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": 50.0,
-  "global_step": 2750,
+  "epoch": 75.0,
+  "global_step": 4125,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -58,529 +58,529 @@
     {
       "epoch": 0.16,
       "learning_rate": 7e-06,
-      "loss": 14.6655,
+      "loss": 14.6656,
       "step": 9
     },
     {
       "epoch": 0.18,
       "learning_rate": 8e-06,
-      "loss": 14.1778,
+      "loss": 14.1779,
       "step": 10
     },
     {
       "epoch": 0.2,
       "learning_rate": 9e-06,
-      "loss": 13.7954,
+      "loss": 13.7956,
       "step": 11
     },
     {
       "epoch": 0.22,
       "learning_rate": 1e-05,
-      "loss": 13.6214,
+      "loss": 13.6213,
       "step": 12
     },
     {
       "epoch": 0.24,
       "learning_rate": 1.1e-05,
-      "loss": 13.9367,
+      "loss": 13.937,
       "step": 13
     },
     {
       "epoch": 0.25,
       "learning_rate": 1.2e-05,
-      "loss": 14.8974,
+      "loss": 14.8979,
       "step": 14
     },
     {
       "epoch": 0.27,
-      "learning_rate": 1.2e-05,
-      "loss": 14.516,
+      "learning_rate": 1.3e-05,
+      "loss": 14.5158,
       "step": 15
     },
     {
       "epoch": 0.29,
-      "learning_rate": 1.3e-05,
-      "loss": 14.7569,
+      "learning_rate": 1.4e-05,
+      "loss": 14.6739,
       "step": 16
     },
     {
       "epoch": 0.31,
-      "learning_rate": 1.4e-05,
-      "loss": 13.3187,
+      "learning_rate": 1.5e-05,
+      "loss": 13.2414,
       "step": 17
     },
     {
       "epoch": 0.33,
-      "learning_rate": 1.5e-05,
-      "loss": 12.677,
+      "learning_rate": 1.6e-05,
+      "loss": 12.5534,
       "step": 18
     },
     {
       "epoch": 0.35,
-      "learning_rate": 1.6e-05,
-      "loss": 14.3528,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 14.2319,
       "step": 19
     },
     {
       "epoch": 0.36,
-      "learning_rate": 1.7000000000000003e-05,
-      "loss": 15.1415,
+      "learning_rate": 1.8e-05,
+      "loss": 15.0286,
       "step": 20
     },
     {
       "epoch": 0.38,
-      "learning_rate": 1.8e-05,
-      "loss": 13.8603,
+      "learning_rate": 1.9e-05,
+      "loss": 13.6824,
       "step": 21
     },
     {
       "epoch": 0.4,
-      "learning_rate": 1.9e-05,
-      "loss": 13.0224,
+      "learning_rate": 2e-05,
+      "loss": 12.8093,
       "step": 22
     },
     {
       "epoch": 0.42,
-      "learning_rate": 2e-05,
-      "loss": 13.2831,
+      "learning_rate": 2.1000000000000002e-05,
+      "loss": 13.0109,
       "step": 23
     },
     {
       "epoch": 0.44,
-      "learning_rate": 2.1000000000000002e-05,
-      "loss": 11.758,
+      "learning_rate": 2.2e-05,
+      "loss": 11.4644,
       "step": 24
     },
     {
       "epoch": 0.45,
       "learning_rate": 2.2e-05,
-      "loss": 13.2083,
+      "loss": 12.8117,
       "step": 25
     },
     {
       "epoch": 0.47,
       "learning_rate": 2.3e-05,
-      "loss": 13.4836,
+      "loss": 13.4644,
       "step": 26
     },
     {
       "epoch": 0.49,
       "learning_rate": 2.4e-05,
-      "loss": 12.6254,
+      "loss": 12.6084,
       "step": 27
     },
     {
       "epoch": 0.51,
       "learning_rate": 2.5e-05,
-      "loss": 11.0464,
+      "loss": 11.0357,
       "step": 28
     },
     {
       "epoch": 0.53,
       "learning_rate": 2.6e-05,
-      "loss": 11.193,
+      "loss": 11.1842,
       "step": 29
     },
     {
       "epoch": 0.55,
       "learning_rate": 2.7e-05,
-      "loss": 9.9618,
+      "loss": 9.9558,
       "step": 30
     },
     {
       "epoch": 0.56,
       "learning_rate": 2.8e-05,
-      "loss": 9.6118,
+      "loss": 9.6082,
       "step": 31
     },
     {
       "epoch": 0.58,
       "learning_rate": 2.9000000000000004e-05,
-      "loss": 10.7847,
+      "loss": 10.7778,
       "step": 32
     },
     {
       "epoch": 0.6,
       "learning_rate": 3e-05,
