loss": 0.202,
+      "step": 9796
+    },
+    {
+      "epoch": 89.88,
+      "learning_rate": 5.322115384615385e-05,
+      "loss": 0.1133,
+      "step": 9797
+    },
+    {
+      "epoch": 89.89,
+      "learning_rate": 5.3173076923076924e-05,
+      "loss": 0.0958,
+      "step": 9798
+    },
+    {
+      "epoch": 89.9,
+      "learning_rate": 5.3125e-05,
+      "loss": 0.08,
+      "step": 9799
+    },
+    {
+      "epoch": 89.91,
+      "learning_rate": 5.3076923076923076e-05,
+      "loss": 0.1898,
+      "step": 9800
+    },
+    {
+      "epoch": 89.92,
+      "learning_rate": 5.3028846153846156e-05,
+      "loss": 0.0953,
+      "step": 9801
+    },
+    {
+      "epoch": 89.93,
+      "learning_rate": 5.2980769230769235e-05,
+      "loss": 0.1002,
+      "step": 9802
+    },
+    {
+      "epoch": 89.94,
+      "learning_rate": 5.293269230769231e-05,
+      "loss": 0.1555,
+      "step": 9803
+    },
+    {
+      "epoch": 89.94,
+      "learning_rate": 5.288461538461539e-05,
+      "loss": 0.0525,
+      "step": 9804
+    },
+    {
+      "epoch": 89.95,
+      "learning_rate": 5.283653846153847e-05,
+      "loss": 0.1208,
+      "step": 9805
+    },
+    {
+      "epoch": 89.96,
+      "learning_rate": 5.278846153846154e-05,
+      "loss": 0.0816,
+      "step": 9806
+    },
+    {
+      "epoch": 89.97,
+      "learning_rate": 5.274038461538461e-05,
+      "loss": 0.0684,
+      "step": 9807
+    },
+    {
+      "epoch": 89.98,
+      "learning_rate": 5.269230769230769e-05,
+      "loss": 0.1235,
+      "step": 9808
+    },
+    {
+      "epoch": 89.99,
+      "learning_rate": 5.264423076923077e-05,
+      "loss": 0.0354,
+      "step": 9809
+    },
+    {
+      "epoch": 90.0,
+      "learning_rate": 5.259615384615385e-05,
+      "loss": 0.1999,
+      "step": 9810
+    },
+    {
+      "epoch": 90.01,
+      "learning_rate": 5.254807692307693e-05,
+      "loss": 0.1737,
+      "step": 9811
+    },
+    {
+      "epoch": 90.02,
+      "learning_rate": 5.25e-05,
+      "loss": 0.18,
+      "step": 9812
+    },
+    {
+      "epoch": 90.03,
+      "learning_rate": 5.2451923076923075e-05,
+      "loss": 0.1485,
+      "step": 9813
+    },
+    {
+      "epoch": 90.04,
+      "learning_rate": 5.2403846153846154e-05,
+      "loss": 0.098,
+      "step": 9814
+    },
+    {
+      "epoch": 90.05,
+      "learning_rate": 5.2355769230769234e-05,
+      "loss": 0.1276,
+      "step": 9815
+    },
+    {
+      "epoch": 90.06,
+      "learning_rate": 5.2307692307692306e-05,
+      "loss": 0.173,
+      "step": 9816
+    },
+    {
+      "epoch": 90.06,
+      "learning_rate": 5.2259615384615386e-05,
+      "loss": 0.0743,
+      "step": 9817
+    },
+    {
+      "epoch": 90.07,
+      "learning_rate": 5.2211538461538465e-05,
+      "loss": 0.1333,
+      "step": 9818
+    },
+    {
+      "epoch": 90.08,
+      "learning_rate": 5.2163461538461544e-05,
+      "loss": 0.1347,
+      "step": 9819
+    },
+    {
+      "epoch": 90.09,
+      "learning_rate": 5.211538461538461e-05,
+      "loss": 0.2129,
+      "step": 9820
+    },
+    {
+      "epoch": 90.1,
+      "learning_rate": 5.206730769230769e-05,
+      "loss": 0.0946,
+      "step": 9821
+    },
+    {
+      "epoch": 90.11,
+      "learning_rate": 5.201923076923077e-05,
+      "loss": 0.1498,
+      "step": 9822
+    },
+    {
+      "epoch": 90.12,
+      "learning_rate": 5.197115384615385e-05,
+      "loss": 0.1288,
+      "step": 9823
+    },
+    {
+      "epoch": 90.13,
+      "learning_rate": 5.192307692307693e-05,
+      "loss": 0.1914,
+      "step": 9824
+    },
+    {
+      "epoch": 90.14,
+      "learning_rate": 5.1875e-05,
+      "loss": 0.1416,
+      "step": 9825
+    },
+    {
+      "epoch": 90.15,
+      "learning_rate": 5.182692307692308e-05,
+      "loss": 0.1143,
+      "step": 9826
+    },
+    {
+      "epoch": 90.16,
+      "learning_rate": 5.177884615384615e-05,
+      "loss": 0.0889,
+      "step": 9827
+    },
+    {
+      "epoch": 90.17,
+      "learning_rate": 5.173076923076923e-05,
+      "loss": 0.1334,
+      "step": 9828
+    },
+    {
+      "epoch": 90.17,
+      "learning_rate": 5.1682692307692305e-05,
+      "loss": 0.1033,
+      "step": 9829
+    },
+    {
+      "epoch": 90.18,
+      "learning_rate": 5.1634615384615384e-05,
+      "loss": 0.1072,
+      "step": 9830
+    },
+    {
+      "epoch": 90.19,
+      "learning_rate": 5.1586538461538463e-05,
+      "loss": 0.0957,
+      "step": 9831
+    },
+    {
+      "epoch": 90.2,
+      "learning_rate": 5.153846153846154e-05,
+      "loss": 0.1315,
+      "step": 9832
+    },
+    {
+      "epoch": 90.21,
+      "learning_rate": 5.149038461538462e-05,
+      "loss": 0.1086,
+      "step": 9833
+    },
+    {
+      "epoch": 90.22,
+      "learning_rate": 5.144230769230769e-05,
+      "loss": 0.0785,
+      "step": 9834
+    },
+    {
+      "epoch": 90.23,
+      "learning_rate": 5.139423076923077e-05,
+      "loss": 0.078,
+      "step": 9835
+    },
+    {
+      "epoch": 90.24,
+      "learning_rate": 5.134615384615385e-05,
+      "loss": 0.1277,
+      "step": 9836
+    },
+    {
+      "epoch": 90.25,
+      "learning_rate": 5.1298076923076926e-05,
+      "loss": 0.0542,
+      "step": 9837
+    },
+    {
+      "epoch": 90.26,
+      "learning_rate": 5.125e-05,
+      "loss": 0.1854,
+      "step": 9838
+    },
+    {
+      "epoch": 90.27,
+      "learning_rate": 5.120192307692308e-05,
+      "loss": 0.1379,
+      "step": 9839
+    },
+    {
+      "epoch": 90.28,
+      "learning_rate": 5.115384615384616e-05,
+      "loss": 0.1413,
+      "step": 9840
+    },
+    {
+      "epoch": 90.28,
+      "learning_rate": 5.110576923076923e-05,
+      "loss": 0.129,
+      "step": 9841
+    },
+    {
+      "epoch": 90.29,
+      "learning_rate": 5.105769230769231e-05,
+      "loss": 0.0942,
+      "step": 9842
+    },
+    {
+      "epoch": 90.3,
+      "learning_rate": 5.100961538461538e-05,
+      "loss": 0.1423,
+      "step": 9843
+    },
+    {
+      "epoch": 90.31,
+      "learning_rate": 5.096153846153846e-05,
+      "loss": 0.1141,
+      "step": 9844
+    },
+    {
+      "epoch": 90.32,
+      "learning_rate": 5.091346153846154e-05,
+      "loss": 0.1463,
+      "step": 9845
+    },
+    {
+      "epoch": 90.33,
+      "learning_rate": 5.086538461538462e-05,
+      "loss": 0.1305,
+      "step": 9846
+    },
+    {
+      "epoch": 90.34,
+      "learning_rate": 5.081730769230769e-05,
+      "loss": 0.0752,
+      "step": 9847
+    },
+    {
+      "epoch": 90.35,
+      "learning_rate": 5.0769230769230766e-05,
+      "loss": 0.1499,
+      "step": 9848
+    },
+    {
+      "epoch": 90.36,
+      "learning_rate": 5.0721153846153845e-05,
+      "loss": 0.1006,
+      "step": 9849
+    },
+    {
+      "epoch": 90.37,
+      "learning_rate": 5.0673076923076925e-05,
+      "loss": 0.1355,
+      "step": 9850
+    },
+    {
+      "epoch": 90.38,
+      "learning_rate": 5.0625000000000004e-05,
+      "loss": 0.1561,
+      "step": 9851
+    },
+    {
+      "epoch": 90.39,
+      "learning_rate": 5.057692307692308e-05,
+      "loss": 0.0938,
+      "step": 9852
+    },
+    {
+      "epoch": 90.39,
+      "learning_rate": 5.0528846153846156e-05,
+      "loss": 0.1614,
+      "step": 9853
+    },
+    {
+      "epoch": 90.4,
+      "learning_rate": 5.0480769230769235e-05,
+      "loss": 0.1453,
+      "step": 9854
+    },
+    {
+      "epoch": 90.41,
+      "learning_rate": 5.043269230769231e-05,
+      "loss": 0.1567,
+      "step": 9855
+    },
+    {
+      "epoch": 90.42,
+      "learning_rate": 5.038461538461538e-05,
+      "loss": 0.1057,
+      "step": 9856
+    },
+    {
+      "epoch": 90.43,
+      "learning_rate": 5.033653846153846e-05,
+      "loss": 0.1242,
+      "step": 9857
+    },
+    {
+      "epoch": 90.44,
+      "learning_rate": 5.028846153846154e-05,
+      "loss": 0.093,
+      "step": 9858
+    },
+    {
+      "epoch": 90.45,
+      "learning_rate": 5.024038461538462e-05,
+      "loss": 0.098,
+      "step": 9859
+    },
+    {
+      "epoch": 90.46,
+      "learning_rate": 5.019230769230769e-05,
+      "loss": 0.1377,
+      "step": 9860
+    },
+    {
+      "epoch": 90.47,
+      "learning_rate": 5.014423076923077e-05,
+      "loss": 0.07,
+      "step": 9861
+    },
+    {
+      "epoch": 90.48,
+      "learning_rate": 5.0096153846153844e-05,
+      "loss": 0.0739,
+      "step": 9862
+    },
+    {
+      "epoch": 90.49,
+      "learning_rate": 5.004807692307692e-05,
+      "loss": 0.114,
+      "step": 9863
+    },
+    {
+      "epoch": 90.5,
+      "learning_rate": 5e-05,
+      "loss": 0.093,
+      "step": 9864
+    },
+    {
+      "epoch": 90.5,
+      "learning_rate": 4.9951923076923075e-05,
+      "loss": 0.1498,
+      "step": 9865
+    },
+    {
+      "epoch": 90.51,
+      "learning_rate": 4.9903846153846154e-05,
+      "loss": 0.1496,
+      "step": 9866
+    },
+    {
+      "epoch": 90.52,
+      "learning_rate": 4.9855769230769234e-05,
+      "loss": 0.1897,
+      "step": 9867
+    },
+    {
+      "epoch": 90.53,
+      "learning_rate": 4.980769230769231e-05,
+      "loss": 0.0937,
+      "step": 9868
+    },
+    {
+      "epoch": 90.54,
+      "learning_rate": 4.975961538461538e-05,
+      "loss": 0.1042,
+      "step": 9869
+    },
+    {
+      "epoch": 90.55,
+      "learning_rate": 4.971153846153846e-05,
+      "loss": 0.1894,
+      "step": 9870
+    },
+    {
+      "epoch": 90.56,
+      "learning_rate": 4.966346153846154e-05,
+      "loss": 0.0946,
+      "step": 9871
+    },
+    {
+      "epoch": 90.57,
+      "learning_rate": 4.961538461538462e-05,
+      "loss": 0.1553,
+      "step": 9872
+    },
+    {
+      "epoch": 90.58,
+      "learning_rate": 4.95673076923077e-05,
+      "loss": 0.1088,
+      "step": 9873
+    },
+    {
+      "epoch": 90.59,
+      "learning_rate": 4.951923076923077e-05,
+      "loss": 0.0901,
+      "step": 9874
+    },
+    {
+      "epoch": 90.6,
+      "learning_rate": 4.947115384615385e-05,
+      "loss": 0.1912,
+      "step": 9875
+    },
+    {
+      "epoch": 90.61,
+      "learning_rate": 4.942307692307693e-05,
+      "loss": 0.1174,
+      "step": 9876
+    },
+    {
+      "epoch": 90.61,
+      "learning_rate": 4.9375e-05,
+      "loss": 0.1119,
+      "step": 9877
+    },
+    {
+      "epoch": 90.62,
+      "learning_rate": 4.932692307692307e-05,
+      "loss": 0.2073,
+      "step": 9878
+    },
+    {
+      "epoch": 90.63,
+      "learning_rate": 4.927884615384615e-05,
+      "loss": 0.1458,
+      "step": 9879
+    },
+    {
+      "epoch": 90.64,
+      "learning_rate": 4.923076923076923e-05,
+      "loss": 0.0812,
+      "step": 9880
+    },
+    {
+      "epoch": 90.65,
+      "learning_rate": 4.918269230769231e-05,
+      "loss": 0.1421,
+      "step": 9881
+    },
+    {
+      "epoch": 90.66,
+      "learning_rate": 4.913461538461539e-05,
+      "loss": 0.0796,
+      "step": 9882
+    },
+    {
+      "epoch": 90.67,
+      "learning_rate": 4.9086538461538464e-05,
+      "loss": 0.1193,
+      "step": 9883
+    },
+    {
+      "epoch": 90.68,
+      "learning_rate": 4.9038461538461536e-05,
+      "loss": 0.0977,
+      "step": 9884
+    },
+    {
+      "epoch": 90.69,
+      "learning_rate": 4.8990384615384616e-05,
+      "loss": 0.0777,
+      "step": 9885
+    },
+    {
+      "epoch": 90.7,
+      "learning_rate": 4.8942307692307695e-05,
+      "loss": 0.1303,
+      "step": 9886
+    },
+    {
+      "epoch": 90.71,
+      "learning_rate": 4.889423076923077e-05,
+      "loss": 0.033,
+      "step": 9887
+    },
+    {
+      "epoch": 90.72,
+      "learning_rate": 4.884615384615385e-05,
+      "loss": 0.1477,
+      "step": 9888
+    },
+    {
+      "epoch": 90.72,
+      "learning_rate": 4.8798076923076926e-05,
+      "loss": 0.1071,
+      "step": 9889
+    },
+    {
+      "epoch": 90.73,
+      "learning_rate": 4.8750000000000006e-05,
+      "loss": 0.1085,
+      "step": 9890
+    },
+    {
+      "epoch": 90.74,
+      "learning_rate": 4.870192307692308e-05,
+      "loss": 0.0329,
+      "step": 9891
+    },
+    {
+      "epoch": 90.75,
+      "learning_rate": 4.865384615384615e-05,
+      "loss": 0.1682,
+      "step": 9892
+    },
+    {
+      "epoch": 90.76,
+      "learning_rate": 4.860576923076923e-05,
+      "loss": 0.1437,
+      "step": 9893
+    },
+    {
+      "epoch": 90.77,
+      "learning_rate": 4.855769230769231e-05,
+      "loss": 0.1518,
+      "step": 9894
+    },
+    {
+      "epoch": 90.78,
+      "learning_rate": 4.850961538461539e-05,
+      "loss": 0.1181,
+      "step": 9895
+    },
+    {
+      "epoch": 90.79,
+      "learning_rate": 4.846153846153846e-05,
+      "loss": 0.1648,
+      "step": 9896
+    },
+    {
+      "epoch": 90.8,
+      "learning_rate": 4.841346153846154e-05,
+      "loss": 0.1735,
+      "step": 9897
+    },
+    {
+      "epoch": 90.81,
+      "learning_rate": 4.8365384615384614e-05,
+      "loss": 0.1026,
+      "step": 9898
+    },
+    {
+      "epoch": 90.82,
+      "learning_rate": 4.8317307692307693e-05,
+      "loss": 0.1823,
+      "step": 9899
+    },
+    {
+      "epoch": 90.83,
+      "learning_rate": 4.826923076923077e-05,
+      "loss": 0.121,
+      "step": 9900
+    },
+    {
+      "epoch": 90.83,
+      "learning_rate": 4.8221153846153845e-05,
+      "loss": 0.0914,
+      "step": 9901
+    },
+    {
+      "epoch": 90.84,
+      "learning_rate": 4.8173076923076925e-05,
+      "loss": 0.1135,
+      "step": 9902
+    },
+    {
+      "epoch": 90.85,
+      "learning_rate": 4.8125000000000004e-05,
+      "loss": 0.1526,
+      "step": 9903
+    },
+    {
+      "epoch": 90.86,
+      "learning_rate": 4.8076923076923084e-05,
+      "loss": 0.1484,
+      "step": 9904
+    },
+    {
+      "epoch": 90.87,
+      "learning_rate": 4.802884615384615e-05,
+      "loss": 0.1109,
+      "step": 9905
+    },
+    {
+      "epoch": 90.88,
+      "learning_rate": 4.798076923076923e-05,
+      "loss": 0.0889,
+      "step": 9906
+    },
+    {
+      "epoch": 90.89,
+      "learning_rate": 4.793269230769231e-05,
+      "loss": 0.1643,
+      "step": 9907
+    },
+    {
+      "epoch": 90.9,
+      "learning_rate": 4.788461538461539e-05,
+      "loss": 0.1107,
+      "step": 9908
+    },
+    {
+      "epoch": 90.91,
+      "learning_rate": 4.783653846153846e-05,
+      "loss": 0.1836,
+      "step": 9909
+    },
+    {
+      "epoch": 90.92,
+      "learning_rate": 4.778846153846154e-05,
+      "loss": 0.1242,
+      "step": 9910
+    },
+    {
+      "epoch": 90.93,
+      "learning_rate": 4.774038461538462e-05,
+      "loss": 0.1339,
+      "step": 9911
+    },
+    {
+      "epoch": 90.94,
+      "learning_rate": 4.769230769230769e-05,
+      "loss": 0.1423,
+      "step": 9912
+    },
+    {
+      "epoch": 90.94,
+      "learning_rate": 4.764423076923077e-05,
+      "loss": 0.0505,
+      "step": 9913
+    },
+    {
+      "epoch": 90.95,
+      "learning_rate": 4.7596153846153844e-05,
+      "loss": 0.0823,
+      "step": 9914
+    },
+    {
+      "epoch": 90.96,
+      "learning_rate": 4.754807692307692e-05,
+      "loss": 0.1505,
+      "step": 9915
+    },
+    {
+      "epoch": 90.97,
+      "learning_rate": 4.75e-05,
+      "loss": 0.2432,
+      "step": 9916
+    },
+    {
+      "epoch": 90.98,
+      "learning_rate": 4.745192307692308e-05,
+      "loss": 0.0943,
+      "step": 9917
+    },
+    {
+      "epoch": 90.99,
+      "learning_rate": 4.7403846153846155e-05,
+      "loss": 0.1046,
+      "step": 9918
+    },
+    {
+      "epoch": 91.0,
+      "learning_rate": 4.735576923076923e-05,
+      "loss": 0.0959,
+      "step": 9919
+    },
+    {
+      "epoch": 91.01,
+      "learning_rate": 4.730769230769231e-05,
+      "loss": 0.1327,
+      "step": 9920
+    },
+    {
+      "epoch": 91.02,
+      "learning_rate": 4.7259615384615386e-05,
+      "loss": 0.1083,
+      "step": 9921
+    },
+    {
+      "epoch": 91.03,
+      "learning_rate": 4.7211538461538465e-05,
+      "loss": 0.1516,
+      "step": 9922
+    },
+    {
+      "epoch": 91.04,
+      "learning_rate": 4.716346153846154e-05,
+      "loss": 0.1293,
+      "step": 9923
+    },
+    {
+      "epoch": 91.05,
+      "learning_rate": 4.711538461538462e-05,
+      "loss": 0.1287,
+      "step": 9924
+    },
+    {
+      "epoch": 91.06,
+      "learning_rate": 4.70673076923077e-05,
+      "loss": 0.1185,
+      "step": 9925
+    },
+    {
+      "epoch": 91.06,
+      "learning_rate": 4.701923076923077e-05,
+      "loss": 0.1244,
+      "step": 9926
+    },
+    {
+      "epoch": 91.07,
+      "learning_rate": 4.697115384615384e-05,
+      "loss": 0.1728,
+      "step": 9927
+    },
+    {
+      "epoch": 91.08,
+      "learning_rate": 4.692307692307692e-05,
+      "loss": 0.1014,
+      "step": 9928
+    },
+    {
+      "epoch": 91.09,
+      "learning_rate": 4.6875e-05,
+      "loss": 0.1232,
+      "step": 9929
+    },
+    {
+      "epoch": 91.1,
+      "learning_rate": 4.682692307692308e-05,
+      "loss": 0.1076,
+      "step": 9930
+    },
+    {
+      "epoch": 91.11,
+      "learning_rate": 4.677884615384616e-05,
+      "loss": 0.1076,
+      "step": 9931
+    },
+    {
+      "epoch": 91.12,
+      "learning_rate": 4.673076923076923e-05,
+      "loss": 0.1171,
+      "step": 9932
+    },
+    {
+      "epoch": 91.13,
+      "learning_rate": 4.6682692307692305e-05,
+      "loss": 0.1216,
+      "step": 9933
+    },
+    {
+      "epoch": 91.14,
+      "learning_rate": 4.6634615384615384e-05,
+      "loss": 0.0811,
+      "step": 9934
+    },
+    {
+      "epoch": 91.15,
+      "learning_rate": 4.6586538461538464e-05,
+      "loss": 0.1514,
+      "step": 9935
+    },
+    {
+      "epoch": 91.16,
+      "learning_rate": 4.6538461538461536e-05,
+      "loss": 0.1293,
+      "step": 9936
+    },
+    {
+      "epoch": 91.17,
+      "learning_rate": 4.6490384615384616e-05,
+      "loss": 0.2084,
+      "step": 9937
+    },
+    {
+      "epoch": 91.17,
+      "learning_rate": 4.6442307692307695e-05,
+      "loss": 0.1065,
+      "step": 9938
+    },
+    {
+      "epoch": 91.18,
+      "learning_rate": 4.6394230769230775e-05,
+      "loss": 0.098,
+      "step": 9939
+    },
+    {
+      "epoch": 91.19,
+      "learning_rate": 4.634615384615385e-05,
+      "loss": 0.0706,
+      "step": 9940
+    },
+    {
+      "epoch": 91.2,
+      "learning_rate": 4.629807692307692e-05,
+      "loss": 0.1571,
+      "step": 9941
+    },
+    {
+      "epoch": 91.21,
+      "learning_rate": 4.625e-05,
+      "loss": 0.0947,
+      "step": 9942
+    },
+    {
+      "epoch": 91.22,
+      "learning_rate": 4.620192307692308e-05,
+      "loss": 0.0844,
+      "step": 9943
+    },
+    {
+      "epoch": 91.23,
+      "learning_rate": 4.615384615384616e-05,
+      "loss": 0.14,
+      "step": 9944
+    },
+    {
+      "epoch": 91.24,
+      "learning_rate": 4.610576923076923e-05,
+      "loss": 0.1321,
+      "step": 9945
+    },
+    {
+      "epoch": 91.25,
+      "learning_rate": 4.605769230769231e-05,
+      "loss": 0.1089,
+      "step": 9946
+    },
+    {
+      "epoch": 91.26,
+      "learning_rate": 4.600961538461538e-05,
+      "loss": 0.1677,
+      "step": 9947
+    },
+    {
+      "epoch": 91.27,
+      "learning_rate": 4.596153846153846e-05,
+      "loss": 0.1399,
+      "step": 9948
+    },
+    {
+      "epoch": 91.28,
+      "learning_rate": 4.591346153846154e-05,
+      "loss": 0.1054,
+      "step": 9949
+    },
+    {
+      "epoch": 91.28,
+      "learning_rate": 4.5865384615384614e-05,
+      "loss": 0.1598,
+      "step": 9950
+    },
+    {
+      "epoch": 91.29,
+      "learning_rate": 4.5817307692307694e-05,
+      "loss": 0.1577,
+      "step": 9951
+    },
+    {
+      "epoch": 91.3,
+      "learning_rate": 4.576923076923077e-05,
+      "loss": 0.0746,
+      "step": 9952
+    },
+    {
+      "epoch": 91.31,
+      "learning_rate": 4.572115384615385e-05,
+      "loss": 0.1525,
+      "step": 9953
+    },
+    {
+      "epoch": 91.32,
+      "learning_rate": 4.567307692307692e-05,
+      "loss": 0.1515,
+      "step": 9954
+    },
+    {
+      "epoch": 91.33,
+      "learning_rate": 4.5625e-05,
+      "loss": 0.1112,
+      "step": 9955
+    },
+    {
+      "epoch": 91.34,
+      "learning_rate": 4.557692307692308e-05,
+      "loss": 0.107,
+      "step": 9956
+    },
+    {
+      "epoch": 91.35,
+      "learning_rate": 4.5528846153846157e-05,
+      "loss": 0.0989,
+      "step": 9957
+    },
+    {
+      "epoch": 91.36,
+      "learning_rate": 4.548076923076923e-05,
+      "loss": 0.1058,
+      "step": 9958
+    },
+    {
+      "epoch": 91.37,
+      "learning_rate": 4.543269230769231e-05,
+      "loss": 0.0784,
+      "step": 9959
+    },
+    {
+      "epoch": 91.38,
+      "learning_rate": 4.538461538461539e-05,
+      "loss": 0.1256,
+      "step": 9960
+    },
+    {
+      "epoch": 91.39,
+      "learning_rate": 4.533653846153846e-05,
+      "loss": 0.0936,
+      "step": 9961
+    },
+    {
+      "epoch": 91.39,
+      "learning_rate": 4.528846153846154e-05,
+      "loss": 0.0977,
+      "step": 9962
+    },
+    {
+      "epoch": 91.4,
+      "learning_rate": 4.524038461538461e-05,
+      "loss": 0.1803,
+      "step": 9963
+    },
+    {
+      "epoch": 91.41,
+      "learning_rate": 4.519230769230769e-05,
+      "loss": 0.126,
+      "step": 9964
+    },
+    {
+      "epoch": 91.42,
+      "learning_rate": 4.514423076923077e-05,
+      "loss": 0.0568,
+      "step": 9965
+    },
+    {
+      "epoch": 91.43,
+      "learning_rate": 4.509615384615385e-05,
+      "loss": 0.0789,
+      "step": 9966
+    },
+    {
+      "epoch": 91.44,
+      "learning_rate": 4.5048076923076923e-05,
+      "loss": 0.1263,
+      "step": 9967
+    },
+    {
+      "epoch": 91.45,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 0.0876,
+      "step": 9968
+    },
+    {
+      "epoch": 91.46,
+      "learning_rate": 4.4951923076923075e-05,
+      "loss": 0.1867,
+      "step": 9969
+    },
+    {
+      "epoch": 91.47,
+      "learning_rate": 4.4903846153846155e-05,
+      "loss": 0.0565,
+      "step": 9970
+    },
+    {
+      "epoch": 91.48,
+      "learning_rate": 4.4855769230769234e-05,
+      "loss": 0.0678,
+      "step": 9971
+    },
+    {
+      "epoch": 91.49,
+      "learning_rate": 4.480769230769231e-05,
+      "loss": 0.1041,
+      "step": 9972
+    },
+    {
+      "epoch": 91.5,
+      "learning_rate": 4.4759615384615386e-05,
+      "loss": 0.1064,
+      "step": 9973
+    },
+    {
+      "epoch": 91.5,
+      "learning_rate": 4.4711538461538466e-05,
+      "loss": 0.1711,
+      "step": 9974
+    },
+    {
+      "epoch": 91.51,
+      "learning_rate": 4.466346153846154e-05,
+      "loss": 0.1847,
+      "step": 9975
+    },
+    {
+      "epoch": 91.52,
+      "learning_rate": 4.461538461538461e-05,
+      "loss": 0.1132,
+      "step": 9976
+    },
+    {
+      "epoch": 91.53,
+      "learning_rate": 4.456730769230769e-05,
+      "loss": 0.1416,
+      "step": 9977
+    },
+    {
+      "epoch": 91.54,
+      "learning_rate": 4.451923076923077e-05,
+      "loss": 0.1128,
+      "step": 9978
+    },
+    {
+      "epoch": 91.55,
+      "learning_rate": 4.447115384615385e-05,
+      "loss": 0.1505,
+      "step": 9979
+    },
+    {
+      "epoch": 91.56,
+      "learning_rate": 4.442307692307693e-05,
+      "loss": 0.1246,
+      "step": 9980
+    },
+    {
+      "epoch": 91.57,
+      "learning_rate": 4.4375e-05,
+      "loss": 0.1025,
+      "step": 9981
+    },
+    {
+      "epoch": 91.58,
+      "learning_rate": 4.4326923076923074e-05,
+      "loss": 0.1691,
+      "step": 9982
+    },
+    {
+      "epoch": 91.59,
+      "learning_rate": 4.427884615384615e-05,
+      "loss": 0.1565,
+      "step": 9983
+    },
+    {
+      "epoch": 91.6,
+      "learning_rate": 4.423076923076923e-05,
+      "loss": 0.0755,
+      "step": 9984
+    },
+    {
+      "epoch": 91.61,
+      "learning_rate": 4.4182692307692305e-05,
+      "loss": 0.139,
+      "step": 9985
+    },
+    {
+      "epoch": 91.61,
+      "learning_rate": 4.4134615384615385e-05,
+      "loss": 0.1305,
+      "step": 9986
+    },
+    {
+      "epoch": 91.62,
+      "learning_rate": 4.4086538461538464e-05,
+      "loss": 0.0951,
+      "step": 9987
+    },
+    {
+      "epoch": 91.63,
+      "learning_rate": 4.4038461538461543e-05,
+      "loss": 0.1412,
+      "step": 9988
+    },
+    {
+      "epoch": 91.64,
+      "learning_rate": 4.399038461538462e-05,
+      "loss": 0.0659,
+      "step": 9989
+    },
+    {
+      "epoch": 91.65,
+      "learning_rate": 4.394230769230769e-05,
+      "loss": 0.1257,
+      "step": 9990
+    },
+    {
+      "epoch": 91.66,
+      "learning_rate": 4.389423076923077e-05,
+      "loss": 0.1257,
+      "step": 9991
+    },
+    {
+      "epoch": 91.67,
+      "learning_rate": 4.384615384615385e-05,
+      "loss": 0.1421,
+      "step": 9992
+    },
+    {
+      "epoch": 91.68,
+      "learning_rate": 4.379807692307693e-05,
+      "loss": 0.0931,
+      "step": 9993
+    },
+    {
+      "epoch": 91.69,
+      "learning_rate": 4.375e-05,
+      "loss": 0.1296,
+      "step": 9994
+    },
+    {
+      "epoch": 91.7,
+      "learning_rate": 4.370192307692308e-05,
+      "loss": 0.0785,
+      "step": 9995
+    },
+    {
+      "epoch": 91.71,
+      "learning_rate": 4.365384615384616e-05,
+      "loss": 0.092,
+      "step": 9996
+    },
+    {
+      "epoch": 91.72,
+      "learning_rate": 4.360576923076923e-05,
+      "loss": 0.1248,
+      "step": 9997
+    },
+    {
+      "epoch": 91.72,
+      "learning_rate": 4.355769230769231e-05,
+      "loss": 0.0506,
+      "step": 9998
+    },
+    {
+      "epoch": 91.73,
+      "learning_rate": 4.350961538461538e-05,
+      "loss": 0.0947,
+      "step": 9999
+    },
+    {
+      "epoch": 91.74,
+      "learning_rate": 4.346153846153846e-05,
+      "loss": 0.0431,
+      "step": 10000
+    },
+    {
+      "epoch": 91.74,
+      "eval_cer": 0.13496454219703444,
+      "eval_loss": 0.7397318482398987,
+      "eval_runtime": 87.5647,
+      "eval_samples_per_second": 18.809,
+      "eval_steps_per_second": 2.353,
+      "eval_wer": 0.47492595240527014,
+      "step": 10000
+    },
+    {
+      "epoch": 91.75,
+      "learning_rate": 4.341346153846154e-05,
+      "loss": 0.2256,
+      "step": 10001
+    },
+    {
+      "epoch": 91.76,
+      "learning_rate": 4.336538461538462e-05,
+      "loss": 0.1548,
+      "step": 10002
+    },
+    {
+      "epoch": 91.77,
+      "learning_rate": 4.3317307692307694e-05,
+      "loss": 0.1369,
+      "step": 10003
+    },
+    {
+      "epoch": 91.78,
+      "learning_rate": 4.3269230769230766e-05,
+      "loss": 0.1429,
+      "step": 10004
+    },
+    {
+      "epoch": 91.79,
+      "learning_rate": 4.3221153846153846e-05,
+      "loss": 0.1607,
+      "step": 10005
+    },
+    {
+      "epoch": 91.8,
+      "learning_rate": 4.3173076923076925e-05,
+      "loss": 0.1191,
+      "step": 10006
+    },
+    {
+      "epoch": 91.81,
+      "learning_rate": 4.3125e-05,
+      "loss": 0.0938,
+      "step": 10007
+    },
+    {
+      "epoch": 91.82,
+      "learning_rate": 4.307692307692308e-05,
+      "loss": 0.0931,
+      "step": 10008
+    },
+    {
+      "epoch": 91.83,
+      "learning_rate": 4.302884615384616e-05,
+      "loss": 0.0949,
+      "step": 10009
+    },
+    {
+      "epoch": 91.83,
+      "learning_rate": 4.2980769230769236e-05,
+      "loss": 0.0797,
+      "step": 10010
+    },
+    {
+      "epoch": 91.84,
+      "learning_rate": 4.293269230769231e-05,
+      "loss": 0.1022,
+      "step": 10011
+    },
+    {
+      "epoch": 91.85,
+      "learning_rate": 4.288461538461538e-05,
+      "loss": 0.101,
+      "step": 10012
+    },
+    {
+      "epoch": 91.86,
+      "learning_rate": 4.283653846153846e-05,
+      "loss": 0.1005,
+      "step": 10013
+    },
+    {
+      "epoch": 91.87,
+      "learning_rate": 4.278846153846154e-05,
+      "loss": 0.1305,
+      "step": 10014
+    },
+    {
+      "epoch": 91.88,
+      "learning_rate": 4.274038461538462e-05,
+      "loss": 0.0678,
+      "step": 10015
+    },
+    {
+      "epoch": 91.89,
+      "learning_rate": 4.269230769230769e-05,
+      "loss": 0.1345,
+      "step": 10016
+    },
+    {
+      "epoch": 91.9,
+      "learning_rate": 4.264423076923077e-05,
+      "loss": 0.1829,
+      "step": 10017
+    },
+    {
+      "epoch": 91.91,
+      "learning_rate": 4.2596153846153844e-05,
+      "loss": 0.0903,
+      "step": 10018
+    },
+    {
+      "epoch": 91.92,
+      "learning_rate": 4.2548076923076924e-05,
+      "loss": 0.0956,
+      "step": 10019
+    },
+    {
+      "epoch": 91.93,
+      "learning_rate": 4.25e-05,
+      "loss": 0.0943,
+      "step": 10020
+    },
+    {
+      "epoch": 91.94,
+      "learning_rate": 4.2451923076923076e-05,
+      "loss": 0.0818,
+      "step": 10021
+    },
+    {
+      "epoch": 91.94,
+      "learning_rate": 4.2403846153846155e-05,
+      "loss": 0.1105,
+      "step": 10022
+    },
+    {
+      "epoch": 91.95,
+      "learning_rate": 4.2355769230769234e-05,
+      "loss": 0.1118,
+      "step": 10023
+    },
+    {
+      "epoch": 91.96,
+      "learning_rate": 4.2307692307692314e-05,
+      "loss": 0.0793,
+      "step": 10024
+    },
+    {
+      "epoch": 91.97,
+      "learning_rate": 4.225961538461538e-05,
+      "loss": 0.0857,
+      "step": 10025
+    },
+    {
+      "epoch": 91.98,
+      "learning_rate": 4.221153846153846e-05,
+      "loss": 0.0949,
+      "step": 10026
+    },
+    {
+      "epoch": 91.99,
+      "learning_rate": 4.216346153846154e-05,
+      "loss": 0.1117,
+      "step": 10027
+    },
+    {
+      "epoch": 92.0,
+      "learning_rate": 4.211538461538462e-05,
+      "loss": 0.088,
+      "step": 10028
+    },
+    {
+      "epoch": 92.01,
+      "learning_rate": 4.20673076923077e-05,
+      "loss": 0.1782,
+      "step": 10029
+    },
+    {
+      "epoch": 92.02,
+      "learning_rate": 4.201923076923077e-05,
+      "loss": 0.1272,
+      "step": 10030
+    },
+    {
+      "epoch": 92.03,
+      "learning_rate": 4.197115384615385e-05,
+      "loss": 0.0972,
+      "step": 10031
+    },
+    {
+      "epoch": 92.04,
+      "learning_rate": 4.192307692307692e-05,
+      "loss": 0.123,
+      "step": 10032
+    },
+    {
+      "epoch": 92.05,
+      "learning_rate": 4.1875e-05,
+      "loss": 0.138,
+      "step": 10033
+    },
+    {
+      "epoch": 92.06,
+      "learning_rate": 4.1826923076923074e-05,
+      "loss": 0.0806,
+      "step": 10034
+    },
+    {
+      "epoch": 92.06,
+      "learning_rate": 4.1778846153846153e-05,
+      "loss": 0.1384,
+      "step": 10035
+    },
+    {
+      "epoch": 92.07,
+      "learning_rate": 4.173076923076923e-05,
+      "loss": 0.1402,
+      "step": 10036
+    },
+    {
+      "epoch": 92.08,
+      "learning_rate": 4.168269230769231e-05,
+      "loss": 0.0697,
+      "step": 10037
+    },
+    {
+      "epoch": 92.09,
+      "learning_rate": 4.163461538461539e-05,
+      "loss": 0.1551,
+      "step": 10038
+    },
+    {
+      "epoch": 92.1,
+      "learning_rate": 4.158653846153846e-05,
+      "loss": 0.1409,
+      "step": 10039
+    },
+    {
+      "epoch": 92.11,
+      "learning_rate": 4.153846153846154e-05,
+      "loss": 0.096,
+      "step": 10040
+    },
+    {
+      "epoch": 92.12,
+      "learning_rate": 4.1490384615384616e-05,
+      "loss": 0.0922,
+      "step": 10041
+    },
+    {
+      "epoch": 92.13,
+      "learning_rate": 4.1442307692307696e-05,
+      "loss": 0.1488,
+      "step": 10042
+    },
+    {
+      "epoch": 92.14,
+      "learning_rate": 4.139423076923077e-05,
+      "loss": 0.1032,
+      "step": 10043
+    },
+    {
+      "epoch": 92.15,
+      "learning_rate": 4.134615384615385e-05,
+      "loss": 0.0587,
+      "step": 10044
+    },
+    {
+      "epoch": 92.16,
+      "learning_rate": 4.129807692307693e-05,
+      "loss": 0.1065,
+      "step": 10045
+    },
+    {
+      "epoch": 92.17,
+      "learning_rate": 4.125e-05,
+      "loss": 0.1299,
+      "step": 10046
+    },
+    {
+      "epoch": 92.17,
+      "learning_rate": 4.120192307692307e-05,
+      "loss": 0.0745,
+      "step": 10047
+    },
+    {
+      "epoch": 92.18,
+      "learning_rate": 4.115384615384615e-05,
+      "loss": 0.0599,
+      "step": 10048
+    },
+    {
+      "epoch": 92.19,
+      "learning_rate": 4.110576923076923e-05,
+      "loss": 0.0755,
+      "step": 10049
+    },
+    {
+      "epoch": 92.2,
+      "learning_rate": 4.105769230769231e-05,
+      "loss": 0.1833,
+      "step": 10050
+    },
+    {
+      "epoch": 92.21,
+      "learning_rate": 4.100961538461539e-05,
+      "loss": 0.1349,
+      "step": 10051
+    },
+    {
+      "epoch": 92.22,
+      "learning_rate": 4.096153846153846e-05,
+      "loss": 0.0948,
+      "step": 10052
+    },
+    {
+      "epoch": 92.23,
+      "learning_rate": 4.0913461538461535e-05,
+      "loss": 0.0969,
+      "step": 10053
+    },
+    {
+      "epoch": 92.24,
+      "learning_rate": 4.0865384615384615e-05,
+      "loss": 0.0488,
+      "step": 10054
+    },
+    {
+      "epoch": 92.25,
+      "learning_rate": 4.0817307692307694e-05,
+      "loss": 0.0934,
+      "step": 10055
+    },
+    {
+      "epoch": 92.26,
+      "learning_rate": 4.076923076923077e-05,
+      "loss": 0.1338,
+      "step": 10056
+    },
+    {
+      "epoch": 92.27,
+      "learning_rate": 4.0721153846153846e-05,
+      "loss": 0.2041,
+      "step": 10057
+    },
+    {
+      "epoch": 92.28,
+      "learning_rate": 4.0673076923076926e-05,
+      "loss": 0.1357,
+      "step": 10058
+    },
+    {
+      "epoch": 92.28,
+      "learning_rate": 4.0625000000000005e-05,
+      "loss": 0.0849,
+      "step": 10059
+    },
+    {
+      "epoch": 92.29,
+      "learning_rate": 4.057692307692308e-05,
+      "loss": 0.1386,
+      "step": 10060
+    },
+    {
+      "epoch": 92.3,
+      "learning_rate": 4.052884615384615e-05,
+      "loss": 0.1264,
+      "step": 10061
+    },
+    {
+      "epoch": 92.31,
+      "learning_rate": 4.048076923076923e-05,
+      "loss": 0.1695,
+      "step": 10062
+    },
+    {
+      "epoch": 92.32,
+      "learning_rate": 4.043269230769231e-05,
+      "loss": 0.0698,
+      "step": 10063
+    },
+    {
+      "epoch": 92.33,
+      "learning_rate": 4.038461538461539e-05,
+      "loss": 0.1343,
+      "step": 10064
+    },
+    {
+      "epoch": 92.34,
+      "learning_rate": 4.033653846153846e-05,
+      "loss": 0.0898,
+      "step": 10065
+    },
+    {
+      "epoch": 92.35,
+      "learning_rate": 4.028846153846154e-05,
+      "loss": 0.1635,
+      "step": 10066
+    },
+    {
+      "epoch": 92.36,
+      "learning_rate": 4.024038461538461e-05,
+      "loss": 0.1123,
+      "step": 10067
+    },
+    {
+      "epoch": 92.37,
+      "learning_rate": 4.019230769230769e-05,
+      "loss": 0.1499,
+      "step": 10068
+    },
+    {
+      "epoch": 92.38,
+      "learning_rate": 4.014423076923077e-05,
+      "loss": 0.0691,
+      "step": 10069
+    },
+    {
+      "epoch": 92.39,
+      "learning_rate": 4.0096153846153844e-05,
+      "loss": 0.1313,
+      "step": 10070
+    },
+    {
+      "epoch": 92.39,
+      "learning_rate": 4.0048076923076924e-05,
+      "loss": 0.139,
+      "step": 10071
+    },
+    {
+      "epoch": 92.4,
+      "learning_rate": 4e-05,
+      "loss": 0.0635,
+      "step": 10072
+    },
+    {
+      "epoch": 92.41,
+      "learning_rate": 3.995192307692308e-05,
+      "loss": 0.0822,
+      "step": 10073
+    },
+    {