-      "loss": 9.4663,
+      "loss": 9.4578,
       "step": 33
     },
     {
       "epoch": 0.62,
       "learning_rate": 3.1e-05,
-      "loss": 8.4026,
+      "loss": 8.3956,
       "step": 34
     },
     {
       "epoch": 0.64,
       "learning_rate": 3.2e-05,
-      "loss": 8.0046,
+      "loss": 8.0003,
       "step": 35
     },
     {
       "epoch": 0.65,
       "learning_rate": 3.3e-05,
-      "loss": 7.3806,
+      "loss": 7.379,
       "step": 36
     },
     {
       "epoch": 0.67,
       "learning_rate": 3.4000000000000007e-05,
-      "loss": 6.7957,
+      "loss": 6.7911,
       "step": 37
     },
     {
       "epoch": 0.69,
       "learning_rate": 3.5000000000000004e-05,
-      "loss": 7.5041,
+      "loss": 7.4992,
       "step": 38
     },
     {
       "epoch": 0.71,
       "learning_rate": 3.6e-05,
-      "loss": 7.0438,
+      "loss": 7.0406,
       "step": 39
     },
     {
       "epoch": 0.73,
       "learning_rate": 3.7e-05,
-      "loss": 6.9028,
+      "loss": 6.8944,
       "step": 40
     },
     {
       "epoch": 0.75,
       "learning_rate": 3.8e-05,
-      "loss": 6.5476,
+      "loss": 6.5413,
       "step": 41
     },
     {
       "epoch": 0.76,
       "learning_rate": 3.9e-05,
-      "loss": 5.9554,
+      "loss": 5.9524,
       "step": 42
     },
     {
       "epoch": 0.78,
       "learning_rate": 4e-05,
-      "loss": 5.89,
+      "loss": 5.8864,
       "step": 43
     },
     {
       "epoch": 0.8,
       "learning_rate": 4.1e-05,
-      "loss": 6.9094,
+      "loss": 6.908,
       "step": 44
     },
     {
       "epoch": 0.82,
       "learning_rate": 4.2000000000000004e-05,
-      "loss": 5.7892,
+      "loss": 5.7879,
       "step": 45
     },
     {
       "epoch": 0.84,
       "learning_rate": 4.2999999999999995e-05,
-      "loss": 5.6216,
+      "loss": 5.6158,
       "step": 46
     },
     {
       "epoch": 0.85,
       "learning_rate": 4.4e-05,
-      "loss": 5.3883,
+      "loss": 5.3846,
       "step": 47
     },
     {
       "epoch": 0.87,
       "learning_rate": 4.4999999999999996e-05,
-      "loss": 5.0965,
+      "loss": 5.0943,
       "step": 48
     },
     {
       "epoch": 0.89,
       "learning_rate": 4.6e-05,
-      "loss": 5.1571,
+      "loss": 5.1507,
       "step": 49
     },
     {
       "epoch": 0.91,
       "learning_rate": 4.7000000000000004e-05,
-      "loss": 5.1086,
+      "loss": 5.1055,
       "step": 50
     },
     {
       "epoch": 0.93,
       "learning_rate": 4.8e-05,
-      "loss": 5.1846,
+      "loss": 5.1817,
       "step": 51
     },
     {
       "epoch": 0.95,
       "learning_rate": 4.9000000000000005e-05,
-      "loss": 5.0277,
+      "loss": 5.0229,
       "step": 52
     },
     {
       "epoch": 0.96,
       "learning_rate": 5e-05,
-      "loss": 4.7516,
+      "loss": 4.7476,
       "step": 53
     },
     {
       "epoch": 0.98,
       "learning_rate": 5.1e-05,
-      "loss": 4.8158,
+      "loss": 4.813,
       "step": 54
     },
     {
       "epoch": 1.0,
       "learning_rate": 5.2e-05,
-      "loss": 5.4187,
+      "loss": 5.4126,
       "step": 55
     },
     {
       "epoch": 1.02,
       "learning_rate": 5.3e-05,
-      "loss": 4.7591,
+      "loss": 4.7531,
       "step": 56
     },
     {
       "epoch": 1.04,
       "learning_rate": 5.4e-05,
-      "loss": 4.7478,
+      "loss": 4.7446,
       "step": 57
     },
     {
       "epoch": 1.05,
       "learning_rate": 5.5e-05,
-      "loss": 4.5594,
+      "loss": 4.5525,
       "step": 58
     },
     {
       "epoch": 1.07,
       "learning_rate": 5.6e-05,
-      "loss": 4.3822,
+      "loss": 4.3812,
       "step": 59
     },
     {
       "epoch": 1.09,
       "learning_rate": 5.7e-05,
-      "loss": 4.3531,
+      "loss": 4.3479,
       "step": 60
     },
     {
       "epoch": 1.11,
       "learning_rate": 5.800000000000001e-05,
-      "loss": 4.6627,
+      "loss": 4.6587,
       "step": 61
     },
     {
       "epoch": 1.13,
       "learning_rate": 5.9e-05,
-      "loss": 4.6832,
+      "loss": 4.6814,
       "step": 62
     },
     {
       "epoch": 1.15,
       "learning_rate": 6e-05,
-      "loss": 4.4216,
+      "loss": 4.4158,
       "step": 63
     },
     {
       "epoch": 1.16,
       "learning_rate": 6.1e-05,
-      "loss": 4.322,
+      "loss": 4.3189,
       "step": 64
     },
     {
       "epoch": 1.18,
       "learning_rate": 6.2e-05,
-      "loss": 4.2335,
+      "loss": 4.2326,
       "step": 65
     },
     {
       "epoch": 1.2,
       "learning_rate": 6.3e-05,
-      "loss": 4.1466,
+      "loss": 4.1423,
       "step": 66
     },
     {
       "epoch": 1.22,
       "learning_rate": 6.4e-05,
-      "loss": 4.0849,
+      "loss": 4.0834,
       "step": 67
     },
     {
       "epoch": 1.24,
       "learning_rate": 6.500000000000001e-05,
-      "loss": 4.2652,
+      "loss": 4.2645,
       "step": 68
     },
     {
       "epoch": 1.25,
       "learning_rate": 6.6e-05,
-      "loss": 4.1821,
+      "loss": 4.1797,
       "step": 69
     },
     {
       "epoch": 1.27,
       "learning_rate": 6.7e-05,
-      "loss": 4.0778,
+      "loss": 4.0752,
       "step": 70
     },
     {
       "epoch": 1.29,
       "learning_rate": 6.800000000000001e-05,
-      "loss": 4.0222,
+      "loss": 4.0217,
       "step": 71
     },
     {
       "epoch": 1.31,
       "learning_rate": 6.900000000000001e-05,
-      "loss": 3.8911,
+      "loss": 3.8892,
       "step": 72
     },
     {
       "epoch": 1.33,
       "learning_rate": 7.000000000000001e-05,
-      "loss": 3.8835,
+      "loss": 3.8817,
       "step": 73
     },
     {
       "epoch": 1.35,
       "learning_rate": 7.099999999999999e-05,
-      "loss": 4.0057,
+      "loss": 4.0051,
       "step": 74
     },
     {
       "epoch": 1.36,
       "learning_rate": 7.2e-05,
-      "loss": 4.012,
+      "loss": 4.0109,
       "step": 75
     },
     {
       "epoch": 1.38,
       "learning_rate": 7.3e-05,
-      "loss": 3.8984,
+      "loss": 3.8969,
       "step": 76
     },
     {
       "epoch": 1.4,
       "learning_rate": 7.4e-05,
-      "loss": 3.8535,
+      "loss": 3.8526,
       "step": 77
     },
     {
       "epoch": 1.42,
       "learning_rate": 7.5e-05,
-      "loss": 3.8227,
+      "loss": 3.8221,
       "step": 78
     },
     {
       "epoch": 1.44,
       "learning_rate": 7.6e-05,
-      "loss": 3.8037,
+      "loss": 3.8026,
       "step": 79
     },
     {
       "epoch": 1.45,
       "learning_rate": 7.7e-05,
-      "loss": 3.916,
+      "loss": 3.9156,
       "step": 80
     },
     {
       "epoch": 1.47,
       "learning_rate": 7.8e-05,
-      "loss": 3.8382,
+      "loss": 3.8375,
       "step": 81
     },
     {
       "epoch": 1.49,
       "learning_rate": 7.9e-05,
-      "loss": 3.7337,
+      "loss": 3.7324,
       "step": 82
     },
     {
       "epoch": 1.51,
       "learning_rate": 8e-05,
-      "loss": 3.7725,
+      "loss": 3.7715,
       "step": 83
     },
     {
       "epoch": 1.53,
       "learning_rate": 8.1e-05,
-      "loss": 3.7002,
+      "loss": 3.6997,
       "step": 84
     },
     {
       "epoch": 1.55,
       "learning_rate": 8.2e-05,
-      "loss": 3.6101,
+      "loss": 3.6093,
       "step": 85
     },
     {
       "epoch": 1.56,
       "learning_rate": 8.300000000000001e-05,
-      "loss": 3.6486,
+      "loss": 3.6481,
       "step": 86
     },
     {
       "epoch": 1.58,
       "learning_rate": 8.400000000000001e-05,
-      "loss": 3.7121,
+      "loss": 3.7118,
       "step": 87
     },
     {
       "epoch": 1.6,
       "learning_rate": 8.5e-05,
-      "loss": 3.6171,
+      "loss": 3.6166,
       "step": 88
     },
     {
       "epoch": 1.62,
       "learning_rate": 8.599999999999999e-05,
-      "loss": 3.5804,
+      "loss": 3.5799,