+      "epoch": 92.42,
+      "learning_rate": 3.990384615384615e-05,
+      "loss": 0.1038,
+      "step": 10074
+    },
+    {
+      "epoch": 92.43,
+      "learning_rate": 3.985576923076923e-05,
+      "loss": 0.0931,
+      "step": 10075
+    },
+    {
+      "epoch": 92.44,
+      "learning_rate": 3.980769230769231e-05,
+      "loss": 0.0668,
+      "step": 10076
+    },
+    {
+      "epoch": 92.45,
+      "learning_rate": 3.975961538461539e-05,
+      "loss": 0.1794,
+      "step": 10077
+    },
+    {
+      "epoch": 92.46,
+      "learning_rate": 3.9711538461538466e-05,
+      "loss": 0.1109,
+      "step": 10078
+    },
+    {
+      "epoch": 92.47,
+      "learning_rate": 3.966346153846154e-05,
+      "loss": 0.1428,
+      "step": 10079
+    },
+    {
+      "epoch": 92.48,
+      "learning_rate": 3.961538461538462e-05,
+      "loss": 0.0889,
+      "step": 10080
+    },
+    {
+      "epoch": 92.49,
+      "learning_rate": 3.956730769230769e-05,
+      "loss": 0.0901,
+      "step": 10081
+    },
+    {
+      "epoch": 92.5,
+      "learning_rate": 3.951923076923077e-05,
+      "loss": 0.0528,
+      "step": 10082
+    },
+    {
+      "epoch": 92.5,
+      "learning_rate": 3.947115384615384e-05,
+      "loss": 0.168,
+      "step": 10083
+    },
+    {
+      "epoch": 92.51,
+      "learning_rate": 3.942307692307692e-05,
+      "loss": 0.1396,
+      "step": 10084
+    },
+    {
+      "epoch": 92.52,
+      "learning_rate": 3.9375e-05,
+      "loss": 0.1847,
+      "step": 10085
+    },
+    {
+      "epoch": 92.53,
+      "learning_rate": 3.932692307692308e-05,
+      "loss": 0.1434,
+      "step": 10086
+    },
+    {
+      "epoch": 92.54,
+      "learning_rate": 3.927884615384616e-05,
+      "loss": 0.1746,
+      "step": 10087
+    },
+    {
+      "epoch": 92.55,
+      "learning_rate": 3.9230769230769226e-05,
+      "loss": 0.1673,
+      "step": 10088
+    },
+    {
+      "epoch": 92.56,
+      "learning_rate": 3.9182692307692306e-05,
+      "loss": 0.1708,
+      "step": 10089
+    },
+    {
+      "epoch": 92.57,
+      "learning_rate": 3.9134615384615385e-05,
+      "loss": 0.1255,
+      "step": 10090
+    },
+    {
+      "epoch": 92.58,
+      "learning_rate": 3.9086538461538464e-05,
+      "loss": 0.1071,
+      "step": 10091
+    },
+    {
+      "epoch": 92.59,
+      "learning_rate": 3.903846153846154e-05,
+      "loss": 0.0946,
+      "step": 10092
+    },
+    {
+      "epoch": 92.6,
+      "learning_rate": 3.8990384615384617e-05,
+      "loss": 0.1692,
+      "step": 10093
+    },
+    {
+      "epoch": 92.61,
+      "learning_rate": 3.8942307692307696e-05,
+      "loss": 0.0985,
+      "step": 10094
+    },
+    {
+      "epoch": 92.61,
+      "learning_rate": 3.8894230769230775e-05,
+      "loss": 0.1141,
+      "step": 10095
+    },
+    {
+      "epoch": 92.62,
+      "learning_rate": 3.884615384615384e-05,
+      "loss": 0.1071,
+      "step": 10096
+    },
+    {
+      "epoch": 92.63,
+      "learning_rate": 3.879807692307692e-05,
+      "loss": 0.0824,
+      "step": 10097
+    },
+    {
+      "epoch": 92.64,
+      "learning_rate": 3.875e-05,
+      "loss": 0.0778,
+      "step": 10098
+    },
+    {
+      "epoch": 92.65,
+      "learning_rate": 3.870192307692308e-05,
+      "loss": 0.0893,
+      "step": 10099
+    },
+    {
+      "epoch": 92.66,
+      "learning_rate": 3.865384615384616e-05,
+      "loss": 0.0844,
+      "step": 10100
+    },
+    {
+      "epoch": 92.67,
+      "learning_rate": 3.860576923076923e-05,
+      "loss": 0.1092,
+      "step": 10101
+    },
+    {
+      "epoch": 92.68,
+      "learning_rate": 3.855769230769231e-05,
+      "loss": 0.0971,
+      "step": 10102
+    },
+    {
+      "epoch": 92.69,
+      "learning_rate": 3.8509615384615383e-05,
+      "loss": 0.1,
+      "step": 10103
+    },
+    {
+      "epoch": 92.7,
+      "learning_rate": 3.846153846153846e-05,
+      "loss": 0.1118,
+      "step": 10104
+    },
+    {
+      "epoch": 92.71,
+      "learning_rate": 3.8413461538461535e-05,
+      "loss": 0.097,
+      "step": 10105
+    },
+    {
+      "epoch": 92.72,
+      "learning_rate": 3.8365384615384615e-05,
+      "loss": 0.1159,
+      "step": 10106
+    },
+    {
+      "epoch": 92.72,
+      "learning_rate": 3.8317307692307694e-05,
+      "loss": 0.104,
+      "step": 10107
+    },
+    {
+      "epoch": 92.73,
+      "learning_rate": 3.8269230769230774e-05,
+      "loss": 0.0678,
+      "step": 10108
+    },
+    {
+      "epoch": 92.74,
+      "learning_rate": 3.822115384615385e-05,
+      "loss": 0.077,
+      "step": 10109
+    },
+    {
+      "epoch": 92.75,
+      "learning_rate": 3.817307692307692e-05,
+      "loss": 0.1754,
+      "step": 10110
+    },
+    {
+      "epoch": 92.76,
+      "learning_rate": 3.8125e-05,
+      "loss": 0.127,
+      "step": 10111
+    },
+    {
+      "epoch": 92.77,
+      "learning_rate": 3.807692307692308e-05,
+      "loss": 0.1291,
+      "step": 10112
+    },
+    {
+      "epoch": 92.78,
+      "learning_rate": 3.802884615384616e-05,
+      "loss": 0.0915,
+      "step": 10113
+    },
+    {
+      "epoch": 92.79,
+      "learning_rate": 3.798076923076923e-05,
+      "loss": 0.1549,
+      "step": 10114
+    },
+    {
+      "epoch": 92.8,
+      "learning_rate": 3.793269230769231e-05,
+      "loss": 0.1068,
+      "step": 10115
+    },
+    {
+      "epoch": 92.81,
+      "learning_rate": 3.788461538461539e-05,
+      "loss": 0.1349,
+      "step": 10116
+    },
+    {
+      "epoch": 92.82,
+      "learning_rate": 3.783653846153846e-05,
+      "loss": 0.153,
+      "step": 10117
+    },
+    {
+      "epoch": 92.83,
+      "learning_rate": 3.778846153846154e-05,
+      "loss": 0.1276,
+      "step": 10118
+    },
+    {
+      "epoch": 92.83,
+      "learning_rate": 3.774038461538461e-05,
+      "loss": 0.0822,
+      "step": 10119
+    },
+    {
+      "epoch": 92.84,
+      "learning_rate": 3.769230769230769e-05,
+      "loss": 0.1227,
+      "step": 10120
+    },
+    {
+      "epoch": 92.85,
+      "learning_rate": 3.764423076923077e-05,
+      "loss": 0.1126,
+      "step": 10121
+    },
+    {
+      "epoch": 92.86,
+      "learning_rate": 3.759615384615385e-05,
+      "loss": 0.1213,
+      "step": 10122
+    },
+    {
+      "epoch": 92.87,
+      "learning_rate": 3.7548076923076924e-05,
+      "loss": 0.1143,
+      "step": 10123
+    },
+    {
+      "epoch": 92.88,
+      "learning_rate": 3.75e-05,
+      "loss": 0.1415,
+      "step": 10124
+    },
+    {
+      "epoch": 92.89,
+      "learning_rate": 3.7451923076923076e-05,
+      "loss": 0.059,
+      "step": 10125
+    },
+    {
+      "epoch": 92.9,
+      "learning_rate": 3.7403846153846156e-05,
+      "loss": 0.0518,
+      "step": 10126
+    },
+    {
+      "epoch": 92.91,
+      "learning_rate": 3.7355769230769235e-05,
+      "loss": 0.1255,
+      "step": 10127
+    },
+    {
+      "epoch": 92.92,
+      "learning_rate": 3.730769230769231e-05,
+      "loss": 0.1182,
+      "step": 10128
+    },
+    {
+      "epoch": 92.93,
+      "learning_rate": 3.725961538461539e-05,
+      "loss": 0.1354,
+      "step": 10129
+    },
+    {
+      "epoch": 92.94,
+      "learning_rate": 3.7211538461538466e-05,
+      "loss": 0.0999,
+      "step": 10130
+    },
+    {
+      "epoch": 92.94,
+      "learning_rate": 3.716346153846154e-05,
+      "loss": 0.1148,
+      "step": 10131
+    },
+    {
+      "epoch": 92.95,
+      "learning_rate": 3.711538461538461e-05,
+      "loss": 0.0874,
+      "step": 10132
+    },
+    {
+      "epoch": 92.96,
+      "learning_rate": 3.706730769230769e-05,
+      "loss": 0.0767,
+      "step": 10133
+    },
+    {
+      "epoch": 92.97,
+      "learning_rate": 3.701923076923077e-05,
+      "loss": 0.1677,
+      "step": 10134
+    },
+    {
+      "epoch": 92.98,
+      "learning_rate": 3.697115384615385e-05,
+      "loss": 0.0744,
+      "step": 10135
+    },
+    {
+      "epoch": 92.99,
+      "learning_rate": 3.692307692307693e-05,
+      "loss": 0.0413,
+      "step": 10136
+    },
+    {
+      "epoch": 93.0,
+      "learning_rate": 3.6875e-05,
+      "loss": 0.1177,
+      "step": 10137
+    },
+    {
+      "epoch": 93.01,
+      "learning_rate": 3.6826923076923074e-05,
+      "loss": 0.1446,
+      "step": 10138
+    },
+    {
+      "epoch": 93.02,
+      "learning_rate": 3.6778846153846154e-05,
+      "loss": 0.1547,
+      "step": 10139
+    },
+    {
+      "epoch": 93.03,
+      "learning_rate": 3.673076923076923e-05,
+      "loss": 0.1496,
+      "step": 10140
+    },
+    {
+      "epoch": 93.04,
+      "learning_rate": 3.6682692307692306e-05,
+      "loss": 0.0964,
+      "step": 10141
+    },
+    {
+      "epoch": 93.05,
+      "learning_rate": 3.6634615384615385e-05,
+      "loss": 0.1074,
+      "step": 10142
+    },
+    {
+      "epoch": 93.06,
+      "learning_rate": 3.6586538461538465e-05,
+      "loss": 0.1076,
+      "step": 10143
+    },
+    {
+      "epoch": 93.06,
+      "learning_rate": 3.6538461538461544e-05,
+      "loss": 0.1029,
+      "step": 10144
+    },
+    {
+      "epoch": 93.07,
+      "learning_rate": 3.649038461538461e-05,
+      "loss": 0.0945,
+      "step": 10145
+    },
+    {
+      "epoch": 93.08,
+      "learning_rate": 3.644230769230769e-05,
+      "loss": 0.1139,
+      "step": 10146
+    },
+    {
+      "epoch": 93.09,
+      "learning_rate": 3.639423076923077e-05,
+      "loss": 0.1683,
+      "step": 10147
+    },
+    {
+      "epoch": 93.1,
+      "learning_rate": 3.634615384615385e-05,
+      "loss": 0.0779,
+      "step": 10148
+    },
+    {
+      "epoch": 93.11,
+      "learning_rate": 3.629807692307693e-05,
+      "loss": 0.0879,
+      "step": 10149
+    },
+    {
+      "epoch": 93.12,
+      "learning_rate": 3.625e-05,
+      "loss": 0.1194,
+      "step": 10150
+    },
+    {
+      "epoch": 93.13,
+      "learning_rate": 3.620192307692308e-05,
+      "loss": 0.1126,
+      "step": 10151
+    },
+    {
+      "epoch": 93.14,
+      "learning_rate": 3.615384615384615e-05,
+      "loss": 0.0916,
+      "step": 10152
+    },
+    {
+      "epoch": 93.15,
+      "learning_rate": 3.610576923076923e-05,
+      "loss": 0.0626,
+      "step": 10153
+    },
+    {
+      "epoch": 93.16,
+      "learning_rate": 3.6057692307692304e-05,
+      "loss": 0.1009,
+      "step": 10154
+    },
+    {
+      "epoch": 93.17,
+      "learning_rate": 3.6009615384615384e-05,
+      "loss": 0.0719,
+      "step": 10155
+    },
+    {
+      "epoch": 93.17,
+      "learning_rate": 3.596153846153846e-05,
+      "loss": 0.0714,
+      "step": 10156
+    },
+    {
+      "epoch": 93.18,
+      "learning_rate": 3.591346153846154e-05,
+      "loss": 0.1088,
+      "step": 10157
+    },
+    {
+      "epoch": 93.19,
+      "learning_rate": 3.586538461538462e-05,
+      "loss": 0.1025,
+      "step": 10158
+    },
+    {
+      "epoch": 93.2,
+      "learning_rate": 3.581730769230769e-05,
+      "loss": 0.0489,
+      "step": 10159
+    },
+    {
+      "epoch": 93.21,
+      "learning_rate": 3.576923076923077e-05,
+      "loss": 0.1591,
+      "step": 10160
+    },
+    {
+      "epoch": 93.22,
+      "learning_rate": 3.5721153846153847e-05,
+      "loss": 0.1934,
+      "step": 10161
+    },
+    {
+      "epoch": 93.23,
+      "learning_rate": 3.5673076923076926e-05,
+      "loss": 0.0666,
+      "step": 10162
+    },
+    {
+      "epoch": 93.24,
+      "learning_rate": 3.5625e-05,
+      "loss": 0.1212,
+      "step": 10163
+    },
+    {
+      "epoch": 93.25,
+      "learning_rate": 3.557692307692308e-05,
+      "loss": 0.0605,
+      "step": 10164
+    },
+    {
+      "epoch": 93.26,
+      "learning_rate": 3.552884615384616e-05,
+      "loss": 0.1958,
+      "step": 10165
+    },
+    {
+      "epoch": 93.27,
+      "learning_rate": 3.548076923076923e-05,
+      "loss": 0.1283,
+      "step": 10166
+    },
+    {
+      "epoch": 93.28,
+      "learning_rate": 3.543269230769231e-05,
+      "loss": 0.1491,
+      "step": 10167
+    },
+    {
+      "epoch": 93.28,
+      "learning_rate": 3.538461538461538e-05,
+      "loss": 0.1399,
+      "step": 10168
+    },
+    {
+      "epoch": 93.29,
+      "learning_rate": 3.533653846153846e-05,
+      "loss": 0.1233,
+      "step": 10169
+    },
+    {
+      "epoch": 93.3,
+      "learning_rate": 3.528846153846154e-05,
+      "loss": 0.0836,
+      "step": 10170
+    },
+    {
+      "epoch": 93.31,
+      "learning_rate": 3.524038461538462e-05,
+      "loss": 0.1349,
+      "step": 10171
+    },
+    {
+      "epoch": 93.32,
+      "learning_rate": 3.519230769230769e-05,
+      "loss": 0.147,
+      "step": 10172
+    },
+    {
+      "epoch": 93.33,
+      "learning_rate": 3.5144230769230766e-05,
+      "loss": 0.1828,
+      "step": 10173
+    },
+    {
+      "epoch": 93.34,
+      "learning_rate": 3.5096153846153845e-05,
+      "loss": 0.145,
+      "step": 10174
+    },
+    {
+      "epoch": 93.35,
+      "learning_rate": 3.5048076923076924e-05,
+      "loss": 0.1288,
+      "step": 10175
+    },
+    {
+      "epoch": 93.36,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 0.0992,
+      "step": 10176
+    },
+    {
+      "epoch": 93.37,
+      "learning_rate": 3.4951923076923076e-05,
+      "loss": 0.0915,
+      "step": 10177
+    },
+    {
+      "epoch": 93.38,
+      "learning_rate": 3.4903846153846156e-05,
+      "loss": 0.0967,
+      "step": 10178
+    },
+    {
+      "epoch": 93.39,
+      "learning_rate": 3.4855769230769235e-05,
+      "loss": 0.1275,
+      "step": 10179
+    },
+    {
+      "epoch": 93.39,
+      "learning_rate": 3.480769230769231e-05,
+      "loss": 0.1242,
+      "step": 10180
+    },
+    {
+      "epoch": 93.4,
+      "learning_rate": 3.475961538461538e-05,
+      "loss": 0.0999,
+      "step": 10181
+    },
+    {
+      "epoch": 93.41,
+      "learning_rate": 3.471153846153846e-05,
+      "loss": 0.1095,
+      "step": 10182
+    },
+    {
+      "epoch": 93.42,
+      "learning_rate": 3.466346153846154e-05,
+      "loss": 0.0972,
+      "step": 10183
+    },
+    {
+      "epoch": 93.43,
+      "learning_rate": 3.461538461538462e-05,
+      "loss": 0.0706,
+      "step": 10184
+    },
+    {
+      "epoch": 93.44,
+      "learning_rate": 3.456730769230769e-05,
+      "loss": 0.0766,
+      "step": 10185
+    },
+    {
+      "epoch": 93.45,
+      "learning_rate": 3.451923076923077e-05,
+      "loss": 0.1121,
+      "step": 10186
+    },
+    {
+      "epoch": 93.46,
+      "learning_rate": 3.447115384615384e-05,
+      "loss": 0.1195,
+      "step": 10187
+    },
+    {
+      "epoch": 93.47,
+      "learning_rate": 3.442307692307692e-05,
+      "loss": 0.0926,
+      "step": 10188
+    },
+    {
+      "epoch": 93.48,
+      "learning_rate": 3.4375e-05,
+      "loss": 0.0643,
+      "step": 10189
+    },
+    {
+      "epoch": 93.49,
+      "learning_rate": 3.4326923076923075e-05,
+      "loss": 0.0596,
+      "step": 10190
+    },
+    {
+      "epoch": 93.5,
+      "learning_rate": 3.4278846153846154e-05,
+      "loss": 0.017,
+      "step": 10191
+    },
+    {
+      "epoch": 93.5,
+      "learning_rate": 3.4230769230769234e-05,
+      "loss": 0.2053,
+      "step": 10192
+    },
+    {
+      "epoch": 93.51,
+      "learning_rate": 3.418269230769231e-05,
+      "loss": 0.1288,
+      "step": 10193
+    },
+    {
+      "epoch": 93.52,
+      "learning_rate": 3.413461538461538e-05,
+      "loss": 0.1218,
+      "step": 10194
+    },
+    {
+      "epoch": 93.53,
+      "learning_rate": 3.408653846153846e-05,
+      "loss": 0.1238,
+      "step": 10195
+    },
+    {
+      "epoch": 93.54,
+      "learning_rate": 3.403846153846154e-05,
+      "loss": 0.1522,
+      "step": 10196
+    },
+    {
+      "epoch": 93.55,
+      "learning_rate": 3.399038461538462e-05,
+      "loss": 0.1157,
+      "step": 10197
+    },
+    {
+      "epoch": 93.56,
+      "learning_rate": 3.3942307692307696e-05,
+      "loss": 0.0646,
+      "step": 10198
+    },
+    {
+      "epoch": 93.57,
+      "learning_rate": 3.389423076923077e-05,
+      "loss": 0.0957,
+      "step": 10199
+    },
+    {
+      "epoch": 93.58,
+      "learning_rate": 3.384615384615385e-05,
+      "loss": 0.1127,
+      "step": 10200
+    },
+    {
+      "epoch": 93.59,
+      "learning_rate": 3.379807692307693e-05,
+      "loss": 0.0987,
+      "step": 10201
+    },
+    {
+      "epoch": 93.6,
+      "learning_rate": 3.375e-05,
+      "loss": 0.0889,
+      "step": 10202
+    },
+    {
+      "epoch": 93.61,
+      "learning_rate": 3.370192307692307e-05,
+      "loss": 0.1063,
+      "step": 10203
+    },
+    {
+      "epoch": 93.61,
+      "learning_rate": 3.365384615384615e-05,
+      "loss": 0.12,
+      "step": 10204
+    },
+    {
+      "epoch": 93.62,
+      "learning_rate": 3.360576923076923e-05,
+      "loss": 0.093,
+      "step": 10205
+    },
+    {
+      "epoch": 93.63,
+      "learning_rate": 3.355769230769231e-05,
+      "loss": 0.134,
+      "step": 10206
+    },
+    {
+      "epoch": 93.64,
+      "learning_rate": 3.350961538461539e-05,
+      "loss": 0.1205,
+      "step": 10207
+    },
+    {
+      "epoch": 93.65,
+      "learning_rate": 3.346153846153846e-05,
+      "loss": 0.1192,
+      "step": 10208
+    },
+    {
+      "epoch": 93.66,
+      "learning_rate": 3.3413461538461536e-05,
+      "loss": 0.0644,
+      "step": 10209
+    },
+    {
+      "epoch": 93.67,
+      "learning_rate": 3.3365384615384615e-05,
+      "loss": 0.0759,
+      "step": 10210
+    },
+    {
+      "epoch": 93.68,
+      "learning_rate": 3.3317307692307695e-05,
+      "loss": 0.1228,
+      "step": 10211
+    },
+    {
+      "epoch": 93.69,
+      "learning_rate": 3.326923076923077e-05,
+      "loss": 0.0554,
+      "step": 10212
+    },
+    {
+      "epoch": 93.7,