       "step": 89
     },
     {
       "epoch": 1.64,
       "learning_rate": 8.7e-05,
-      "loss": 3.5441,
+      "loss": 3.5437,
       "step": 90
     },
     {
       "epoch": 1.65,
       "learning_rate": 8.8e-05,
-      "loss": 3.5081,
+      "loss": 3.5078,
       "step": 91
     },
     {
       "epoch": 1.67,
       "learning_rate": 8.9e-05,
-      "loss": 3.5926,
+      "loss": 3.5923,
       "step": 92
     },
     {
       "epoch": 1.69,
       "learning_rate": 8.999999999999999e-05,
-      "loss": 3.7213,
+      "loss": 3.7209,
       "step": 93
     },
     {
       "epoch": 1.71,
       "learning_rate": 9.1e-05,
-      "loss": 3.4969,
+      "loss": 3.4968,
       "step": 94
     },
     {
       "epoch": 1.73,
       "learning_rate": 9.2e-05,
-      "loss": 3.4836,
+      "loss": 3.4833,
       "step": 95
     },
     {
       "epoch": 1.75,
       "learning_rate": 9.3e-05,
-      "loss": 3.4579,
+      "loss": 3.4577,
       "step": 96
     },
     {
@@ -592,13 +592,13 @@
     {
       "epoch": 1.78,
       "learning_rate": 9.5e-05,
-      "loss": 3.4132,
+      "loss": 3.4131,
       "step": 98
     },
     {
       "epoch": 1.8,
       "learning_rate": 9.6e-05,
-      "loss": 3.4926,
+      "loss": 3.4923,
       "step": 99
     },
     {
@@ -616,13 +616,13 @@
     {
       "epoch": 1.85,
       "learning_rate": 9.900000000000001e-05,
-      "loss": 3.3274,
+      "loss": 3.3271,
       "step": 102
     },
     {
       "epoch": 1.87,
       "learning_rate": 0.0001,
-      "loss": 3.3582,
+      "loss": 3.3581,
       "step": 103
     },
     {
@@ -634,19 +634,19 @@
     {
       "epoch": 1.91,
       "learning_rate": 0.000102,
-      "loss": 3.373,
+      "loss": 3.3727,
       "step": 105
     },
     {
       "epoch": 1.93,
       "learning_rate": 0.000103,
-      "loss": 3.2887,
+      "loss": 3.2884,
       "step": 106
     },
     {
       "epoch": 1.95,
       "learning_rate": 0.000104,
-      "loss": 3.2888,
+      "loss": 3.2889,
       "step": 107
     },
     {
@@ -658,25 +658,25 @@
     {
       "epoch": 1.98,
       "learning_rate": 0.000106,
-      "loss": 3.3123,
+      "loss": 3.3119,
       "step": 109
     },
     {
       "epoch": 2.0,
       "learning_rate": 0.000107,
-      "loss": 3.3547,
+      "loss": 3.3544,
       "step": 110
     },
     {
       "epoch": 2.02,
       "learning_rate": 0.000108,
-      "loss": 3.2866,
+      "loss": 3.2876,
       "step": 111
     },
     {
       "epoch": 2.04,
       "learning_rate": 0.000109,
-      "loss": 3.2461,
+      "loss": 3.246,
       "step": 112
     },
     {
@@ -688,31 +688,31 @@
     {
       "epoch": 2.07,
       "learning_rate": 0.000111,
-      "loss": 3.2288,
+      "loss": 3.229,
       "step": 114
     },
     {
       "epoch": 2.09,
       "learning_rate": 0.000112,
-      "loss": 3.2111,
+      "loss": 3.2113,
       "step": 115
     },
     {
       "epoch": 2.11,
       "learning_rate": 0.00011300000000000001,
-      "loss": 3.2571,
+      "loss": 3.2576,
       "step": 116
     },
     {
       "epoch": 2.13,
       "learning_rate": 0.000114,
-      "loss": 3.2541,
+      "loss": 3.2543,
       "step": 117
     },
     {
       "epoch": 2.15,
       "learning_rate": 0.000115,
-      "loss": 3.2047,
+      "loss": 3.2046,
       "step": 118
     },
     {
@@ -730,15846 +730,24126 @@
     {
       "epoch": 2.2,
       "learning_rate": 0.000118,
-      "loss": 3.1901,
+      "loss": 3.1903,
       "step": 121
     },
     {
       "epoch": 2.22,
       "learning_rate": 0.00011899999999999999,
-      "loss": 3.2236,
+      "loss": 3.2241,
       "step": 122
     },
     {
       "epoch": 2.24,
       "learning_rate": 0.00012,
-      "loss": 3.3453,
+      "loss": 3.3425,
       "step": 123
     },
     {
       "epoch": 2.25,
       "learning_rate": 0.000121,
-      "loss": 3.1857,
+      "loss": 3.1854,
       "step": 124