+      "learning_rate": 3.322115384615385e-05,
+      "loss": 0.0698,
+      "step": 10213
+    },
+    {
+      "epoch": 93.71,
+      "learning_rate": 3.3173076923076926e-05,
+      "loss": 0.1048,
+      "step": 10214
+    },
+    {
+      "epoch": 93.72,
+      "learning_rate": 3.3125000000000006e-05,
+      "loss": 0.1421,
+      "step": 10215
+    },
+    {
+      "epoch": 93.72,
+      "learning_rate": 3.307692307692308e-05,
+      "loss": 0.1083,
+      "step": 10216
+    },
+    {
+      "epoch": 93.73,
+      "learning_rate": 3.302884615384615e-05,
+      "loss": 0.1051,
+      "step": 10217
+    },
+    {
+      "epoch": 93.74,
+      "learning_rate": 3.298076923076923e-05,
+      "loss": 0.0208,
+      "step": 10218
+    },
+    {
+      "epoch": 93.75,
+      "learning_rate": 3.293269230769231e-05,
+      "loss": 0.1642,
+      "step": 10219
+    },
+    {
+      "epoch": 93.76,
+      "learning_rate": 3.288461538461539e-05,
+      "loss": 0.1583,
+      "step": 10220
+    },
+    {
+      "epoch": 93.77,
+      "learning_rate": 3.283653846153846e-05,
+      "loss": 0.0896,
+      "step": 10221
+    },
+    {
+      "epoch": 93.78,
+      "learning_rate": 3.278846153846154e-05,
+      "loss": 0.1608,
+      "step": 10222
+    },
+    {
+      "epoch": 93.79,
+      "learning_rate": 3.2740384615384614e-05,
+      "loss": 0.1214,
+      "step": 10223
+    },
+    {
+      "epoch": 93.8,
+      "learning_rate": 3.269230769230769e-05,
+      "loss": 0.115,
+      "step": 10224
+    },
+    {
+      "epoch": 93.81,
+      "learning_rate": 3.264423076923077e-05,
+      "loss": 0.1085,
+      "step": 10225
+    },
+    {
+      "epoch": 93.82,
+      "learning_rate": 3.2596153846153845e-05,
+      "loss": 0.1319,
+      "step": 10226
+    },
+    {
+      "epoch": 93.83,
+      "learning_rate": 3.2548076923076925e-05,
+      "loss": 0.1037,
+      "step": 10227
+    },
+    {
+      "epoch": 93.83,
+      "learning_rate": 3.2500000000000004e-05,
+      "loss": 0.0838,
+      "step": 10228
+    },
+    {
+      "epoch": 93.84,
+      "learning_rate": 3.245192307692308e-05,
+      "loss": 0.1393,
+      "step": 10229
+    },
+    {
+      "epoch": 93.85,
+      "learning_rate": 3.240384615384615e-05,
+      "loss": 0.1019,
+      "step": 10230
+    },
+    {
+      "epoch": 93.86,
+      "learning_rate": 3.235576923076923e-05,
+      "loss": 0.1108,
+      "step": 10231
+    },
+    {
+      "epoch": 93.87,
+      "learning_rate": 3.230769230769231e-05,
+      "loss": 0.1861,
+      "step": 10232
+    },
+    {
+      "epoch": 93.88,
+      "learning_rate": 3.225961538461539e-05,
+      "loss": 0.1384,
+      "step": 10233
+    },
+    {
+      "epoch": 93.89,
+      "learning_rate": 3.221153846153846e-05,
+      "loss": 0.133,
+      "step": 10234
+    },
+    {
+      "epoch": 93.9,
+      "learning_rate": 3.216346153846154e-05,
+      "loss": 0.0918,
+      "step": 10235
+    },
+    {
+      "epoch": 93.91,
+      "learning_rate": 3.211538461538462e-05,
+      "loss": 0.1632,
+      "step": 10236
+    },
+    {
+      "epoch": 93.92,
+      "learning_rate": 3.206730769230769e-05,
+      "loss": 0.1264,
+      "step": 10237
+    },
+    {
+      "epoch": 93.93,
+      "learning_rate": 3.201923076923077e-05,
+      "loss": 0.0668,
+      "step": 10238
+    },
+    {
+      "epoch": 93.94,
+      "learning_rate": 3.1971153846153843e-05,
+      "loss": 0.0822,
+      "step": 10239
+    },
+    {
+      "epoch": 93.94,
+      "learning_rate": 3.192307692307692e-05,
+      "loss": 0.0765,
+      "step": 10240
+    },
+    {
+      "epoch": 93.95,
+      "learning_rate": 3.1875e-05,
+      "loss": 0.0743,
+      "step": 10241
+    },
+    {
+      "epoch": 93.96,
+      "learning_rate": 3.182692307692308e-05,
+      "loss": 0.0473,
+      "step": 10242
+    },
+    {
+      "epoch": 93.97,
+      "learning_rate": 3.1778846153846154e-05,
+      "loss": 0.0663,
+      "step": 10243
+    },
+    {
+      "epoch": 93.98,
+      "learning_rate": 3.173076923076923e-05,
+      "loss": 0.0511,
+      "step": 10244
+    },
+    {
+      "epoch": 93.99,
+      "learning_rate": 3.1682692307692306e-05,
+      "loss": 0.0611,
+      "step": 10245
+    },
+    {
+      "epoch": 94.0,
+      "learning_rate": 3.1634615384615386e-05,
+      "loss": 0.1075,
+      "step": 10246
+    },
+    {
+      "epoch": 94.01,
+      "learning_rate": 3.1586538461538465e-05,
+      "loss": 0.1615,
+      "step": 10247
+    },
+    {
+      "epoch": 94.02,
+      "learning_rate": 3.153846153846154e-05,
+      "loss": 0.1101,
+      "step": 10248
+    },
+    {
+      "epoch": 94.03,
+      "learning_rate": 3.149038461538462e-05,
+      "loss": 0.1132,
+      "step": 10249
+    },
+    {
+      "epoch": 94.04,
+      "learning_rate": 3.1442307692307697e-05,
+      "loss": 0.1086,
+      "step": 10250
+    },
+    {
+      "epoch": 94.05,
+      "learning_rate": 3.139423076923077e-05,
+      "loss": 0.1522,
+      "step": 10251
+    },
+    {
+      "epoch": 94.06,
+      "learning_rate": 3.134615384615384e-05,
+      "loss": 0.1205,
+      "step": 10252
+    },
+    {
+      "epoch": 94.06,
+      "learning_rate": 3.129807692307692e-05,
+      "loss": 0.1039,
+      "step": 10253
+    },
+    {
+      "epoch": 94.07,
+      "learning_rate": 3.125e-05,
+      "loss": 0.0767,
+      "step": 10254
+    },
+    {
+      "epoch": 94.08,
+      "learning_rate": 3.120192307692308e-05,
+      "loss": 0.0964,
+      "step": 10255
+    },
+    {
+      "epoch": 94.09,
+      "learning_rate": 3.115384615384615e-05,
+      "loss": 0.0621,
+      "step": 10256
+    },
+    {
+      "epoch": 94.1,
+      "learning_rate": 3.110576923076923e-05,
+      "loss": 0.1051,
+      "step": 10257
+    },
+    {
+      "epoch": 94.11,
+      "learning_rate": 3.1057692307692305e-05,
+      "loss": 0.1167,
+      "step": 10258
+    },
+    {
+      "epoch": 94.12,
+      "learning_rate": 3.1009615384615384e-05,
+      "loss": 0.0907,
+      "step": 10259
+    },
+    {
+      "epoch": 94.13,
+      "learning_rate": 3.0961538461538464e-05,
+      "loss": 0.1198,
+      "step": 10260
+    },
+    {
+      "epoch": 94.14,
+      "learning_rate": 3.091346153846154e-05,
+      "loss": 0.19,
+      "step": 10261
+    },
+    {
+      "epoch": 94.15,
+      "learning_rate": 3.0865384615384616e-05,
+      "loss": 0.1433,
+      "step": 10262
+    },
+    {
+      "epoch": 94.16,
+      "learning_rate": 3.0817307692307695e-05,
+      "loss": 0.1479,
+      "step": 10263
+    },
+    {
+      "epoch": 94.17,
+      "learning_rate": 3.0769230769230774e-05,
+      "loss": 0.1083,
+      "step": 10264
+    },
+    {
+      "epoch": 94.17,
+      "learning_rate": 3.072115384615385e-05,
+      "loss": 0.0833,
+      "step": 10265
+    },
+    {
+      "epoch": 94.18,
+      "learning_rate": 3.0673076923076926e-05,
+      "loss": 0.0965,
+      "step": 10266
+    },
+    {
+      "epoch": 94.19,
+      "learning_rate": 3.0625e-05,
+      "loss": 0.0915,
+      "step": 10267
+    },
+    {
+      "epoch": 94.2,
+      "learning_rate": 3.057692307692308e-05,
+      "loss": 0.0925,
+      "step": 10268
+    },
+    {
+      "epoch": 94.21,
+      "learning_rate": 3.052884615384615e-05,
+      "loss": 0.1332,
+      "step": 10269
+    },
+    {
+      "epoch": 94.22,
+      "learning_rate": 3.048076923076923e-05,
+      "loss": 0.123,
+      "step": 10270
+    },
+    {
+      "epoch": 94.23,
+      "learning_rate": 3.0432692307692306e-05,
+      "loss": 0.0439,
+      "step": 10271
+    },
+    {
+      "epoch": 94.24,
+      "learning_rate": 3.0384615384615386e-05,
+      "loss": 0.0847,
+      "step": 10272
+    },
+    {
+      "epoch": 94.25,
+      "learning_rate": 3.0336538461538462e-05,
+      "loss": 0.0226,
+      "step": 10273
+    },
+    {
+      "epoch": 94.26,
+      "learning_rate": 3.0288461538461538e-05,
+      "loss": 0.1764,
+      "step": 10274
+    },
+    {
+      "epoch": 94.27,
+      "learning_rate": 3.0240384615384617e-05,
+      "loss": 0.0916,
+      "step": 10275
+    },
+    {
+      "epoch": 94.28,
+      "learning_rate": 3.0192307692307693e-05,
+      "loss": 0.158,
+      "step": 10276
+    },
+    {
+      "epoch": 94.28,
+      "learning_rate": 3.014423076923077e-05,
+      "loss": 0.1164,
+      "step": 10277
+    },
+    {
+      "epoch": 94.29,
+      "learning_rate": 3.0096153846153845e-05,
+      "loss": 0.1426,
+      "step": 10278
+    },
+    {
+      "epoch": 94.3,
+      "learning_rate": 3.0048076923076925e-05,
+      "loss": 0.1876,
+      "step": 10279
+    },
+    {
+      "epoch": 94.31,
+      "learning_rate": 3e-05,
+      "loss": 0.0944,
+      "step": 10280
+    },
+    {
+      "epoch": 94.32,
+      "learning_rate": 2.9951923076923077e-05,
+      "loss": 0.0718,
+      "step": 10281
+    },
+    {
+      "epoch": 94.33,
+      "learning_rate": 2.9903846153846153e-05,
+      "loss": 0.1078,
+      "step": 10282
+    },
+    {
+      "epoch": 94.34,
+      "learning_rate": 2.9855769230769232e-05,
+      "loss": 0.0755,
+      "step": 10283
+    },
+    {
+      "epoch": 94.35,
+      "learning_rate": 2.980769230769231e-05,
+      "loss": 0.1343,
+      "step": 10284
+    },
+    {
+      "epoch": 94.36,
+      "learning_rate": 2.9759615384615384e-05,
+      "loss": 0.0866,
+      "step": 10285
+    },
+    {
+      "epoch": 94.37,
+      "learning_rate": 2.9711538461538464e-05,
+      "loss": 0.1301,
+      "step": 10286
+    },
+    {
+      "epoch": 94.38,
+      "learning_rate": 2.966346153846154e-05,
+      "loss": 0.1119,
+      "step": 10287
+    },
+    {
+      "epoch": 94.39,
+      "learning_rate": 2.9615384615384616e-05,
+      "loss": 0.1099,
+      "step": 10288
+    },
+    {
+      "epoch": 94.39,
+      "learning_rate": 2.956730769230769e-05,
+      "loss": 0.0826,
+      "step": 10289
+    },
+    {
+      "epoch": 94.4,
+      "learning_rate": 2.951923076923077e-05,
+      "loss": 0.1095,
+      "step": 10290
+    },
+    {
+      "epoch": 94.41,
+      "learning_rate": 2.9471153846153847e-05,
+      "loss": 0.0387,
+      "step": 10291
+    },
+    {
+      "epoch": 94.42,
+      "learning_rate": 2.9423076923076923e-05,
+      "loss": 0.0601,
+      "step": 10292
+    },
+    {
+      "epoch": 94.43,
+      "learning_rate": 2.9375e-05,
+      "loss": 0.1277,
+      "step": 10293
+    },
+    {
+      "epoch": 94.44,
+      "learning_rate": 2.932692307692308e-05,
+      "loss": 0.1025,
+      "step": 10294
+    },
+    {
+      "epoch": 94.45,
+      "learning_rate": 2.9278846153846155e-05,
+      "loss": 0.1713,
+      "step": 10295
+    },
+    {
+      "epoch": 94.46,
+      "learning_rate": 2.923076923076923e-05,
+      "loss": 0.0652,
+      "step": 10296
+    },
+    {
+      "epoch": 94.47,
+      "learning_rate": 2.918269230769231e-05,
+      "loss": 0.1037,
+      "step": 10297
+    },
+    {
+      "epoch": 94.48,
+      "learning_rate": 2.9134615384615386e-05,
+      "loss": 0.1505,
+      "step": 10298
+    },
+    {
+      "epoch": 94.49,
+      "learning_rate": 2.9086538461538462e-05,
+      "loss": 0.1242,
+      "step": 10299
+    },
+    {
+      "epoch": 94.5,
+      "learning_rate": 2.9038461538461538e-05,
+      "loss": 0.0627,
+      "step": 10300
+    },
+    {
+      "epoch": 94.5,
+      "learning_rate": 2.8990384615384617e-05,
+      "loss": 0.1684,
+      "step": 10301
+    },
+    {
+      "epoch": 94.51,
+      "learning_rate": 2.894230769230769e-05,
+      "loss": 0.1232,
+      "step": 10302
+    },
+    {
+      "epoch": 94.52,
+      "learning_rate": 2.889423076923077e-05,
+      "loss": 0.0839,
+      "step": 10303
+    },
+    {
+      "epoch": 94.53,
+      "learning_rate": 2.884615384615385e-05,
+      "loss": 0.1426,
+      "step": 10304
+    },
+    {
+      "epoch": 94.54,
+      "learning_rate": 2.8798076923076925e-05,
+      "loss": 0.1038,
+      "step": 10305
+    },
+    {
+      "epoch": 94.55,
+      "learning_rate": 2.875e-05,
+      "loss": 0.1793,
+      "step": 10306
+    },
+    {
+      "epoch": 94.56,
+      "learning_rate": 2.8701923076923077e-05,
+      "loss": 0.1592,
+      "step": 10307
+    },
+    {
+      "epoch": 94.57,
+      "learning_rate": 2.8653846153846156e-05,
+      "loss": 0.1394,
+      "step": 10308
+    },
+    {
+      "epoch": 94.58,
+      "learning_rate": 2.860576923076923e-05,
+      "loss": 0.1063,
+      "step": 10309
+    },
+    {
+      "epoch": 94.59,
+      "learning_rate": 2.855769230769231e-05,
+      "loss": 0.0773,
+      "step": 10310
+    },
+    {
+      "epoch": 94.6,
+      "learning_rate": 2.8509615384615384e-05,
+      "loss": 0.0971,
+      "step": 10311
+    },
+    {
+      "epoch": 94.61,
+      "learning_rate": 2.8461538461538464e-05,
+      "loss": 0.1118,
+      "step": 10312
+    },
+    {
+      "epoch": 94.61,
+      "learning_rate": 2.8413461538461536e-05,
+      "loss": 0.0736,
+      "step": 10313
+    },
+    {
+      "epoch": 94.62,
+      "learning_rate": 2.8365384615384616e-05,
+      "loss": 0.0608,
+      "step": 10314
+    },
+    {
+      "epoch": 94.63,
+      "learning_rate": 2.8317307692307695e-05,
+      "loss": 0.1535,
+      "step": 10315
+    },
+    {
+      "epoch": 94.64,
+      "learning_rate": 2.8269230769230768e-05,
+      "loss": 0.0762,
+      "step": 10316
+    },
+    {
+      "epoch": 94.65,
+      "learning_rate": 2.8221153846153847e-05,
+      "loss": 0.0629,
+      "step": 10317
+    },
+    {
+      "epoch": 94.66,
+      "learning_rate": 2.8173076923076923e-05,
+      "loss": 0.1276,
+      "step": 10318
+    },
+    {
+      "epoch": 94.67,
+      "learning_rate": 2.8125000000000003e-05,
+      "loss": 0.1351,
+      "step": 10319
+    },
+    {
+      "epoch": 94.68,
+      "learning_rate": 2.8076923076923075e-05,
+      "loss": 0.0763,
+      "step": 10320
+    },
+    {
+      "epoch": 94.69,
+      "learning_rate": 2.8028846153846155e-05,
+      "loss": 0.1101,
+      "step": 10321
+    },
+    {
+      "epoch": 94.7,
+      "learning_rate": 2.798076923076923e-05,
+      "loss": 0.1259,
+      "step": 10322
+    },
+    {
+      "epoch": 94.71,
+      "learning_rate": 2.7932692307692307e-05,
+      "loss": 0.0863,
+      "step": 10323
+    },
+    {
+      "epoch": 94.72,
+      "learning_rate": 2.7884615384615386e-05,
+      "loss": 0.0939,
+      "step": 10324
+    },
+    {
+      "epoch": 94.72,
+      "learning_rate": 2.7836538461538462e-05,
+      "loss": 0.1472,
+      "step": 10325
+    },
+    {
+      "epoch": 94.73,
+      "learning_rate": 2.778846153846154e-05,
+      "loss": 0.0644,
+      "step": 10326
+    },
+    {
+      "epoch": 94.74,
+      "learning_rate": 2.7740384615384614e-05,
+      "loss": 0.1037,
+      "step": 10327
+    },
+    {
+      "epoch": 94.75,
+      "learning_rate": 2.7692307692307694e-05,
+      "loss": 0.1787,
+      "step": 10328
+    },
+    {
+      "epoch": 94.76,
+      "learning_rate": 2.764423076923077e-05,
+      "loss": 0.1126,
+      "step": 10329
+    },
+    {
+      "epoch": 94.77,
+      "learning_rate": 2.7596153846153846e-05,
+      "loss": 0.142,
+      "step": 10330
+    },
+    {
+      "epoch": 94.78,
+      "learning_rate": 2.754807692307692e-05,
+      "loss": 0.1122,
+      "step": 10331
+    },
+    {
+      "epoch": 94.79,
+      "learning_rate": 2.75e-05,
+      "loss": 0.1342,
+      "step": 10332
+    },
+    {
+      "epoch": 94.8,
+      "learning_rate": 2.745192307692308e-05,
+      "loss": 0.11,
+      "step": 10333
+    },
+    {
+      "epoch": 94.81,
+      "learning_rate": 2.7403846153846153e-05,
+      "loss": 0.1532,
+      "step": 10334
+    },
+    {
+      "epoch": 94.82,
+      "learning_rate": 2.7355769230769232e-05,
+      "loss": 0.1504,
+      "step": 10335
+    },
+    {
+      "epoch": 94.83,
+      "learning_rate": 2.730769230769231e-05,
+      "loss": 0.1017,
+      "step": 10336
+    },
+    {
+      "epoch": 94.83,
+      "learning_rate": 2.7259615384615388e-05,
+      "loss": 0.1168,
+      "step": 10337
+    },
+    {
+      "epoch": 94.84,
+      "learning_rate": 2.721153846153846e-05,
+      "loss": 0.1131,
+      "step": 10338
+    },
+    {
+      "epoch": 94.85,
+      "learning_rate": 2.716346153846154e-05,
+      "loss": 0.0945,
+      "step": 10339
+    },
+    {
+      "epoch": 94.86,
+      "learning_rate": 2.7115384615384616e-05,
+      "loss": 0.0908,
+      "step": 10340
+    },
+    {
+      "epoch": 94.87,
+      "learning_rate": 2.7067307692307692e-05,
+      "loss": 0.1069,
+      "step": 10341
+    },
+    {
+      "epoch": 94.88,
+      "learning_rate": 2.7019230769230768e-05,
+      "loss": 0.0468,
+      "step": 10342
+    },
+    {
+      "epoch": 94.89,
+      "learning_rate": 2.6971153846153847e-05,
+      "loss": 0.0881,
+      "step": 10343
+    },
+    {
+      "epoch": 94.9,
+      "learning_rate": 2.6923076923076927e-05,
+      "loss": 0.0844,
+      "step": 10344
+    },
+    {
+      "epoch": 94.91,
+      "learning_rate": 2.6875e-05,
+      "loss": 0.0787,
+      "step": 10345
+    },
+    {
+      "epoch": 94.92,
+      "learning_rate": 2.682692307692308e-05,
+      "loss": 0.0481,
+      "step": 10346
+    },
+    {
+      "epoch": 94.93,
+      "learning_rate": 2.6778846153846155e-05,
+      "loss": 0.1173,
+      "step": 10347
+    },
+    {
+      "epoch": 94.94,
+      "learning_rate": 2.673076923076923e-05,
+      "loss": 0.0917,
+      "step": 10348
+    },
+    {
+      "epoch": 94.94,
+      "learning_rate": 2.6682692307692307e-05,
+      "loss": 0.0779,
+      "step": 10349
+    },
+    {
+      "epoch": 94.95,
+      "learning_rate": 2.6634615384615386e-05,
+      "loss": 0.0557,
+      "step": 10350
+    },
+    {
+      "epoch": 94.96,
+      "learning_rate": 2.6586538461538462e-05,
+      "loss": 0.04,
+      "step": 10351
+    },
+    {
+      "epoch": 94.97,
+      "learning_rate": 2.6538461538461538e-05,
+      "loss": 0.0863,
+      "step": 10352
+    },
+    {
+      "epoch": 94.98,
+      "learning_rate": 2.6490384615384618e-05,
+      "loss": 0.0604,
+      "step": 10353
+    },
+    {
+      "epoch": 94.99,
+      "learning_rate": 2.6442307692307694e-05,
+      "loss": 0.0352,
+      "step": 10354
+    },
+    {
+      "epoch": 95.0,
+      "learning_rate": 2.639423076923077e-05,
+      "loss": 0.1061,
+      "step": 10355
+    },
+    {
+      "epoch": 95.01,
+      "learning_rate": 2.6346153846153846e-05,
+      "loss": 0.1261,
+      "step": 10356
+    },
+    {
+      "epoch": 95.02,
+      "learning_rate": 2.6298076923076925e-05,
+      "loss": 0.1403,
+      "step": 10357
+    },
+    {
+      "epoch": 95.03,
+      "learning_rate": 2.625e-05,
+      "loss": 0.168,
+      "step": 10358
+    },
+    {
+      "epoch": 95.04,
+      "learning_rate": 2.6201923076923077e-05,
+      "loss": 0.1503,
+      "step": 10359
+    },
+    {
+      "epoch": 95.05,
+      "learning_rate": 2.6153846153846153e-05,
+      "loss": 0.1545,
+      "step": 10360
+    },
+    {
+      "epoch": 95.06,
+      "learning_rate": 2.6105769230769233e-05,
+      "loss": 0.0899,
+      "step": 10361
+    },
+    {
+      "epoch": 95.06,
+      "learning_rate": 2.6057692307692305e-05,
+      "loss": 0.0815,
+      "step": 10362
+    },
+    {
+      "epoch": 95.07,
+      "learning_rate": 2.6009615384615385e-05,
+      "loss": 0.1047,
+      "step": 10363
+    },
+    {
+      "epoch": 95.08,
+      "learning_rate": 2.5961538461538464e-05,
+      "loss": 0.0766,
+      "step": 10364
+    },
+    {
+      "epoch": 95.09,
+      "learning_rate": 2.591346153846154e-05,
+      "loss": 0.057,
+      "step": 10365
+    },
+    {
+      "epoch": 95.1,
+      "learning_rate": 2.5865384615384616e-05,
+      "loss": 0.1026,
+      "step": 10366
+    },
+    {
+      "epoch": 95.11,
+      "learning_rate": 2.5817307692307692e-05,
+      "loss": 0.1225,
+      "step": 10367
+    },
+    {
+      "epoch": 95.12,
+      "learning_rate": 2.576923076923077e-05,
+      "loss": 0.1344,
+      "step": 10368
+    },
+    {
+      "epoch": 95.13,
+      "learning_rate": 2.5721153846153844e-05,
+      "loss": 0.0917,
+      "step": 10369
+    },
+    {
+      "epoch": 95.14,
+      "learning_rate": 2.5673076923076923e-05,
+      "loss": 0.069,
+      "step": 10370
+    },
+    {
+      "epoch": 95.15,
+      "learning_rate": 2.5625e-05,
+      "loss": 0.0681,
+      "step": 10371
+    },
+    {
+      "epoch": 95.16,
+      "learning_rate": 2.557692307692308e-05,
+      "loss": 0.0747,
+      "step": 10372
+    },
+    {
+      "epoch": 95.17,
+      "learning_rate": 2.5528846153846155e-05,
+      "loss": 0.0769,
+      "step": 10373
+    },
+    {
+      "epoch": 95.17,
+      "learning_rate": 2.548076923076923e-05,
+      "loss": 0.1076,
+      "step": 10374
+    },
+    {
+      "epoch": 95.18,
+      "learning_rate": 2.543269230769231e-05,
+      "loss": 0.0868,
+      "step": 10375
+    },
+    {
+      "epoch": 95.19,
+      "learning_rate": 2.5384615384615383e-05,
+      "loss": 0.0662,
+      "step": 10376
+    },
+    {
+      "epoch": 95.2,
+      "learning_rate": 2.5336538461538462e-05,
+      "loss": 0.0991,
+      "step": 10377
+    },
+    {
+      "epoch": 95.21,
+      "learning_rate": 2.528846153846154e-05,
+      "loss": 0.0931,
+      "step": 10378
+    },
+    {
+      "epoch": 95.22,
+      "learning_rate": 2.5240384615384618e-05,
+      "loss": 0.1044,
+      "step": 10379
+    },
+    {
+      "epoch": 95.23,
+      "learning_rate": 2.519230769230769e-05,
+      "loss": 0.1033,
+      "step": 10380
+    },
+    {
+      "epoch": 95.24,
+      "learning_rate": 2.514423076923077e-05,
+      "loss": 0.1033,
+      "step": 10381
+    },
+    {
+      "epoch": 95.25,
+      "learning_rate": 2.5096153846153846e-05,
+      "loss": 0.0904,
+      "step": 10382
+    },
+    {
+      "epoch": 95.26,
+      "learning_rate": 2.5048076923076922e-05,
+      "loss": 0.1408,
+      "step": 10383
+    },
+    {
+      "epoch": 95.27,
+      "learning_rate": 2.5e-05,
+      "loss": 0.1025,
+      "step": 10384
+    },
+    {
+      "epoch": 95.28,
+      "learning_rate": 2.4951923076923077e-05,
+      "loss": 0.1448,
+      "step": 10385
+    },
+    {
+      "epoch": 95.28,
+      "learning_rate": 2.4903846153846157e-05,
+      "loss": 0.1196,
+      "step": 10386
+    },
+    {
+      "epoch": 95.29,
+      "learning_rate": 2.485576923076923e-05,
+      "loss": 0.1284,
+      "step": 10387
+    },
+    {
+      "epoch": 95.3,
+      "learning_rate": 2.480769230769231e-05,
+      "loss": 0.1257,
+      "step": 10388
+    },
+    {
+      "epoch": 95.31,
+      "learning_rate": 2.4759615384615385e-05,
+      "loss": 0.0918,
+      "step": 10389
+    },
+    {
+      "epoch": 95.32,
+      "learning_rate": 2.4711538461538464e-05,
+      "loss": 0.1501,
+      "step": 10390
+    },
+    {
+      "epoch": 95.33,
+      "learning_rate": 2.4663461538461537e-05,
+      "loss": 0.1074,
+      "step": 10391
+    },
+    {
+      "epoch": 95.34,
+      "learning_rate": 2.4615384615384616e-05,
+      "loss": 0.1432,
+      "step": 10392
+    },
+    {
+      "epoch": 95.35,
+      "learning_rate": 2.4567307692307695e-05,
+      "loss": 0.0969,
+      "step": 10393
+    },
+    {
+      "epoch": 95.36,
+      "learning_rate": 2.4519230769230768e-05,
+      "loss": 0.0923,
+      "step": 10394
+    },
+    {
+      "epoch": 95.37,
+      "learning_rate": 2.4471153846153848e-05,
+      "loss": 0.0898,
+      "step": 10395
+    },
+    {
+      "epoch": 95.38,
+      "learning_rate": 2.4423076923076924e-05,
+      "loss": 0.1103,
+      "step": 10396
+    },
+    {
+      "epoch": 95.39,
+      "learning_rate": 2.4375000000000003e-05,
+      "loss": 0.0857,
+      "step": 10397
+    },
+    {
+      "epoch": 95.39,
+      "learning_rate": 2.4326923076923076e-05,
+      "loss": 0.1517,
+      "step": 10398
+    },
+    {
+      "epoch": 95.4,
+      "learning_rate": 2.4278846153846155e-05,
+      "loss": 0.0987,
+      "step": 10399
+    },
+    {
+      "epoch": 95.41,
+      "learning_rate": 2.423076923076923e-05,
+      "loss": 0.0877,
+      "step": 10400
+    },
+    {
+      "epoch": 95.42,
+      "learning_rate": 2.4182692307692307e-05,
+      "loss": 0.1588,
+      "step": 10401
+    },
+    {
+      "epoch": 95.43,
+      "learning_rate": 2.4134615384615386e-05,
+      "loss": 0.1773,
+      "step": 10402
+    },
+    {
+      "epoch": 95.44,
+      "learning_rate": 2.4086538461538462e-05,
+      "loss": 0.0895,
+      "step": 10403
+    },
+    {
+      "epoch": 95.45,
+      "learning_rate": 2.4038461538461542e-05,
+      "loss": 0.1514,
+      "step": 10404
+    },
+    {
+      "epoch": 95.46,
+      "learning_rate": 2.3990384615384614e-05,
+      "loss": 0.0451,
+      "step": 10405
+    },
+    {
+      "epoch": 95.47,
+      "learning_rate": 2.3942307692307694e-05,
+      "loss": 0.0803,
+      "step": 10406
+    },
+    {
+      "epoch": 95.48,
+      "learning_rate": 2.389423076923077e-05,
+      "loss": 0.0743,
+      "step": 10407
+    },
+    {
+      "epoch": 95.49,
+      "learning_rate": 2.3846153846153846e-05,
+      "loss": 0.0761,
+      "step": 10408
+    },
+    {
+      "epoch": 95.5,
+      "learning_rate": 2.3798076923076922e-05,
+      "loss": 0.0481,
+      "step": 10409
+    },
+    {
+      "epoch": 95.5,
+      "learning_rate": 2.375e-05,
+      "loss": 0.2069,
+      "step": 10410
+    },
+    {
+      "epoch": 95.51,
+      "learning_rate": 2.3701923076923077e-05,
+      "loss": 0.1305,
+      "step": 10411
+    },
+    {
+      "epoch": 95.52,
+      "learning_rate": 2.3653846153846153e-05,
+      "loss": 0.082,
+      "step": 10412
+    },
+    {
+      "epoch": 95.53,
+      "learning_rate": 2.3605769230769233e-05,
+      "loss": 0.1016,
+      "step": 10413
+    },
+    {
+      "epoch": 95.54,
+      "learning_rate": 2.355769230769231e-05,
+      "loss": 0.1111,
+      "step": 10414
+    },
+    {
+      "epoch": 95.55,
+      "learning_rate": 2.3509615384615385e-05,
+      "loss": 0.1116,
+      "step": 10415
+    },
+    {
+      "epoch": 95.56,
+      "learning_rate": 2.346153846153846e-05,
+      "loss": 0.1121,
+      "step": 10416
+    },
+    {
+      "epoch": 95.57,
+      "learning_rate": 2.341346153846154e-05,
+      "loss": 0.0977,
+      "step": 10417
+    },
+    {
+      "epoch": 95.58,
+      "learning_rate": 2.3365384615384616e-05,
+      "loss": 0.1107,
+      "step": 10418
+    },
+    {
+      "epoch": 95.59,
+      "learning_rate": 2.3317307692307692e-05,
+      "loss": 0.1586,
+      "step": 10419
+    },
+    {
+      "epoch": 95.6,
+      "learning_rate": 2.3269230769230768e-05,
+      "loss": 0.0818,
+      "step": 10420
+    },
+    {
+      "epoch": 95.61,
+      "learning_rate": 2.3221153846153848e-05,
+      "loss": 0.1068,
+      "step": 10421
+    },
+    {
+      "epoch": 95.61,
+      "learning_rate": 2.3173076923076924e-05,
+      "loss": 0.1073,
+      "step": 10422
+    },
+    {
+      "epoch": 95.62,
+      "learning_rate": 2.3125e-05,
+      "loss": 0.168,
+      "step": 10423
+    },
+    {
+      "epoch": 95.63,
+      "learning_rate": 2.307692307692308e-05,
+      "loss": 0.0877,
+      "step": 10424
+    },
+    {
+      "epoch": 95.64,
+      "learning_rate": 2.3028846153846155e-05,
+      "loss": 0.1067,
+      "step": 10425
+    },
+    {
+      "epoch": 95.65,
+      "learning_rate": 2.298076923076923e-05,
+      "loss": 0.0979,
+      "step": 10426
+    },
+    {
+      "epoch": 95.66,
+      "learning_rate": 2.2932692307692307e-05,
+      "loss": 0.1119,
+      "step": 10427
+    },
+    {
+      "epoch": 95.67,
+      "learning_rate": 2.2884615384615387e-05,
+      "loss": 0.0888,
+      "step": 10428
+    },
+    {
+      "epoch": 95.68,
+      "learning_rate": 2.283653846153846e-05,
+      "loss": 0.1139,
+      "step": 10429
+    },
+    {
+      "epoch": 95.69,
+      "learning_rate": 2.278846153846154e-05,
+      "loss": 0.0899,
+      "step": 10430
+    },
+    {
+      "epoch": 95.7,
+      "learning_rate": 2.2740384615384615e-05,
+      "loss": 0.0854,
+      "step": 10431
+    },
+    {
+      "epoch": 95.71,
+      "learning_rate": 2.2692307692307694e-05,
+      "loss": 0.1057,
+      "step": 10432
+    },
+    {
+      "epoch": 95.72,
+      "learning_rate": 2.264423076923077e-05,
+      "loss": 0.0539,
+      "step": 10433
+    },
+    {
+      "epoch": 95.72,
+      "learning_rate": 2.2596153846153846e-05,
+      "loss": 0.1261,
+      "step": 10434
+    },
+    {
+      "epoch": 95.73,
+      "learning_rate": 2.2548076923076925e-05,
+      "loss": 0.0676,
+      "step": 10435
+    },
+    {
+      "epoch": 95.74,
+      "learning_rate": 2.2499999999999998e-05,
+      "loss": 0.0486,
+      "step": 10436
+    },
+    {
+      "epoch": 95.75,
+      "learning_rate": 2.2451923076923077e-05,
+      "loss": 0.1968,
+      "step": 10437
+    },
+    {
+      "epoch": 95.76,
+      "learning_rate": 2.2403846153846153e-05,
+      "loss": 0.1346,
+      "step": 10438
+    },
+    {
+      "epoch": 95.77,
+      "learning_rate": 2.2355769230769233e-05,
+      "loss": 0.1679,
+      "step": 10439
+    },
+    {
+      "epoch": 95.78,
+      "learning_rate": 2.2307692307692305e-05,
+      "loss": 0.0738,
+      "step": 10440
+    },
+    {
+      "epoch": 95.79,
+      "learning_rate": 2.2259615384615385e-05,
+      "loss": 0.0982,
+      "step": 10441
+    },
+    {
+      "epoch": 95.8,
+      "learning_rate": 2.2211538461538464e-05,
+      "loss": 0.1282,
+      "step": 10442
+    },
+    {
+      "epoch": 95.81,
+      "learning_rate": 2.2163461538461537e-05,
+      "loss": 0.0614,
+      "step": 10443
+    },
+    {
+      "epoch": 95.82,
+      "learning_rate": 2.2115384615384616e-05,
+      "loss": 0.1405,
+      "step": 10444
+    },
+    {
+      "epoch": 95.83,
+      "learning_rate": 2.2067307692307692e-05,
+      "loss": 0.0785,
+      "step": 10445
+    },
+    {
+      "epoch": 95.83,
+      "learning_rate": 2.2019230769230772e-05,
+      "loss": 0.099,
+      "step": 10446
+    },
+    {
+      "epoch": 95.84,
+      "learning_rate": 2.1971153846153844e-05,
+      "loss": 0.1392,
+      "step": 10447
+    },
+    {
+      "epoch": 95.85,
+      "learning_rate": 2.1923076923076924e-05,
+      "loss": 0.0903,
+      "step": 10448
+    },
+    {
+      "epoch": 95.86,
+      "learning_rate": 2.1875e-05,
+      "loss": 0.0886,
+      "step": 10449
+    },
+    {
+      "epoch": 95.87,
+      "learning_rate": 2.182692307692308e-05,
+      "loss": 0.0732,
+      "step": 10450
+    },
+    {
+      "epoch": 95.88,
+      "learning_rate": 2.1778846153846155e-05,
+      "loss": 0.0887,
+      "step": 10451
+    },
+    {
+      "epoch": 95.89,
+      "learning_rate": 2.173076923076923e-05,
+      "loss": 0.0669,
+      "step": 10452
+    },
+    {
+      "epoch": 95.9,
+      "learning_rate": 2.168269230769231e-05,
+      "loss": 0.0938,
+      "step": 10453
+    },
+    {
+      "epoch": 95.91,
+      "learning_rate": 2.1634615384615383e-05,
+      "loss": 0.04,
+      "step": 10454
+    },
+    {
+      "epoch": 95.92,
+      "learning_rate": 2.1586538461538463e-05,
+      "loss": 0.1259,
+      "step": 10455
+    },
+    {
+      "epoch": 95.93,
+      "learning_rate": 2.153846153846154e-05,
+      "loss": 0.1132,
+      "step": 10456
+    },
+    {
+      "epoch": 95.94,
+      "learning_rate": 2.1490384615384618e-05,
+      "loss": 0.058,
+      "step": 10457
+    },
+    {
+      "epoch": 95.94,
+      "learning_rate": 2.144230769230769e-05,
+      "loss": 0.0931,
+      "step": 10458
+    },
+    {
+      "epoch": 95.95,
+      "learning_rate": 2.139423076923077e-05,
+      "loss": 0.0897,
+      "step": 10459
+    },
+    {
+      "epoch": 95.96,
+      "learning_rate": 2.1346153846153846e-05,
+      "loss": 0.1169,
+      "step": 10460
+    },
+    {
+      "epoch": 95.97,
+      "learning_rate": 2.1298076923076922e-05,
+      "loss": 0.2418,
+      "step": 10461
+    },
+    {
+      "epoch": 95.98,
+      "learning_rate": 2.125e-05,
+      "loss": 0.0707,
+      "step": 10462
+    },
+    {
+      "epoch": 95.99,
+      "learning_rate": 2.1201923076923078e-05,
+      "loss": 0.0848,
+      "step": 10463
+    },
+    {
+      "epoch": 96.0,
+      "learning_rate": 2.1153846153846157e-05,
+      "loss": 0.1187,
+      "step": 10464
+    },
+    {
+      "epoch": 96.01,
+      "learning_rate": 2.110576923076923e-05,
+      "loss": 0.1566,
+      "step": 10465
+    },
+    {
+      "epoch": 96.02,
+      "learning_rate": 2.105769230769231e-05,
+      "loss": 0.1352,
+      "step": 10466
+    },
+    {
+      "epoch": 96.03,
+      "learning_rate": 2.1009615384615385e-05,
+      "loss": 0.1075,
+      "step": 10467
+    },
+    {
+      "epoch": 96.04,
+      "learning_rate": 2.096153846153846e-05,
+      "loss": 0.0872,
+      "step": 10468
+    },
+    {
+      "epoch": 96.05,
+      "learning_rate": 2.0913461538461537e-05,
+      "loss": 0.1282,
+      "step": 10469
+    },
+    {
+      "epoch": 96.06,
+      "learning_rate": 2.0865384615384616e-05,
+      "loss": 0.123,
+      "step": 10470
+    },
+    {
+      "epoch": 96.06,
+      "learning_rate": 2.0817307692307696e-05,
+      "loss": 0.1014,
+      "step": 10471
+    },
+    {
+      "epoch": 96.07,
+      "learning_rate": 2.076923076923077e-05,
+      "loss": 0.1413,
+      "step": 10472
+    },
+    {
+      "epoch": 96.08,
+      "learning_rate": 2.0721153846153848e-05,
+      "loss": 0.1095,
+      "step": 10473
+    },
+    {
+      "epoch": 96.09,
+      "learning_rate": 2.0673076923076924e-05,
+      "loss": 0.0891,
+      "step": 10474
+    },
+    {
+      "epoch": 96.1,
+      "learning_rate": 2.0625e-05,
+      "loss": 0.0818,
+      "step": 10475
+    },
+    {
+      "epoch": 96.11,
+      "learning_rate": 2.0576923076923076e-05,
+      "loss": 0.1239,
+      "step": 10476
+    },
+    {
+      "epoch": 96.12,
+      "learning_rate": 2.0528846153846155e-05,
+      "loss": 0.1177,
+      "step": 10477
+    },
+    {
+      "epoch": 96.13,
+      "learning_rate": 2.048076923076923e-05,
+      "loss": 0.0672,
+      "step": 10478
+    },
+    {
+      "epoch": 96.14,
+      "learning_rate": 2.0432692307692307e-05,
+      "loss": 0.1126,
+      "step": 10479
+    },
+    {
+      "epoch": 96.15,
+      "learning_rate": 2.0384615384615383e-05,
+      "loss": 0.077,
+      "step": 10480
+    },
+    {
+      "epoch": 96.16,
+      "learning_rate": 2.0336538461538463e-05,
+      "loss": 0.1095,
+      "step": 10481
+    },
+    {
+      "epoch": 96.17,
+      "learning_rate": 2.028846153846154e-05,
+      "loss": 0.0811,
+      "step": 10482
+    },
+    {
+      "epoch": 96.17,
+      "learning_rate": 2.0240384615384615e-05,
+      "loss": 0.0802,
+      "step": 10483
+    },
+    {
+      "epoch": 96.18,
+      "learning_rate": 2.0192307692307694e-05,
+      "loss": 0.1463,
+      "step": 10484
+    },
+    {
+      "epoch": 96.19,
+      "learning_rate": 2.014423076923077e-05,
+      "loss": 0.1093,
+      "step": 10485
+    },
+    {
+      "epoch": 96.2,
+      "learning_rate": 2.0096153846153846e-05,
+      "loss": 0.0946,
+      "step": 10486
+    },
+    {
+      "epoch": 96.21,
+      "learning_rate": 2.0048076923076922e-05,
+      "loss": 0.0635,
+      "step": 10487
+    },
+    {
+      "epoch": 96.22,
+      "learning_rate": 2e-05,
+      "loss": 0.1061,
+      "step": 10488
+    },
+    {
+      "epoch": 96.23,
+      "learning_rate": 1.9951923076923074e-05,
+      "loss": 0.0618,
+      "step": 10489
+    },
+    {
+      "epoch": 96.24,
+      "learning_rate": 1.9903846153846154e-05,