     },
     {
       "epoch": 2.27,
       "learning_rate": 0.000122,
-      "loss": 3.1782,
+      "loss": 3.1776,
       "step": 125
     },
     {
       "epoch": 2.29,
       "learning_rate": 0.000123,
-      "loss": 3.1883,
+      "loss": 3.1878,
       "step": 126
     },
     {
       "epoch": 2.31,
       "learning_rate": 0.000124,
-      "loss": 3.1754,
+      "loss": 3.1746,
       "step": 127
     },
     {
       "epoch": 2.33,
       "learning_rate": 0.000125,
-      "loss": 3.1663,
+      "loss": 3.1652,
       "step": 128
     },
     {
       "epoch": 2.35,
       "learning_rate": 0.000126,
-      "loss": 3.2371,
+      "loss": 3.2359,
       "step": 129
     },
     {
       "epoch": 2.36,
       "learning_rate": 0.000127,
-      "loss": 3.1454,
+      "loss": 3.1451,
       "step": 130
     },
     {
       "epoch": 2.38,
       "learning_rate": 0.000128,
-      "loss": 3.1598,
+      "loss": 3.1579,
       "step": 131
     },
     {
       "epoch": 2.4,
       "learning_rate": 0.00012900000000000002,
-      "loss": 3.1433,
+      "loss": 3.1437,
       "step": 132
     },
     {
       "epoch": 2.42,
       "learning_rate": 0.00013000000000000002,
-      "loss": 3.1531,
+      "loss": 3.1516,
       "step": 133
     },
     {
       "epoch": 2.44,
       "learning_rate": 0.000131,
-      "loss": 3.1467,
+      "loss": 3.1461,
       "step": 134
     },
     {
       "epoch": 2.45,
       "learning_rate": 0.000132,
-      "loss": 3.1744,
+      "loss": 3.173,
       "step": 135
     },
     {
       "epoch": 2.47,
       "learning_rate": 0.000133,
-      "loss": 3.1956,
+      "loss": 3.2164,
       "step": 136
     },
     {
       "epoch": 2.49,
       "learning_rate": 0.000134,
-      "loss": 3.1084,
+      "loss": 3.1152,
       "step": 137
     },
     {
       "epoch": 2.51,
       "learning_rate": 0.000135,
-      "loss": 3.1265,
+      "loss": 3.134,
       "step": 138
     },
     {
       "epoch": 2.53,
       "learning_rate": 0.00013600000000000003,
-      "loss": 3.1289,
+      "loss": 3.1382,
       "step": 139
     },
     {
       "epoch": 2.55,
       "learning_rate": 0.00013700000000000002,
-      "loss": 3.1217,
+      "loss": 3.1325,
       "step": 140
     },
     {
       "epoch": 2.56,
       "learning_rate": 0.00013800000000000002,
-      "loss": 3.1812,
+      "loss": 3.1821,
       "step": 141
     },
     {
       "epoch": 2.58,
       "learning_rate": 0.00013900000000000002,
-      "loss": 3.1556,
+      "loss": 3.1651,
       "step": 142
     },
     {
       "epoch": 2.6,
       "learning_rate": 0.00014000000000000001,
-      "loss": 3.11,
+      "loss": 3.1145,
       "step": 143
     },
     {
       "epoch": 2.62,
       "learning_rate": 0.00014099999999999998,
-      "loss": 3.113,
+      "loss": 3.1212,
       "step": 144
     },
     {
       "epoch": 2.64,
       "learning_rate": 0.00014199999999999998,
-      "loss": 3.1041,
+      "loss": 3.1171,
       "step": 145
     },
     {
       "epoch": 2.65,
       "learning_rate": 0.00014299999999999998,
-      "loss": 3.142,
+      "loss": 3.1338,
       "step": 146
     },
     {
       "epoch": 2.67,
       "learning_rate": 0.000144,
-      "loss": 3.1028,
+      "loss": 3.1331,
       "step": 147
     },
     {
       "epoch": 2.69,
       "learning_rate": 0.000145,
-      "loss": 3.162,
+      "loss": 3.1985,
       "step": 148
     },
     {
       "epoch": 2.71,
       "learning_rate": 0.000146,
-      "loss": 3.0899,
+      "loss": 3.1058,
       "step": 149
     },
     {
       "epoch": 2.73,
       "learning_rate": 0.000147,
-      "loss": 3.0961,
+      "loss": 3.1062,
       "step": 150
     },
     {
       "epoch": 2.75,
       "learning_rate": 0.000148,
-      "loss": 3.08,
+      "loss": 3.0872,
       "step": 151
     },
     {
       "epoch": 2.76,
       "learning_rate": 0.000149,