+      "loss": 0.0505,
+      "step": 10490
+    },
+    {
+      "epoch": 96.25,
+      "learning_rate": 1.9855769230769233e-05,
+      "loss": 0.1072,
+      "step": 10491
+    },
+    {
+      "epoch": 96.26,
+      "learning_rate": 1.980769230769231e-05,
+      "loss": 0.1766,
+      "step": 10492
+    },
+    {
+      "epoch": 96.27,
+      "learning_rate": 1.9759615384615385e-05,
+      "loss": 0.1332,
+      "step": 10493
+    },
+    {
+      "epoch": 96.28,
+      "learning_rate": 1.971153846153846e-05,
+      "loss": 0.092,
+      "step": 10494
+    },
+    {
+      "epoch": 96.28,
+      "learning_rate": 1.966346153846154e-05,
+      "loss": 0.1285,
+      "step": 10495
+    },
+    {
+      "epoch": 96.29,
+      "learning_rate": 1.9615384615384613e-05,
+      "loss": 0.1155,
+      "step": 10496
+    },
+    {
+      "epoch": 96.3,
+      "learning_rate": 1.9567307692307693e-05,
+      "loss": 0.0806,
+      "step": 10497
+    },
+    {
+      "epoch": 96.31,
+      "learning_rate": 1.951923076923077e-05,
+      "loss": 0.1119,
+      "step": 10498
+    },
+    {
+      "epoch": 96.32,
+      "learning_rate": 1.9471153846153848e-05,
+      "loss": 0.1328,
+      "step": 10499
+    },
+    {
+      "epoch": 96.33,
+      "learning_rate": 1.942307692307692e-05,
+      "loss": 0.0586,
+      "step": 10500
+    },
+    {
+      "epoch": 96.33,
+      "eval_cer": 0.13390005847164124,
+      "eval_loss": 0.7359636425971985,
+      "eval_runtime": 87.8802,
+      "eval_samples_per_second": 18.741,
+      "eval_steps_per_second": 2.344,
+      "eval_wer": 0.4754366254723726,
+      "step": 10500
+    },
+    {
+      "epoch": 96.34,
+      "learning_rate": 1.9375e-05,
+      "loss": 0.1608,
+      "step": 10501
+    },
+    {
+      "epoch": 96.35,
+      "learning_rate": 1.932692307692308e-05,
+      "loss": 0.1755,
+      "step": 10502
+    },
+    {
+      "epoch": 96.36,
+      "learning_rate": 1.9278846153846155e-05,
+      "loss": 0.0978,
+      "step": 10503
+    },
+    {
+      "epoch": 96.37,
+      "learning_rate": 1.923076923076923e-05,
+      "loss": 0.1605,
+      "step": 10504
+    },
+    {
+      "epoch": 96.38,
+      "learning_rate": 1.9182692307692307e-05,
+      "loss": 0.1321,
+      "step": 10505
+    },
+    {
+      "epoch": 96.39,
+      "learning_rate": 1.9134615384615387e-05,
+      "loss": 0.1023,
+      "step": 10506
+    },
+    {
+      "epoch": 96.39,
+      "learning_rate": 1.908653846153846e-05,
+      "loss": 0.0784,
+      "step": 10507
+    },
+    {
+      "epoch": 96.4,
+      "learning_rate": 1.903846153846154e-05,
+      "loss": 0.069,
+      "step": 10508
+    },
+    {
+      "epoch": 96.41,
+      "learning_rate": 1.8990384615384615e-05,
+      "loss": 0.0694,
+      "step": 10509
+    },
+    {
+      "epoch": 96.42,
+      "learning_rate": 1.8942307692307694e-05,
+      "loss": 0.1372,
+      "step": 10510
+    },
+    {
+      "epoch": 96.43,
+      "learning_rate": 1.889423076923077e-05,
+      "loss": 0.0912,
+      "step": 10511
+    },
+    {
+      "epoch": 96.44,
+      "learning_rate": 1.8846153846153846e-05,
+      "loss": 0.1011,
+      "step": 10512
+    },
+    {
+      "epoch": 96.45,
+      "learning_rate": 1.8798076923076926e-05,
+      "loss": 0.1284,
+      "step": 10513
+    },
+    {
+      "epoch": 96.46,
+      "learning_rate": 1.875e-05,
+      "loss": 0.1563,
+      "step": 10514
+    },
+    {
+      "epoch": 96.47,
+      "learning_rate": 1.8701923076923078e-05,
+      "loss": 0.0666,
+      "step": 10515
+    },
+    {
+      "epoch": 96.48,
+      "learning_rate": 1.8653846153846154e-05,
+      "loss": 0.1044,
+      "step": 10516
+    },
+    {
+      "epoch": 96.49,
+      "learning_rate": 1.8605769230769233e-05,
+      "loss": 0.0935,
+      "step": 10517
+    },
+    {
+      "epoch": 96.5,
+      "learning_rate": 1.8557692307692306e-05,
+      "loss": 0.112,
+      "step": 10518
+    },
+    {
+      "epoch": 96.5,
+      "learning_rate": 1.8509615384615385e-05,
+      "loss": 0.126,
+      "step": 10519
+    },
+    {
+      "epoch": 96.51,
+      "learning_rate": 1.8461538461538465e-05,
+      "loss": 0.1627,
+      "step": 10520
+    },
+    {
+      "epoch": 96.52,
+      "learning_rate": 1.8413461538461537e-05,
+      "loss": 0.1363,
+      "step": 10521
+    },
+    {
+      "epoch": 96.53,
+      "learning_rate": 1.8365384615384617e-05,
+      "loss": 0.1524,
+      "step": 10522
+    },
+    {
+      "epoch": 96.54,
+      "learning_rate": 1.8317307692307693e-05,
+      "loss": 0.1185,
+      "step": 10523
+    },
+    {
+      "epoch": 96.55,
+      "learning_rate": 1.8269230769230772e-05,
+      "loss": 0.111,
+      "step": 10524
+    },
+    {
+      "epoch": 96.56,
+      "learning_rate": 1.8221153846153845e-05,
+      "loss": 0.1093,
+      "step": 10525
+    },
+    {
+      "epoch": 96.57,
+      "learning_rate": 1.8173076923076924e-05,
+      "loss": 0.058,
+      "step": 10526
+    },
+    {
+      "epoch": 96.58,
+      "learning_rate": 1.8125e-05,
+      "loss": 0.0667,
+      "step": 10527
+    },
+    {
+      "epoch": 96.59,
+      "learning_rate": 1.8076923076923076e-05,
+      "loss": 0.1126,
+      "step": 10528
+    },
+    {
+      "epoch": 96.6,
+      "learning_rate": 1.8028846153846152e-05,
+      "loss": 0.0928,
+      "step": 10529
+    },
+    {
+      "epoch": 96.61,
+      "learning_rate": 1.798076923076923e-05,
+      "loss": 0.0966,
+      "step": 10530
+    },
+    {
+      "epoch": 96.61,
+      "learning_rate": 1.793269230769231e-05,
+      "loss": 0.0637,
+      "step": 10531
+    },
+    {
+      "epoch": 96.62,
+      "learning_rate": 1.7884615384615384e-05,
+      "loss": 0.0786,
+      "step": 10532
+    },
+    {
+      "epoch": 96.63,
+      "learning_rate": 1.7836538461538463e-05,
+      "loss": 0.1502,
+      "step": 10533
+    },
+    {
+      "epoch": 96.64,
+      "learning_rate": 1.778846153846154e-05,
+      "loss": 0.0982,
+      "step": 10534
+    },
+    {
+      "epoch": 96.65,
+      "learning_rate": 1.7740384615384615e-05,
+      "loss": 0.1326,
+      "step": 10535
+    },
+    {
+      "epoch": 96.66,
+      "learning_rate": 1.769230769230769e-05,
+      "loss": 0.0662,
+      "step": 10536
+    },
+    {
+      "epoch": 96.67,
+      "learning_rate": 1.764423076923077e-05,
+      "loss": 0.1025,
+      "step": 10537
+    },
+    {
+      "epoch": 96.68,
+      "learning_rate": 1.7596153846153846e-05,
+      "loss": 0.1385,
+      "step": 10538
+    },
+    {
+      "epoch": 96.69,
+      "learning_rate": 1.7548076923076922e-05,
+      "loss": 0.0851,
+      "step": 10539
+    },
+    {
+      "epoch": 96.7,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 0.0788,
+      "step": 10540
+    },
+    {
+      "epoch": 96.71,
+      "learning_rate": 1.7451923076923078e-05,
+      "loss": 0.0495,
+      "step": 10541
+    },
+    {
+      "epoch": 96.72,
+      "learning_rate": 1.7403846153846154e-05,
+      "loss": 0.0613,
+      "step": 10542
+    },
+    {
+      "epoch": 96.72,
+      "learning_rate": 1.735576923076923e-05,
+      "loss": 0.083,
+      "step": 10543
+    },
+    {
+      "epoch": 96.73,
+      "learning_rate": 1.730769230769231e-05,
+      "loss": 0.1314,
+      "step": 10544
+    },
+    {
+      "epoch": 96.74,
+      "learning_rate": 1.7259615384615385e-05,
+      "loss": 0.1278,
+      "step": 10545
+    },
+    {
+      "epoch": 96.75,
+      "learning_rate": 1.721153846153846e-05,
+      "loss": 0.1405,
+      "step": 10546
+    },
+    {
+      "epoch": 96.76,
+      "learning_rate": 1.7163461538461537e-05,
+      "loss": 0.0983,
+      "step": 10547
+    },
+    {
+      "epoch": 96.77,
+      "learning_rate": 1.7115384615384617e-05,
+      "loss": 0.1169,
+      "step": 10548
+    },
+    {
+      "epoch": 96.78,
+      "learning_rate": 1.706730769230769e-05,
+      "loss": 0.0914,
+      "step": 10549
+    },
+    {
+      "epoch": 96.79,
+      "learning_rate": 1.701923076923077e-05,
+      "loss": 0.0971,
+      "step": 10550
+    },
+    {
+      "epoch": 96.8,
+      "learning_rate": 1.6971153846153848e-05,
+      "loss": 0.0951,
+      "step": 10551
+    },
+    {
+      "epoch": 96.81,
+      "learning_rate": 1.6923076923076924e-05,
+      "loss": 0.0813,
+      "step": 10552
+    },
+    {
+      "epoch": 96.82,
+      "learning_rate": 1.6875e-05,
+      "loss": 0.0793,
+      "step": 10553
+    },
+    {
+      "epoch": 96.83,
+      "learning_rate": 1.6826923076923076e-05,
+      "loss": 0.0885,
+      "step": 10554
+    },
+    {
+      "epoch": 96.83,
+      "learning_rate": 1.6778846153846156e-05,
+      "loss": 0.073,
+      "step": 10555
+    },
+    {
+      "epoch": 96.84,
+      "learning_rate": 1.673076923076923e-05,
+      "loss": 0.0892,
+      "step": 10556
+    },
+    {
+      "epoch": 96.85,
+      "learning_rate": 1.6682692307692308e-05,
+      "loss": 0.0787,
+      "step": 10557
+    },
+    {
+      "epoch": 96.86,
+      "learning_rate": 1.6634615384615384e-05,
+      "loss": 0.0782,
+      "step": 10558
+    },
+    {
+      "epoch": 96.87,
+      "learning_rate": 1.6586538461538463e-05,
+      "loss": 0.0622,
+      "step": 10559
+    },
+    {
+      "epoch": 96.88,
+      "learning_rate": 1.653846153846154e-05,
+      "loss": 0.0956,
+      "step": 10560
+    },
+    {
+      "epoch": 96.89,
+      "learning_rate": 1.6490384615384615e-05,
+      "loss": 0.0741,
+      "step": 10561
+    },
+    {
+      "epoch": 96.9,
+      "learning_rate": 1.6442307692307695e-05,
+      "loss": 0.1073,
+      "step": 10562
+    },
+    {
+      "epoch": 96.91,
+      "learning_rate": 1.639423076923077e-05,
+      "loss": 0.0747,
+      "step": 10563
+    },
+    {
+      "epoch": 96.92,
+      "learning_rate": 1.6346153846153847e-05,
+      "loss": 0.1143,
+      "step": 10564
+    },
+    {
+      "epoch": 96.93,
+      "learning_rate": 1.6298076923076923e-05,
+      "loss": 0.0906,
+      "step": 10565
+    },
+    {
+      "epoch": 96.94,
+      "learning_rate": 1.6250000000000002e-05,
+      "loss": 0.1303,
+      "step": 10566
+    },
+    {
+      "epoch": 96.94,
+      "learning_rate": 1.6201923076923075e-05,
+      "loss": 0.1188,
+      "step": 10567
+    },
+    {
+      "epoch": 96.95,
+      "learning_rate": 1.6153846153846154e-05,
+      "loss": 0.0773,
+      "step": 10568
+    },
+    {
+      "epoch": 96.96,
+      "learning_rate": 1.610576923076923e-05,
+      "loss": 0.1376,
+      "step": 10569
+    },
+    {
+      "epoch": 96.97,
+      "learning_rate": 1.605769230769231e-05,
+      "loss": 0.091,
+      "step": 10570
+    },
+    {
+      "epoch": 96.98,
+      "learning_rate": 1.6009615384615385e-05,
+      "loss": 0.0843,
+      "step": 10571
+    },
+    {
+      "epoch": 96.99,
+      "learning_rate": 1.596153846153846e-05,
+      "loss": 0.1066,
+      "step": 10572
+    },
+    {
+      "epoch": 97.0,
+      "learning_rate": 1.591346153846154e-05,
+      "loss": 0.0846,
+      "step": 10573
+    },
+    {
+      "epoch": 97.01,
+      "learning_rate": 1.5865384615384613e-05,
+      "loss": 0.1501,
+      "step": 10574
+    },
+    {
+      "epoch": 97.02,
+      "learning_rate": 1.5817307692307693e-05,
+      "loss": 0.1045,
+      "step": 10575
+    },
+    {
+      "epoch": 97.03,
+      "learning_rate": 1.576923076923077e-05,
+      "loss": 0.1179,
+      "step": 10576
+    },
+    {
+      "epoch": 97.04,
+      "learning_rate": 1.5721153846153848e-05,
+      "loss": 0.1085,
+      "step": 10577
+    },
+    {
+      "epoch": 97.05,
+      "learning_rate": 1.567307692307692e-05,
+      "loss": 0.1578,
+      "step": 10578
+    },
+    {
+      "epoch": 97.06,
+      "learning_rate": 1.5625e-05,
+      "loss": 0.0982,
+      "step": 10579
+    },
+    {
+      "epoch": 97.06,
+      "learning_rate": 1.5576923076923076e-05,
+      "loss": 0.0764,
+      "step": 10580
+    },
+    {
+      "epoch": 97.07,
+      "learning_rate": 1.5528846153846152e-05,
+      "loss": 0.1044,
+      "step": 10581
+    },
+    {
+      "epoch": 97.08,
+      "learning_rate": 1.5480769230769232e-05,
+      "loss": 0.1084,
+      "step": 10582
+    },
+    {
+      "epoch": 97.09,
+      "learning_rate": 1.5432692307692308e-05,
+      "loss": 0.1081,
+      "step": 10583
+    },
+    {
+      "epoch": 97.1,
+      "learning_rate": 1.5384615384615387e-05,
+      "loss": 0.0863,
+      "step": 10584
+    },
+    {
+      "epoch": 97.11,
+      "learning_rate": 1.5336538461538463e-05,
+      "loss": 0.1192,
+      "step": 10585
+    },
+    {
+      "epoch": 97.12,
+      "learning_rate": 1.528846153846154e-05,
+      "loss": 0.125,
+      "step": 10586
+    },
+    {
+      "epoch": 97.13,
+      "learning_rate": 1.5240384615384615e-05,
+      "loss": 0.0623,
+      "step": 10587
+    },
+    {
+      "epoch": 97.14,
+      "learning_rate": 1.5192307692307693e-05,
+      "loss": 0.0727,
+      "step": 10588
+    },
+    {
+      "epoch": 97.15,
+      "learning_rate": 1.5144230769230769e-05,
+      "loss": 0.0703,
+      "step": 10589
+    },
+    {
+      "epoch": 97.16,
+      "learning_rate": 1.5096153846153847e-05,
+      "loss": 0.113,
+      "step": 10590
+    },
+    {
+      "epoch": 97.17,
+      "learning_rate": 1.5048076923076923e-05,
+      "loss": 0.0971,
+      "step": 10591
+    },
+    {
+      "epoch": 97.17,
+      "learning_rate": 1.5e-05,
+      "loss": 0.0691,
+      "step": 10592
+    },
+    {
+      "epoch": 97.18,
+      "learning_rate": 1.4951923076923076e-05,
+      "loss": 0.1618,
+      "step": 10593
+    },
+    {
+      "epoch": 97.19,
+      "learning_rate": 1.4903846153846156e-05,
+      "loss": 0.086,
+      "step": 10594
+    },
+    {
+      "epoch": 97.2,
+      "learning_rate": 1.4855769230769232e-05,
+      "loss": 0.1307,
+      "step": 10595
+    },
+    {
+      "epoch": 97.21,
+      "learning_rate": 1.4807692307692308e-05,
+      "loss": 0.0762,
+      "step": 10596
+    },
+    {
+      "epoch": 97.22,
+      "learning_rate": 1.4759615384615386e-05,
+      "loss": 0.1148,
+      "step": 10597
+    },
+    {
+      "epoch": 97.23,
+      "learning_rate": 1.4711538461538462e-05,
+      "loss": 0.0723,
+      "step": 10598
+    },
+    {
+      "epoch": 97.24,
+      "learning_rate": 1.466346153846154e-05,
+      "loss": 0.0957,
+      "step": 10599
+    },
+    {
+      "epoch": 97.25,
+      "learning_rate": 1.4615384615384615e-05,
+      "loss": 0.085,
+      "step": 10600
+    },
+    {
+      "epoch": 97.26,
+      "learning_rate": 1.4567307692307693e-05,
+      "loss": 0.1458,
+      "step": 10601
+    },
+    {
+      "epoch": 97.27,
+      "learning_rate": 1.4519230769230769e-05,
+      "loss": 0.1268,
+      "step": 10602
+    },
+    {
+      "epoch": 97.28,
+      "learning_rate": 1.4471153846153845e-05,
+      "loss": 0.1339,
+      "step": 10603
+    },
+    {
+      "epoch": 97.28,
+      "learning_rate": 1.4423076923076924e-05,
+      "loss": 0.1093,
+      "step": 10604
+    },
+    {
+      "epoch": 97.29,
+      "learning_rate": 1.4375e-05,
+      "loss": 0.1123,
+      "step": 10605
+    },
+    {
+      "epoch": 97.3,
+      "learning_rate": 1.4326923076923078e-05,
+      "loss": 0.1232,
+      "step": 10606
+    },
+    {
+      "epoch": 97.31,
+      "learning_rate": 1.4278846153846154e-05,
+      "loss": 0.0946,
+      "step": 10607
+    },
+    {
+      "epoch": 97.32,
+      "learning_rate": 1.4230769230769232e-05,
+      "loss": 0.0789,
+      "step": 10608
+    },
+    {
+      "epoch": 97.33,
+      "learning_rate": 1.4182692307692308e-05,
+      "loss": 0.1177,
+      "step": 10609
+    },
+    {
+      "epoch": 97.34,
+      "learning_rate": 1.4134615384615384e-05,
+      "loss": 0.0587,
+      "step": 10610
+    },
+    {
+      "epoch": 97.35,
+      "learning_rate": 1.4086538461538462e-05,
+      "loss": 0.088,
+      "step": 10611
+    },
+    {
+      "epoch": 97.36,
+      "learning_rate": 1.4038461538461538e-05,
+      "loss": 0.0568,
+      "step": 10612
+    },
+    {
+      "epoch": 97.37,
+      "learning_rate": 1.3990384615384615e-05,
+      "loss": 0.096,
+      "step": 10613
+    },
+    {
+      "epoch": 97.38,
+      "learning_rate": 1.3942307692307693e-05,
+      "loss": 0.1626,
+      "step": 10614
+    },
+    {
+      "epoch": 97.39,
+      "learning_rate": 1.389423076923077e-05,
+      "loss": 0.0847,
+      "step": 10615
+    },
+    {
+      "epoch": 97.39,
+      "learning_rate": 1.3846153846153847e-05,
+      "loss": 0.1289,
+      "step": 10616
+    },
+    {
+      "epoch": 97.4,
+      "learning_rate": 1.3798076923076923e-05,
+      "loss": 0.0813,
+      "step": 10617
+    },
+    {
+      "epoch": 97.41,
+      "learning_rate": 1.375e-05,
+      "loss": 0.0931,
+      "step": 10618
+    },
+    {
+      "epoch": 97.42,
+      "learning_rate": 1.3701923076923077e-05,
+      "loss": 0.1229,
+      "step": 10619
+    },
+    {
+      "epoch": 97.43,
+      "learning_rate": 1.3653846153846154e-05,
+      "loss": 0.1042,
+      "step": 10620
+    },
+    {
+      "epoch": 97.44,
+      "learning_rate": 1.360576923076923e-05,
+      "loss": 0.1809,
+      "step": 10621
+    },
+    {
+      "epoch": 97.45,
+      "learning_rate": 1.3557692307692308e-05,
+      "loss": 0.163,
+      "step": 10622
+    },
+    {
+      "epoch": 97.46,
+      "learning_rate": 1.3509615384615384e-05,
+      "loss": 0.0389,
+      "step": 10623
+    },
+    {
+      "epoch": 97.47,
+      "learning_rate": 1.3461538461538463e-05,
+      "loss": 0.1089,
+      "step": 10624
+    },
+    {
+      "epoch": 97.48,
+      "learning_rate": 1.341346153846154e-05,
+      "loss": 0.0626,
+      "step": 10625
+    },
+    {
+      "epoch": 97.49,
+      "learning_rate": 1.3365384615384615e-05,
+      "loss": 0.0596,
+      "step": 10626
+    },
+    {
+      "epoch": 97.5,
+      "learning_rate": 1.3317307692307693e-05,
+      "loss": 0.0524,
+      "step": 10627
+    },
+    {
+      "epoch": 97.5,
+      "learning_rate": 1.3269230769230769e-05,
+      "loss": 0.1355,
+      "step": 10628
+    },
+    {
+      "epoch": 97.51,
+      "learning_rate": 1.3221153846153847e-05,
+      "loss": 0.1323,
+      "step": 10629
+    },
+    {
+      "epoch": 97.52,
+      "learning_rate": 1.3173076923076923e-05,
+      "loss": 0.1384,
+      "step": 10630
+    },
+    {
+      "epoch": 97.53,
+      "learning_rate": 1.3125e-05,
+      "loss": 0.0982,
+      "step": 10631
+    },
+    {
+      "epoch": 97.54,
+      "learning_rate": 1.3076923076923077e-05,
+      "loss": 0.0849,
+      "step": 10632
+    },
+    {
+      "epoch": 97.55,
+      "learning_rate": 1.3028846153846153e-05,
+      "loss": 0.0658,
+      "step": 10633
+    },
+    {
+      "epoch": 97.56,
+      "learning_rate": 1.2980769230769232e-05,
+      "loss": 0.0823,
+      "step": 10634
+    },
+    {
+      "epoch": 97.57,
+      "learning_rate": 1.2932692307692308e-05,
+      "loss": 0.1178,
+      "step": 10635
+    },
+    {
+      "epoch": 97.58,
+      "learning_rate": 1.2884615384615386e-05,
+      "loss": 0.0725,
+      "step": 10636
+    },
+    {
+      "epoch": 97.59,
+      "learning_rate": 1.2836538461538462e-05,
+      "loss": 0.0602,
+      "step": 10637
+    },
+    {
+      "epoch": 97.6,
+      "learning_rate": 1.278846153846154e-05,
+      "loss": 0.0946,
+      "step": 10638
+    },
+    {
+      "epoch": 97.61,
+      "learning_rate": 1.2740384615384615e-05,
+      "loss": 0.0841,
+      "step": 10639
+    },
+    {
+      "epoch": 97.61,
+      "learning_rate": 1.2692307692307691e-05,
+      "loss": 0.0951,
+      "step": 10640
+    },
+    {
+      "epoch": 97.62,
+      "learning_rate": 1.264423076923077e-05,
+      "loss": 0.121,
+      "step": 10641
+    },
+    {
+      "epoch": 97.63,
+      "learning_rate": 1.2596153846153845e-05,
+      "loss": 0.0961,
+      "step": 10642
+    },
+    {
+      "epoch": 97.64,
+      "learning_rate": 1.2548076923076923e-05,
+      "loss": 0.0875,
+      "step": 10643
+    },
+    {
+      "epoch": 97.65,
+      "learning_rate": 1.25e-05,
+      "loss": 0.0905,
+      "step": 10644
+    },
+    {
+      "epoch": 97.66,
+      "learning_rate": 1.2451923076923078e-05,
+      "loss": 0.087,
+      "step": 10645
+    },
+    {
+      "epoch": 97.67,
+      "learning_rate": 1.2403846153846154e-05,
+      "loss": 0.0773,
+      "step": 10646
+    },
+    {
+      "epoch": 97.68,
+      "learning_rate": 1.2355769230769232e-05,
+      "loss": 0.0643,
+      "step": 10647
+    },
+    {
+      "epoch": 97.69,
+      "learning_rate": 1.2307692307692308e-05,
+      "loss": 0.0863,
+      "step": 10648
+    },
+    {
+      "epoch": 97.7,
+      "learning_rate": 1.2259615384615384e-05,
+      "loss": 0.0755,
+      "step": 10649
+    },
+    {
+      "epoch": 97.71,
+      "learning_rate": 1.2211538461538462e-05,
+      "loss": 0.094,
+      "step": 10650
+    },
+    {
+      "epoch": 97.72,
+      "learning_rate": 1.2163461538461538e-05,
+      "loss": 0.0653,
+      "step": 10651
+    },
+    {
+      "epoch": 97.72,
+      "learning_rate": 1.2115384615384615e-05,
+      "loss": 0.0869,
+      "step": 10652
+    },
+    {
+      "epoch": 97.73,
+      "learning_rate": 1.2067307692307693e-05,
+      "loss": 0.0762,
+      "step": 10653
+    },
+    {
+      "epoch": 97.74,
+      "learning_rate": 1.2019230769230771e-05,
+      "loss": 0.0212,
+      "step": 10654
+    },
+    {
+      "epoch": 97.75,
+      "learning_rate": 1.1971153846153847e-05,
+      "loss": 0.1531,
+      "step": 10655
+    },
+    {
+      "epoch": 97.76,
+      "learning_rate": 1.1923076923076923e-05,
+      "loss": 0.1183,
+      "step": 10656
+    },
+    {
+      "epoch": 97.77,
+      "learning_rate": 1.1875e-05,
+      "loss": 0.0916,
+      "step": 10657
+    },
+    {
+      "epoch": 97.78,
+      "learning_rate": 1.1826923076923077e-05,
+      "loss": 0.086,
+      "step": 10658
+    },
+    {
+      "epoch": 97.79,
+      "learning_rate": 1.1778846153846154e-05,
+      "loss": 0.1233,
+      "step": 10659
+    },
+    {
+      "epoch": 97.8,
+      "learning_rate": 1.173076923076923e-05,
+      "loss": 0.0986,
+      "step": 10660
+    },
+    {
+      "epoch": 97.81,
+      "learning_rate": 1.1682692307692308e-05,
+      "loss": 0.1258,
+      "step": 10661
+    },
+    {
+      "epoch": 97.82,
+      "learning_rate": 1.1634615384615384e-05,
+      "loss": 0.1512,
+      "step": 10662
+    },
+    {
+      "epoch": 97.83,
+      "learning_rate": 1.1586538461538462e-05,
+      "loss": 0.0741,
+      "step": 10663
+    },
+    {
+      "epoch": 97.83,
+      "learning_rate": 1.153846153846154e-05,
+      "loss": 0.1224,
+      "step": 10664
+    },
+    {
+      "epoch": 97.84,
+      "learning_rate": 1.1490384615384616e-05,
+      "loss": 0.1029,
+      "step": 10665
+    },
+    {
+      "epoch": 97.85,
+      "learning_rate": 1.1442307692307693e-05,
+      "loss": 0.1199,
+      "step": 10666
+    },
+    {
+      "epoch": 97.86,
+      "learning_rate": 1.139423076923077e-05,
+      "loss": 0.0765,
+      "step": 10667
+    },
+    {
+      "epoch": 97.87,
+      "learning_rate": 1.1346153846153847e-05,
+      "loss": 0.1132,
+      "step": 10668
+    },
+    {
+      "epoch": 97.88,
+      "learning_rate": 1.1298076923076923e-05,
+      "loss": 0.1571,
+      "step": 10669
+    },
+    {
+      "epoch": 97.89,
+      "learning_rate": 1.1249999999999999e-05,
+      "loss": 0.0853,
+      "step": 10670
+    },
+    {
+      "epoch": 97.9,
+      "learning_rate": 1.1201923076923077e-05,
+      "loss": 0.1177,
+      "step": 10671
+    },
+    {
+      "epoch": 97.91,
+      "learning_rate": 1.1153846153846153e-05,
+      "loss": 0.1029,
+      "step": 10672
+    },
+    {
+      "epoch": 97.92,
+      "learning_rate": 1.1105769230769232e-05,
+      "loss": 0.0737,
+      "step": 10673
+    },
+    {
+      "epoch": 97.93,
+      "learning_rate": 1.1057692307692308e-05,
+      "loss": 0.071,
+      "step": 10674
+    },
+    {
+      "epoch": 97.94,
+      "learning_rate": 1.1009615384615386e-05,
+      "loss": 0.0793,
+      "step": 10675
+    },
+    {
+      "epoch": 97.94,
+      "learning_rate": 1.0961538461538462e-05,
+      "loss": 0.0541,
+      "step": 10676
+    },
+    {
+      "epoch": 97.95,
+      "learning_rate": 1.091346153846154e-05,
+      "loss": 0.0715,
+      "step": 10677
+    },
+    {
+      "epoch": 97.96,
+      "learning_rate": 1.0865384615384616e-05,
+      "loss": 0.0896,
+      "step": 10678
+    },
+    {
+      "epoch": 97.97,
+      "learning_rate": 1.0817307692307692e-05,
+      "loss": 0.1121,
+      "step": 10679
+    },
+    {
+      "epoch": 97.98,
+      "learning_rate": 1.076923076923077e-05,
+      "loss": 0.0806,
+      "step": 10680
+    },
+    {
+      "epoch": 97.99,
+      "learning_rate": 1.0721153846153845e-05,
+      "loss": 0.039,
+      "step": 10681
+    },
+    {
+      "epoch": 98.0,
+      "learning_rate": 1.0673076923076923e-05,
+      "loss": 0.0613,
+      "step": 10682
+    },
+    {
+      "epoch": 98.01,
+      "learning_rate": 1.0625e-05,
+      "loss": 0.1759,
+      "step": 10683
+    },
+    {
+      "epoch": 98.02,
+      "learning_rate": 1.0576923076923078e-05,
+      "loss": 0.1243,
+      "step": 10684
+    },
+    {
+      "epoch": 98.03,
+      "learning_rate": 1.0528846153846154e-05,
+      "loss": 0.1195,
+      "step": 10685
+    },
+    {
+      "epoch": 98.04,
+      "learning_rate": 1.048076923076923e-05,
+      "loss": 0.0845,
+      "step": 10686
+    },
+    {
+      "epoch": 98.05,
+      "learning_rate": 1.0432692307692308e-05,
+      "loss": 0.1433,
+      "step": 10687
+    },
+    {
+      "epoch": 98.06,
+      "learning_rate": 1.0384615384615384e-05,
+      "loss": 0.1045,
+      "step": 10688
+    },
+    {
+      "epoch": 98.06,
+      "learning_rate": 1.0336538461538462e-05,
+      "loss": 0.1342,
+      "step": 10689
+    },
+    {
+      "epoch": 98.07,
+      "learning_rate": 1.0288461538461538e-05,
+      "loss": 0.1045,
+      "step": 10690
+    },
+    {
+      "epoch": 98.08,
+      "learning_rate": 1.0240384615384616e-05,
+      "loss": 0.1304,
+      "step": 10691
+    },
+    {
+      "epoch": 98.09,
+      "learning_rate": 1.0192307692307692e-05,
+      "loss": 0.1098,
+      "step": 10692
+    },
+    {
+      "epoch": 98.1,
+      "learning_rate": 1.014423076923077e-05,
+      "loss": 0.1077,
+      "step": 10693
+    },
+    {
+      "epoch": 98.11,
+      "learning_rate": 1.0096153846153847e-05,
+      "loss": 0.1068,
+      "step": 10694
+    },
+    {
+      "epoch": 98.12,
+      "learning_rate": 1.0048076923076923e-05,
+      "loss": 0.1021,
+      "step": 10695
+    },
+    {
+      "epoch": 98.13,
+      "learning_rate": 1e-05,
+      "loss": 0.0923,
+      "step": 10696
+    },
+    {
+      "epoch": 98.14,
+      "learning_rate": 9.951923076923077e-06,
+      "loss": 0.1061,
+      "step": 10697
+    },
+    {
+      "epoch": 98.15,
+      "learning_rate": 9.903846153846155e-06,
+      "loss": 0.1195,
+      "step": 10698
+    },
+    {
+      "epoch": 98.16,
+      "learning_rate": 9.85576923076923e-06,
+      "loss": 0.0543,
+      "step": 10699
+    },
+    {
+      "epoch": 98.17,
+      "learning_rate": 9.807692307692307e-06,
+      "loss": 0.1391,
+      "step": 10700
+    },
+    {
+      "epoch": 98.17,
+      "learning_rate": 9.759615384615384e-06,
+      "loss": 0.0912,
+      "step": 10701
+    },
+    {
+      "epoch": 98.18,
+      "learning_rate": 9.71153846153846e-06,
+      "loss": 0.0989,
+      "step": 10702
+    },
+    {
+      "epoch": 98.19,
+      "learning_rate": 9.66346153846154e-06,
+      "loss": 0.1448,
+      "step": 10703
+    },
+    {
+      "epoch": 98.2,
+      "learning_rate": 9.615384615384616e-06,
+      "loss": 0.0309,
+      "step": 10704
+    },
+    {
+      "epoch": 98.21,
+      "learning_rate": 9.567307692307693e-06,
+      "loss": 0.0515,
+      "step": 10705
+    },
+    {
+      "epoch": 98.22,
+      "learning_rate": 9.51923076923077e-06,
+      "loss": 0.0897,
+      "step": 10706
+    },
+    {
+      "epoch": 98.23,
+      "learning_rate": 9.471153846153847e-06,
+      "loss": 0.0825,
+      "step": 10707
+    },
+    {
+      "epoch": 98.24,
+      "learning_rate": 9.423076923076923e-06,
+      "loss": 0.07,
+      "step": 10708
+    },
+    {
+      "epoch": 98.25,
+      "learning_rate": 9.375e-06,
+      "loss": 0.0725,
+      "step": 10709
+    },
+    {
+      "epoch": 98.26,
+      "learning_rate": 9.326923076923077e-06,
+      "loss": 0.1557,
+      "step": 10710
+    },
+    {
+      "epoch": 98.27,
+      "learning_rate": 9.278846153846153e-06,
+      "loss": 0.1375,
+      "step": 10711
+    },
+    {
+      "epoch": 98.28,
+      "learning_rate": 9.230769230769232e-06,
+      "loss": 0.1014,
+      "step": 10712
+    },
+    {
+      "epoch": 98.28,
+      "learning_rate": 9.182692307692308e-06,
+      "loss": 0.1071,
+      "step": 10713
+    },
+    {
+      "epoch": 98.29,
+      "learning_rate": 9.134615384615386e-06,
+      "loss": 0.1428,
+      "step": 10714
+    },
+    {
+      "epoch": 98.3,
+      "learning_rate": 9.086538461538462e-06,
+      "loss": 0.1182,
+      "step": 10715
+    },
+    {
+      "epoch": 98.31,
+      "learning_rate": 9.038461538461538e-06,
+      "loss": 0.1034,
+      "step": 10716
+    },
+    {
+      "epoch": 98.32,
+      "learning_rate": 8.990384615384616e-06,
+      "loss": 0.1138,
+      "step": 10717
+    },
+    {
+      "epoch": 98.33,
+      "learning_rate": 8.942307692307692e-06,
+      "loss": 0.1355,
+      "step": 10718
+    },
+    {
+      "epoch": 98.34,
+      "learning_rate": 8.89423076923077e-06,
+      "loss": 0.0971,
+      "step": 10719
+    },
+    {
+      "epoch": 98.35,
+      "learning_rate": 8.846153846153846e-06,
+      "loss": 0.1325,
+      "step": 10720
+    },
+    {
+      "epoch": 98.36,
+      "learning_rate": 8.798076923076923e-06,
+      "loss": 0.1429,
+      "step": 10721
+    },
+    {
+      "epoch": 98.37,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 0.1403,
+      "step": 10722
+    },
+    {
+      "epoch": 98.38,
+      "learning_rate": 8.701923076923077e-06,
+      "loss": 0.1079,
+      "step": 10723
+    },
+    {
+      "epoch": 98.39,
+      "learning_rate": 8.653846153846155e-06,
+      "loss": 0.1083,
+      "step": 10724
+    },
+    {
+      "epoch": 98.39,
+      "learning_rate": 8.60576923076923e-06,
+      "loss": 0.1192,
+      "step": 10725
+    },
+    {
+      "epoch": 98.4,
+      "learning_rate": 8.557692307692308e-06,
+      "loss": 0.1056,
+      "step": 10726
+    },
+    {
+      "epoch": 98.41,
+      "learning_rate": 8.509615384615384e-06,
+      "loss": 0.1136,
+      "step": 10727
+    },
+    {
+      "epoch": 98.42,
+      "learning_rate": 8.461538461538462e-06,
+      "loss": 0.088,
+      "step": 10728
+    },
+    {
+      "epoch": 98.43,
+      "learning_rate": 8.413461538461538e-06,
+      "loss": 0.0983,
+      "step": 10729
+    },
+    {
+      "epoch": 98.44,
+      "learning_rate": 8.365384615384616e-06,
+      "loss": 0.1053,
+      "step": 10730
+    },
+    {
+      "epoch": 98.45,
+      "learning_rate": 8.317307692307692e-06,
+      "loss": 0.0869,
+      "step": 10731
+    },
+    {
+      "epoch": 98.46,
+      "learning_rate": 8.26923076923077e-06,
+      "loss": 0.071,
+      "step": 10732
+    },
+    {
+      "epoch": 98.47,
+      "learning_rate": 8.221153846153847e-06,
+      "loss": 0.1736,
+      "step": 10733
+    },
+    {
+      "epoch": 98.48,
+      "learning_rate": 8.173076923076923e-06,
+      "loss": 0.0977,
+      "step": 10734
+    },
+    {
+      "epoch": 98.49,
+      "learning_rate": 8.125000000000001e-06,
+      "loss": 0.0914,
+      "step": 10735
+    },
+    {
+      "epoch": 98.5,
+      "learning_rate": 8.076923076923077e-06,
+      "loss": 0.0306,
+      "step": 10736
+    },
+    {
+      "epoch": 98.5,
+      "learning_rate": 8.028846153846155e-06,
+      "loss": 0.1421,
+      "step": 10737
+    },
+    {
+      "epoch": 98.51,
+      "learning_rate": 7.98076923076923e-06,
+      "loss": 0.1401,
+      "step": 10738
+    },
+    {
+      "epoch": 98.52,
+      "learning_rate": 7.932692307692307e-06,
+      "loss": 0.1104,
+      "step": 10739
+    },
+    {
+      "epoch": 98.53,
+      "learning_rate": 7.884615384615384e-06,
+      "loss": 0.1197,
+      "step": 10740
+    },
+    {
+      "epoch": 98.54,
+      "learning_rate": 7.83653846153846e-06,
+      "loss": 0.133,
+      "step": 10741
+    },
+    {
+      "epoch": 98.55,
+      "learning_rate": 7.788461538461538e-06,
+      "loss": 0.1172,
+      "step": 10742
+    },
+    {
+      "epoch": 98.56,
+      "learning_rate": 7.740384615384616e-06,
+      "loss": 0.1163,
+      "step": 10743
+    },
+    {
+      "epoch": 98.57,
+      "learning_rate": 7.692307692307694e-06,
+      "loss": 0.1027,
+      "step": 10744
+    },
+    {
+      "epoch": 98.58,
+      "learning_rate": 7.64423076923077e-06,
+      "loss": 0.1016,
+      "step": 10745
+    },
+    {
+      "epoch": 98.59,
+      "learning_rate": 7.5961538461538465e-06,
+      "loss": 0.123,
+      "step": 10746
+    },
+    {
+      "epoch": 98.6,
+      "learning_rate": 7.548076923076923e-06,
+      "loss": 0.135,
+      "step": 10747
+    },
+    {
+      "epoch": 98.61,
+      "learning_rate": 7.5e-06,
+      "loss": 0.1509,
+      "step": 10748
+    },
+    {
+      "epoch": 98.61,
+      "learning_rate": 7.451923076923078e-06,
+      "loss": 0.0623,
+      "step": 10749
+    },
+    {
+      "epoch": 98.62,
+      "learning_rate": 7.403846153846154e-06,
+      "loss": 0.0884,
+      "step": 10750
+    },
+    {
+      "epoch": 98.63,
+      "learning_rate": 7.355769230769231e-06,
+      "loss": 0.1031,
+      "step": 10751
+    },
+    {
+      "epoch": 98.64,
+      "learning_rate": 7.307692307692308e-06,
+      "loss": 0.0448,
+      "step": 10752
+    },
+    {
+      "epoch": 98.65,
+      "learning_rate": 7.2596153846153845e-06,
+      "loss": 0.0493,
+      "step": 10753
+    },
+    {
+      "epoch": 98.66,
+      "learning_rate": 7.211538461538462e-06,
+      "loss": 0.0913,
+      "step": 10754
+    },
+    {
+      "epoch": 98.67,
+      "learning_rate": 7.163461538461539e-06,
+      "loss": 0.0801,
+      "step": 10755
+    },
+    {
+      "epoch": 98.68,
+      "learning_rate": 7.115384615384616e-06,
+      "loss": 0.1115,
+      "step": 10756
+    },
+    {
+      "epoch": 98.69,
+      "learning_rate": 7.067307692307692e-06,
+      "loss": 0.0649,
+      "step": 10757
+    },
+    {
+      "epoch": 98.7,
+      "learning_rate": 7.019230769230769e-06,
+      "loss": 0.1154,
+      "step": 10758
+    },
+    {
+      "epoch": 98.71,
+      "learning_rate": 6.9711538461538465e-06,
+      "loss": 0.0921,
+      "step": 10759
+    },
+    {
+      "epoch": 98.72,
+      "learning_rate": 6.923076923076923e-06,
+      "loss": 0.1203,
+      "step": 10760
+    },
+    {
+      "epoch": 98.72,
+      "learning_rate": 6.875e-06,
+      "loss": 0.07,
+      "step": 10761
+    },
+    {
+      "epoch": 98.73,
+      "learning_rate": 6.826923076923077e-06,
+      "loss": 0.0516,
+      "step": 10762
+    },
+    {
+      "epoch": 98.74,
+      "learning_rate": 6.778846153846154e-06,
+      "loss": 0.0659,
+      "step": 10763
+    },
+    {
+      "epoch": 98.75,
+      "learning_rate": 6.730769230769232e-06,
+      "loss": 0.1547,
+      "step": 10764
+    },
+    {
+      "epoch": 98.76,
+      "learning_rate": 6.682692307692308e-06,
+      "loss": 0.0945,