-      "loss": 3.0853,
+      "loss": 3.1206,
       "step": 152
     },
     {
       "epoch": 2.78,
       "learning_rate": 0.00015,
-      "loss": 3.0903,
+      "loss": 3.1206,
       "step": 153
     },
     {
       "epoch": 2.8,
       "learning_rate": 0.000151,
-      "loss": 3.0899,
+      "loss": 3.1173,
       "step": 154
     },
     {
       "epoch": 2.82,
       "learning_rate": 0.000152,
-      "loss": 3.0976,
+      "loss": 3.1028,
       "step": 155
     },
     {
       "epoch": 2.84,
       "learning_rate": 0.000153,
-      "loss": 3.0625,
+      "loss": 3.0681,
       "step": 156
     },
     {
       "epoch": 2.85,
       "learning_rate": 0.000154,
-      "loss": 3.0899,
+      "loss": 3.0681,
       "step": 157
     },
     {
       "epoch": 2.87,
       "learning_rate": 0.000155,
-      "loss": 3.0633,
+      "loss": 3.0745,
       "step": 158
     },
     {
       "epoch": 2.89,
       "learning_rate": 0.000156,
-      "loss": 3.0872,
+      "loss": 3.0973,
       "step": 159
     },
     {
       "epoch": 2.91,
       "learning_rate": 0.000157,
-      "loss": 3.112,
+      "loss": 3.126,
       "step": 160
     },
     {
       "epoch": 2.93,
       "learning_rate": 0.000158,
-      "loss": 3.0724,
+      "loss": 3.0665,
       "step": 161
     },
     {
       "epoch": 2.95,
       "learning_rate": 0.00015900000000000002,
-      "loss": 3.0704,
+      "loss": 3.0838,
       "step": 162
     },
     {
       "epoch": 2.96,
       "learning_rate": 0.00016,
-      "loss": 3.0534,
+      "loss": 3.0537,
       "step": 163
     },
     {
       "epoch": 2.98,
       "learning_rate": 0.000161,
-      "loss": 3.0429,
+      "loss": 3.0644,
       "step": 164
     },
     {
       "epoch": 3.0,
       "learning_rate": 0.000162,
-      "loss": 3.0736,
+      "loss": 3.0884,
       "step": 165
     },
     {
       "epoch": 3.02,
       "learning_rate": 0.000163,
-      "loss": 3.0946,
+      "loss": 3.0656,
       "step": 166
     },
     {
       "epoch": 3.04,
       "learning_rate": 0.000164,
-      "loss": 3.0545,
+      "loss": 3.0522,
       "step": 167
     },
     {
       "epoch": 3.05,
       "learning_rate": 0.000165,
-      "loss": 3.0656,
+      "loss": 3.0562,
       "step": 168
     },
     {
       "epoch": 3.07,
       "learning_rate": 0.00016600000000000002,
-      "loss": 3.0758,
+      "loss": 3.0571,
       "step": 169
     },
     {
       "epoch": 3.09,
       "learning_rate": 0.00016700000000000002,
-      "loss": 3.0627,
+      "loss": 3.0812,
       "step": 170
     },
     {
       "epoch": 3.11,
       "learning_rate": 0.00016800000000000002,
-      "loss": 3.1162,
+      "loss": 3.1032,
       "step": 171
     },
     {
       "epoch": 3.13,
       "learning_rate": 0.00016900000000000002,
-      "loss": 3.0719,
+      "loss": 3.0541,
       "step": 172
     },
     {
       "epoch": 3.15,
       "learning_rate": 0.00017,
-      "loss": 3.0263,
+      "loss": 3.035,
       "step": 173
     },
     {
       "epoch": 3.16,
       "learning_rate": 0.000171,
-      "loss": 3.054,
+      "loss": 3.0512,
       "step": 174
     },
     {
       "epoch": 3.18,
       "learning_rate": 0.00017199999999999998,
-      "loss": 3.058,
+      "loss": 3.0539,
       "step": 175
     },
     {
       "epoch": 3.2,
       "learning_rate": 0.000173,
-      "loss": 3.0366,
+      "loss": 3.0328,
       "step": 176
     },
     {
       "epoch": 3.22,
       "learning_rate": 0.000174,
-      "loss": 3.0335,
+      "loss": 3.0367,
       "step": 177
     },
     {
       "epoch": 3.24,
       "learning_rate": 0.000175,
-      "loss": 3.078,
+      "loss": 3.0608,
       "step": 178
     },
     {
       "epoch": 3.25,
       "learning_rate": 0.000176,
-      "loss": 3.0564,
+      "loss": 3.0598,
       "step": 179
     },
     {
       "epoch": 3.27,
       "learning_rate": 0.000177,
-      "loss": 3.0048,