+      "step": 10765
+    },
+    {
+      "epoch": 98.77,
+      "learning_rate": 6.6346153846153846e-06,
+      "loss": 0.1401,
+      "step": 10766
+    },
+    {
+      "epoch": 98.78,
+      "learning_rate": 6.586538461538461e-06,
+      "loss": 0.1126,
+      "step": 10767
+    },
+    {
+      "epoch": 98.79,
+      "learning_rate": 6.538461538461538e-06,
+      "loss": 0.1139,
+      "step": 10768
+    },
+    {
+      "epoch": 98.8,
+      "learning_rate": 6.490384615384616e-06,
+      "loss": 0.1408,
+      "step": 10769
+    },
+    {
+      "epoch": 98.81,
+      "learning_rate": 6.442307692307693e-06,
+      "loss": 0.112,
+      "step": 10770
+    },
+    {
+      "epoch": 98.82,
+      "learning_rate": 6.39423076923077e-06,
+      "loss": 0.0731,
+      "step": 10771
+    },
+    {
+      "epoch": 98.83,
+      "learning_rate": 6.346153846153846e-06,
+      "loss": 0.1398,
+      "step": 10772
+    },
+    {
+      "epoch": 98.83,
+      "learning_rate": 6.298076923076923e-06,
+      "loss": 0.0956,
+      "step": 10773
+    },
+    {
+      "epoch": 98.84,
+      "learning_rate": 6.25e-06,
+      "loss": 0.1047,
+      "step": 10774
+    },
+    {
+      "epoch": 98.85,
+      "learning_rate": 6.201923076923077e-06,
+      "loss": 0.0632,
+      "step": 10775
+    },
+    {
+      "epoch": 98.86,
+      "learning_rate": 6.153846153846154e-06,
+      "loss": 0.0971,
+      "step": 10776
+    },
+    {
+      "epoch": 98.87,
+      "learning_rate": 6.105769230769231e-06,
+      "loss": 0.1108,
+      "step": 10777
+    },
+    {
+      "epoch": 98.88,
+      "learning_rate": 6.057692307692308e-06,
+      "loss": 0.154,
+      "step": 10778
+    },
+    {
+      "epoch": 98.89,
+      "learning_rate": 6.0096153846153855e-06,
+      "loss": 0.0705,
+      "step": 10779
+    },
+    {
+      "epoch": 98.9,
+      "learning_rate": 5.9615384615384615e-06,
+      "loss": 0.1322,
+      "step": 10780
+    },
+    {
+      "epoch": 98.91,
+      "learning_rate": 5.913461538461538e-06,
+      "loss": 0.0784,
+      "step": 10781
+    },
+    {
+      "epoch": 98.92,
+      "learning_rate": 5.865384615384615e-06,
+      "loss": 0.0775,
+      "step": 10782
+    },
+    {
+      "epoch": 98.93,
+      "learning_rate": 5.817307692307692e-06,
+      "loss": 0.1013,
+      "step": 10783
+    },
+    {
+      "epoch": 98.94,
+      "learning_rate": 5.76923076923077e-06,
+      "loss": 0.0877,
+      "step": 10784
+    },
+    {
+      "epoch": 98.94,
+      "learning_rate": 5.721153846153847e-06,
+      "loss": 0.1856,
+      "step": 10785
+    },
+    {
+      "epoch": 98.95,
+      "learning_rate": 5.6730769230769235e-06,
+      "loss": 0.0554,
+      "step": 10786
+    },
+    {
+      "epoch": 98.96,
+      "learning_rate": 5.6249999999999995e-06,
+      "loss": 0.0956,
+      "step": 10787
+    },
+    {
+      "epoch": 98.97,
+      "learning_rate": 5.576923076923076e-06,
+      "loss": 0.0672,
+      "step": 10788
+    },
+    {
+      "epoch": 98.98,
+      "learning_rate": 5.528846153846154e-06,
+      "loss": 0.1047,
+      "step": 10789
+    },
+    {
+      "epoch": 98.99,
+      "learning_rate": 5.480769230769231e-06,
+      "loss": 0.0257,
+      "step": 10790
+    },
+    {
+      "epoch": 99.0,
+      "learning_rate": 5.432692307692308e-06,
+      "loss": 0.1263,
+      "step": 10791
+    },
+    {
+      "epoch": 99.01,
+      "learning_rate": 5.384615384615385e-06,
+      "loss": 0.1653,
+      "step": 10792
+    },
+    {
+      "epoch": 99.02,
+      "learning_rate": 5.3365384615384615e-06,
+      "loss": 0.1337,
+      "step": 10793
+    },
+    {
+      "epoch": 99.03,
+      "learning_rate": 5.288461538461539e-06,
+      "loss": 0.1166,
+      "step": 10794
+    },
+    {
+      "epoch": 99.04,
+      "learning_rate": 5.240384615384615e-06,
+      "loss": 0.1185,
+      "step": 10795
+    },
+    {
+      "epoch": 99.05,
+      "learning_rate": 5.192307692307692e-06,
+      "loss": 0.1101,
+      "step": 10796
+    },
+    {
+      "epoch": 99.06,
+      "learning_rate": 5.144230769230769e-06,
+      "loss": 0.1459,
+      "step": 10797
+    },
+    {
+      "epoch": 99.06,
+      "learning_rate": 5.096153846153846e-06,
+      "loss": 0.1221,
+      "step": 10798
+    },
+    {
+      "epoch": 99.07,
+      "learning_rate": 5.0480769230769235e-06,
+      "loss": 0.106,
+      "step": 10799
+    },
+    {
+      "epoch": 99.08,
+      "learning_rate": 5e-06,
+      "loss": 0.1366,
+      "step": 10800
+    },
+    {
+      "epoch": 99.09,
+      "learning_rate": 4.951923076923077e-06,
+      "loss": 0.0836,
+      "step": 10801
+    },
+    {
+      "epoch": 99.1,
+      "learning_rate": 4.903846153846153e-06,
+      "loss": 0.1291,
+      "step": 10802
+    },
+    {
+      "epoch": 99.11,
+      "learning_rate": 4.85576923076923e-06,
+      "loss": 0.1328,
+      "step": 10803
+    },
+    {
+      "epoch": 99.12,
+      "learning_rate": 4.807692307692308e-06,
+      "loss": 0.0745,
+      "step": 10804
+    },
+    {
+      "epoch": 99.13,
+      "learning_rate": 4.759615384615385e-06,
+      "loss": 0.1366,
+      "step": 10805
+    },
+    {
+      "epoch": 99.14,
+      "learning_rate": 4.711538461538462e-06,
+      "loss": 0.0914,
+      "step": 10806
+    },
+    {
+      "epoch": 99.15,
+      "learning_rate": 4.6634615384615384e-06,
+      "loss": 0.1165,
+      "step": 10807
+    },
+    {
+      "epoch": 99.16,
+      "learning_rate": 4.615384615384616e-06,
+      "loss": 0.0281,
+      "step": 10808
+    },
+    {
+      "epoch": 99.17,
+      "learning_rate": 4.567307692307693e-06,
+      "loss": 0.1348,
+      "step": 10809
+    },
+    {
+      "epoch": 99.17,
+      "learning_rate": 4.519230769230769e-06,
+      "loss": 0.0689,
+      "step": 10810
+    },
+    {
+      "epoch": 99.18,
+      "learning_rate": 4.471153846153846e-06,
+      "loss": 0.1174,
+      "step": 10811
+    },
+    {
+      "epoch": 99.19,
+      "learning_rate": 4.423076923076923e-06,
+      "loss": 0.079,
+      "step": 10812
+    },
+    {
+      "epoch": 99.2,
+      "learning_rate": 4.3750000000000005e-06,
+      "loss": 0.1269,
+      "step": 10813
+    },
+    {
+      "epoch": 99.21,
+      "learning_rate": 4.326923076923077e-06,
+      "loss": 0.0556,
+      "step": 10814
+    },
+    {
+      "epoch": 99.22,
+      "learning_rate": 4.278846153846154e-06,
+      "loss": 0.0863,
+      "step": 10815
+    },
+    {
+      "epoch": 99.23,
+      "learning_rate": 4.230769230769231e-06,
+      "loss": 0.0922,
+      "step": 10816
+    },
+    {
+      "epoch": 99.24,
+      "learning_rate": 4.182692307692308e-06,
+      "loss": 0.0968,
+      "step": 10817
+    },
+    {
+      "epoch": 99.25,
+      "learning_rate": 4.134615384615385e-06,
+      "loss": 0.077,
+      "step": 10818
+    },
+    {
+      "epoch": 99.26,
+      "learning_rate": 4.086538461538462e-06,
+      "loss": 0.1543,
+      "step": 10819
+    },
+    {
+      "epoch": 99.27,
+      "learning_rate": 4.0384615384615385e-06,
+      "loss": 0.1635,
+      "step": 10820
+    },
+    {
+      "epoch": 99.28,
+      "learning_rate": 3.990384615384615e-06,
+      "loss": 0.0929,
+      "step": 10821
+    },
+    {
+      "epoch": 99.28,
+      "learning_rate": 3.942307692307692e-06,
+      "loss": 0.1135,
+      "step": 10822
+    },
+    {
+      "epoch": 99.29,
+      "learning_rate": 3.894230769230769e-06,
+      "loss": 0.1106,
+      "step": 10823
+    },
+    {
+      "epoch": 99.3,
+      "learning_rate": 3.846153846153847e-06,
+      "loss": 0.1107,
+      "step": 10824
+    },
+    {
+      "epoch": 99.31,
+      "learning_rate": 3.7980769230769232e-06,
+      "loss": 0.1289,
+      "step": 10825
+    },
+    {
+      "epoch": 99.32,
+      "learning_rate": 3.75e-06,
+      "loss": 0.0919,
+      "step": 10826
+    },
+    {
+      "epoch": 99.33,
+      "learning_rate": 3.701923076923077e-06,
+      "loss": 0.1318,
+      "step": 10827
+    },
+    {
+      "epoch": 99.34,
+      "learning_rate": 3.653846153846154e-06,
+      "loss": 0.0655,
+      "step": 10828
+    },
+    {
+      "epoch": 99.35,
+      "learning_rate": 3.605769230769231e-06,
+      "loss": 0.0838,
+      "step": 10829
+    },
+    {
+      "epoch": 99.36,
+      "learning_rate": 3.557692307692308e-06,
+      "loss": 0.0583,
+      "step": 10830
+    },
+    {
+      "epoch": 99.37,
+      "learning_rate": 3.5096153846153844e-06,
+      "loss": 0.0635,
+      "step": 10831
+    },
+    {
+      "epoch": 99.38,
+      "learning_rate": 3.4615384615384617e-06,
+      "loss": 0.1053,
+      "step": 10832
+    },
+    {
+      "epoch": 99.39,
+      "learning_rate": 3.4134615384615386e-06,
+      "loss": 0.1396,
+      "step": 10833
+    },
+    {
+      "epoch": 99.39,
+      "learning_rate": 3.365384615384616e-06,
+      "loss": 0.0984,
+      "step": 10834
+    },
+    {
+      "epoch": 99.4,
+      "learning_rate": 3.3173076923076923e-06,
+      "loss": 0.0749,
+      "step": 10835
+    },
+    {
+      "epoch": 99.41,
+      "learning_rate": 3.269230769230769e-06,
+      "loss": 0.1174,
+      "step": 10836
+    },
+    {
+      "epoch": 99.42,
+      "learning_rate": 3.2211538461538464e-06,
+      "loss": 0.0928,
+      "step": 10837
+    },
+    {
+      "epoch": 99.43,
+      "learning_rate": 3.173076923076923e-06,
+      "loss": 0.0549,
+      "step": 10838
+    },
+    {
+      "epoch": 99.44,
+      "learning_rate": 3.125e-06,
+      "loss": 0.0872,
+      "step": 10839
+    },
+    {
+      "epoch": 99.45,
+      "learning_rate": 3.076923076923077e-06,
+      "loss": 0.0704,
+      "step": 10840
+    },
+    {
+      "epoch": 99.46,
+      "learning_rate": 3.028846153846154e-06,
+      "loss": 0.0398,
+      "step": 10841
+    },
+    {
+      "epoch": 99.47,
+      "learning_rate": 2.9807692307692307e-06,
+      "loss": 0.094,
+      "step": 10842
+    },
+    {
+      "epoch": 99.48,
+      "learning_rate": 2.9326923076923076e-06,
+      "loss": 0.0902,
+      "step": 10843
+    },
+    {
+      "epoch": 99.49,
+      "learning_rate": 2.884615384615385e-06,
+      "loss": 0.0994,
+      "step": 10844
+    },
+    {
+      "epoch": 99.5,
+      "learning_rate": 2.8365384615384617e-06,
+      "loss": 0.0513,
+      "step": 10845
+    },
+    {
+      "epoch": 99.5,
+      "learning_rate": 2.788461538461538e-06,
+      "loss": 0.2055,
+      "step": 10846
+    },
+    {
+      "epoch": 99.51,
+      "learning_rate": 2.7403846153846155e-06,
+      "loss": 0.1316,
+      "step": 10847
+    },
+    {
+      "epoch": 99.52,
+      "learning_rate": 2.6923076923076923e-06,
+      "loss": 0.1219,
+      "step": 10848
+    },
+    {
+      "epoch": 99.53,
+      "learning_rate": 2.6442307692307696e-06,
+      "loss": 0.1132,
+      "step": 10849
+    },
+    {
+      "epoch": 99.54,
+      "learning_rate": 2.596153846153846e-06,
+      "loss": 0.1072,
+      "step": 10850
+    },
+    {
+      "epoch": 99.55,
+      "learning_rate": 2.548076923076923e-06,
+      "loss": 0.0983,
+      "step": 10851
+    },
+    {
+      "epoch": 99.56,
+      "learning_rate": 2.5e-06,
+      "loss": 0.1457,
+      "step": 10852
+    },
+    {
+      "epoch": 99.57,
+      "learning_rate": 2.4519230769230766e-06,
+      "loss": 0.0745,
+      "step": 10853
+    },
+    {
+      "epoch": 99.58,
+      "learning_rate": 2.403846153846154e-06,
+      "loss": 0.0876,
+      "step": 10854
+    },
+    {
+      "epoch": 99.59,
+      "learning_rate": 2.355769230769231e-06,
+      "loss": 0.1684,
+      "step": 10855
+    },
+    {
+      "epoch": 99.6,
+      "learning_rate": 2.307692307692308e-06,
+      "loss": 0.0859,
+      "step": 10856
+    },
+    {
+      "epoch": 99.61,
+      "learning_rate": 2.2596153846153845e-06,
+      "loss": 0.1476,
+      "step": 10857
+    },
+    {
+      "epoch": 99.61,
+      "learning_rate": 2.2115384615384614e-06,
+      "loss": 0.0793,
+      "step": 10858
+    },
+    {
+      "epoch": 99.62,
+      "learning_rate": 2.1634615384615387e-06,
+      "loss": 0.1262,
+      "step": 10859
+    },
+    {
+      "epoch": 99.63,
+      "learning_rate": 2.1153846153846155e-06,
+      "loss": 0.1145,
+      "step": 10860
+    },
+    {
+      "epoch": 99.64,
+      "learning_rate": 2.0673076923076924e-06,
+      "loss": 0.0778,
+      "step": 10861
+    },
+    {
+      "epoch": 99.65,
+      "learning_rate": 2.0192307692307692e-06,
+      "loss": 0.067,
+      "step": 10862
+    },
+    {
+      "epoch": 99.66,
+      "learning_rate": 1.971153846153846e-06,
+      "loss": 0.0686,
+      "step": 10863
+    },
+    {
+      "epoch": 99.67,
+      "learning_rate": 1.9230769230769234e-06,
+      "loss": 0.0865,
+      "step": 10864
+    },
+    {
+      "epoch": 99.68,
+      "learning_rate": 1.875e-06,
+      "loss": 0.1209,
+      "step": 10865
+    },
+    {
+      "epoch": 99.69,
+      "learning_rate": 1.826923076923077e-06,
+      "loss": 0.0729,
+      "step": 10866
+    },
+    {
+      "epoch": 99.7,
+      "learning_rate": 1.778846153846154e-06,
+      "loss": 0.1472,
+      "step": 10867
+    },
+    {
+      "epoch": 99.71,
+      "learning_rate": 1.7307692307692308e-06,
+      "loss": 0.0516,
+      "step": 10868
+    },
+    {
+      "epoch": 99.72,
+      "learning_rate": 1.682692307692308e-06,
+      "loss": 0.0901,
+      "step": 10869
+    },
+    {
+      "epoch": 99.72,
+      "learning_rate": 1.6346153846153846e-06,
+      "loss": 0.0505,
+      "step": 10870
+    },
+    {
+      "epoch": 99.73,
+      "learning_rate": 1.5865384615384614e-06,
+      "loss": 0.0629,
+      "step": 10871
+    },
+    {
+      "epoch": 99.74,
+      "learning_rate": 1.5384615384615385e-06,
+      "loss": 0.2219,
+      "step": 10872
+    },
+    {
+      "epoch": 99.75,
+      "learning_rate": 1.4903846153846154e-06,
+      "loss": 0.1274,
+      "step": 10873
+    },
+    {
+      "epoch": 99.76,
+      "learning_rate": 1.4423076923076924e-06,
+      "loss": 0.1312,
+      "step": 10874
+    },
+    {
+      "epoch": 99.77,
+      "learning_rate": 1.394230769230769e-06,
+      "loss": 0.1167,
+      "step": 10875
+    },
+    {
+      "epoch": 99.78,
+      "learning_rate": 1.3461538461538462e-06,
+      "loss": 0.0919,
+      "step": 10876
+    },
+    {
+      "epoch": 99.79,
+      "learning_rate": 1.298076923076923e-06,
+      "loss": 0.17,
+      "step": 10877
+    },
+    {
+      "epoch": 99.8,
+      "learning_rate": 1.25e-06,
+      "loss": 0.1601,
+      "step": 10878
+    },
+    {
+      "epoch": 99.81,
+      "learning_rate": 1.201923076923077e-06,
+      "loss": 0.0664,
+      "step": 10879
+    },
+    {
+      "epoch": 99.82,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 0.08,
+      "step": 10880
+    },
+    {
+      "epoch": 99.83,
+      "learning_rate": 1.1057692307692307e-06,
+      "loss": 0.1484,
+      "step": 10881
+    },
+    {
+      "epoch": 99.83,
+      "learning_rate": 1.0576923076923078e-06,
+      "loss": 0.1291,
+      "step": 10882
+    },
+    {
+      "epoch": 99.84,
+      "learning_rate": 1.0096153846153846e-06,
+      "loss": 0.1387,
+      "step": 10883
+    },
+    {
+      "epoch": 99.85,
+      "learning_rate": 9.615384615384617e-07,
+      "loss": 0.1047,
+      "step": 10884
+    },
+    {
+      "epoch": 99.86,
+      "learning_rate": 9.134615384615385e-07,
+      "loss": 0.1096,
+      "step": 10885
+    },
+    {
+      "epoch": 99.87,
+      "learning_rate": 8.653846153846154e-07,
+      "loss": 0.1193,
+      "step": 10886
+    },
+    {
+      "epoch": 99.88,
+      "learning_rate": 8.173076923076923e-07,
+      "loss": 0.1043,
+      "step": 10887
+    },
+    {
+      "epoch": 99.89,
+      "learning_rate": 7.692307692307693e-07,
+      "loss": 0.178,
+      "step": 10888
+    },
+    {
+      "epoch": 99.9,
+      "learning_rate": 7.211538461538462e-07,
+      "loss": 0.0839,
+      "step": 10889
+    },
+    {
+      "epoch": 99.91,
+      "learning_rate": 6.730769230769231e-07,
+      "loss": 0.0829,
+      "step": 10890
+    },
+    {
+      "epoch": 99.92,
+      "learning_rate": 6.25e-07,
+      "loss": 0.1001,
+      "step": 10891
+    },
+    {
+      "epoch": 99.93,
+      "learning_rate": 5.76923076923077e-07,
+      "loss": 0.0748,
+      "step": 10892
+    },
+    {
+      "epoch": 99.94,
+      "learning_rate": 5.288461538461539e-07,
+      "loss": 0.0612,
+      "step": 10893
+    },
+    {
+      "epoch": 99.94,
+      "learning_rate": 4.807692307692308e-07,
+      "loss": 0.1034,
+      "step": 10894
+    },
+    {
+      "epoch": 99.95,
+      "learning_rate": 4.326923076923077e-07,
+      "loss": 0.0369,
+      "step": 10895
+    },
+    {
+      "epoch": 99.96,
+      "learning_rate": 3.8461538461538463e-07,
+      "loss": 0.1068,
+      "step": 10896
+    },
+    {
+      "epoch": 99.97,
+      "learning_rate": 3.3653846153846154e-07,
+      "loss": 0.0879,
+      "step": 10897
+    },
+    {
+      "epoch": 99.98,
+      "learning_rate": 2.884615384615385e-07,
+      "loss": 0.1357,
+      "step": 10898
+    },
+    {
+      "epoch": 99.99,
+      "learning_rate": 2.403846153846154e-07,
+      "loss": 0.0615,
+      "step": 10899
+    },
+    {
+      "epoch": 100.0,
+      "learning_rate": 1.9230769230769231e-07,
+      "loss": 0.0967,
+      "step": 10900
+    },
+    {
+      "epoch": 100.0,
+      "step": 10900,
+      "total_flos": 1.3360714276118677e+20,
+      "train_loss": 0.44451899920906357,
+      "train_runtime": 28212.0393,
+      "train_samples_per_second": 12.328,
+      "train_steps_per_second": 0.386
     }
   ],
-  "max_steps": 109,
-  "num_train_epochs": 1,
-  "total_flos": 1.3366257233080689e+18,
+  "max_steps": 10900,
+  "num_train_epochs": 100,
+  "total_flos": 1.3360714276118677e+20,
   "trial_name": null,
   "trial_params": null
 }