+      "loss": 3.0025,
       "step": 180
     },
     {
       "epoch": 3.29,
       "learning_rate": 0.000178,
-      "loss": 3.0359,
+      "loss": 3.0337,
       "step": 181
     },
     {
       "epoch": 3.31,
       "learning_rate": 0.000179,
-      "loss": 3.0737,
+      "loss": 3.0694,
       "step": 182
     },
     {
       "epoch": 3.33,
       "learning_rate": 0.00017999999999999998,
-      "loss": 3.0093,
+      "loss": 3.0073,
       "step": 183
     },
     {
       "epoch": 3.35,
       "learning_rate": 0.000181,
-      "loss": 3.0337,
+      "loss": 3.0346,
       "step": 184
     },
     {
       "epoch": 3.36,
       "learning_rate": 0.000182,
-      "loss": 3.0376,
+      "loss": 3.0386,
       "step": 185
     },
     {
       "epoch": 3.38,
       "learning_rate": 0.000183,
-      "loss": 3.0566,
+      "loss": 3.0529,
       "step": 186
     },
     {
       "epoch": 3.4,
       "learning_rate": 0.000184,
-      "loss": 3.0338,
+      "loss": 3.0314,
       "step": 187
     },
     {
       "epoch": 3.42,
       "learning_rate": 0.000185,
-      "loss": 3.0305,
+      "loss": 3.031,
       "step": 188
     },
     {
       "epoch": 3.44,
       "learning_rate": 0.000186,
-      "loss": 3.0281,
+      "loss": 3.0292,
       "step": 189
     },
     {
       "epoch": 3.45,
       "learning_rate": 0.000187,
-      "loss": 3.0338,
+      "loss": 3.0277,
       "step": 190
     },
     {
       "epoch": 3.47,
       "learning_rate": 0.00018800000000000002,
-      "loss": 3.0555,
+      "loss": 3.0583,
       "step": 191
     },
     {
       "epoch": 3.49,
       "learning_rate": 0.000189,
-      "loss": 3.0434,
+      "loss": 3.0499,
       "step": 192
     },
     {
       "epoch": 3.51,
       "learning_rate": 0.00019,
-      "loss": 3.0229,
+      "loss": 3.0235,
       "step": 193
     },
     {
       "epoch": 3.53,
       "learning_rate": 0.000191,
-      "loss": 3.0268,
+      "loss": 3.037,
       "step": 194
     },
     {
       "epoch": 3.55,
       "learning_rate": 0.000192,
-      "loss": 3.0304,
+      "loss": 3.0311,
       "step": 195
     },
     {
       "epoch": 3.56,
       "learning_rate": 0.000193,
-      "loss": 3.0047,
+      "loss": 3.014,
       "step": 196
     },
     {
       "epoch": 3.58,
       "learning_rate": 0.000194,
-      "loss": 3.0204,
+      "loss": 3.0291,
       "step": 197
     },
     {
       "epoch": 3.6,
       "learning_rate": 0.00019500000000000002,
-      "loss": 3.0288,
+      "loss": 3.0309,
       "step": 198
     },
     {
       "epoch": 3.62,
       "learning_rate": 0.00019600000000000002,
-      "loss": 2.9999,
+      "loss": 3.0076,
       "step": 199
     },
     {
       "epoch": 3.64,
       "learning_rate": 0.00019700000000000002,
-      "loss": 3.0234,
+      "loss": 3.0286,
       "step": 200
     },
     {
       "epoch": 3.65,
       "learning_rate": 0.00019800000000000002,
-      "loss": 3.0044,
+      "loss": 3.0056,
       "step": 201
     },
     {
       "epoch": 3.67,
       "learning_rate": 0.000199,
-      "loss": 2.9642,
+      "loss": 2.9736,
       "step": 202
     },
     {
       "epoch": 3.69,
       "learning_rate": 0.0002,
-      "loss": 3.0243,
+      "loss": 3.0267,
       "step": 203
     },
     {
       "epoch": 3.71,
       "learning_rate": 0.000201,
-      "loss": 3.0057,
+      "loss": 3.0209,
       "step": 204
     },
     {
       "epoch": 3.73,
       "learning_rate": 0.000202,
-      "loss": 2.9896,
+      "loss": 2.9988,
       "step": 205
     },
     {
       "epoch": 3.75,
       "learning_rate": 0.00020300000000000003,
-      "loss": 3.017,
+      "loss": 2.9928,
       "step": 206
     },
     {
       "epoch": 3.76,
       "learning_rate": 0.000204,
-      "loss": 2.9893,
+      "loss": 2.988,
       "step": 207
     },
     {
       "epoch": 3.78,
       "learning_rate": 0.000205,
-      "loss": 2.9906,
+      "loss": 2.9929,
       "step": 208
     },
     {
       "epoch": 3.8,
       "learning_rate": 0.000206,
-      "loss": 3.0004,
+      "loss": 2.9748,
       "step": 209
     },
     {
       "epoch": 3.82,
       "learning_rate": 0.000207,
-      "loss": 2.9901,
+      "loss": 2.9805,
       "step": 210
     },
     {
       "epoch": 3.84,
       "learning_rate": 0.000208,
-      "loss": 3.0208,
+      "loss": 2.9961,
       "step": 211
     },
     {
       "epoch": 3.85,
       "learning_rate": 0.00020899999999999998,
-      "loss": 2.9932,
+      "loss": 2.9694,
       "step": 212
     },
     {
       "epoch": 3.87,
       "learning_rate": 0.00021,
-      "loss": 2.9674,
+      "loss": 2.9456,
       "step": 213
     },
     {
       "epoch": 3.89,
       "learning_rate": 0.000211,
-      "loss": 2.9687,
+      "loss": 2.9481,
       "step": 214
     },
     {
       "epoch": 3.91,
       "learning_rate": 0.000212,
-      "loss": 3.0161,
+      "loss": 3.01,
       "step": 215
     },
     {
       "epoch": 3.93,
       "learning_rate": 0.000213,
-      "loss": 2.9991,
+      "loss": 2.98,
       "step": 216
     },
     {
       "epoch": 3.95,
       "learning_rate": 0.000214,
-      "loss": 2.9617,
+      "loss": 2.9409,
       "step": 217
     },
     {
       "epoch": 3.96,
       "learning_rate": 0.000215,
-      "loss": 2.9637,
+      "loss": 2.9379,
       "step": 218
     },
     {
       "epoch": 3.98,
       "learning_rate": 0.000216,
-      "loss": 2.9238,
+      "loss": 2.9003,
       "step": 219
     },
     {
       "epoch": 4.0,
       "learning_rate": 0.00021700000000000002,
-      "loss": 2.9462,
+      "loss": 2.9401,
       "step": 220
     },
     {
       "epoch": 4.02,
       "learning_rate": 0.000218,
-      "loss": 2.9644,
+      "loss": 2.9351,
       "step": 221
     },
     {
       "epoch": 4.04,
       "learning_rate": 0.000219,
-      "loss": 2.9597,
+      "loss": 2.952,
       "step": 222
     },
     {
       "epoch": 4.05,
       "learning_rate": 0.00022,
-      "loss": 2.926,
+      "loss": 2.9292,
       "step": 223
     },
     {
       "epoch": 4.07,
       "learning_rate": 0.000221,
-      "loss": 2.9654,
+      "loss": 2.939,
       "step": 224
     },
     {
       "epoch": 4.09,
       "learning_rate": 0.000222,
-      "loss": 2.9364,
+      "loss": 2.9059,
       "step": 225
     },
     {
       "epoch": 4.11,
       "learning_rate": 0.000223,
-      "loss": 2.9098,
+      "loss": 2.8869,
       "step": 226
     },
     {
       "epoch": 4.13,
       "learning_rate": 0.000224,
-      "loss": 2.9645,
+      "loss": 2.9294,
       "step": 227
     },
     {
       "epoch": 4.15,
       "learning_rate": 0.00022500000000000002,
-      "loss": 2.9431,
+      "loss": 2.9215,
       "step": 228
     },
     {
       "epoch": 4.16,
       "learning_rate": 0.00022600000000000002,
-      "loss": 2.911,
+      "loss": 2.8795,
       "step": 229
     },
     {
       "epoch": 4.18,
       "learning_rate": 0.00022700000000000002,
-      "loss": 2.9217,
+      "loss": 2.8728,
       "step": 230
     },
     {
       "epoch": 4.2,
       "learning_rate": 0.000228,
-      "loss": 2.8868,
+      "loss": 2.8408,
       "step": 231
     },
     {
       "epoch": 4.22,
       "learning_rate": 0.000229,
-      "loss": 2.8742,
+      "loss": 2.8284,
       "step": 232
     },
     {
       "epoch": 4.24,
       "learning_rate": 0.00023,
-      "loss": 2.8968,
+      "loss": 2.8592,
       "step": 233
     },
     {
       "epoch": 4.25,
       "learning_rate": 0.000231,
-      "loss": 2.9068,
+      "loss": 2.8417,
       "step": 234
     },
     {
       "epoch": 4.27,
       "learning_rate": 0.00023200000000000003,
-      "loss": 2.8477,
+      "loss": 2.7846,
       "step": 235
     },
     {
       "epoch": 4.29,
       "learning_rate": 0.00023300000000000003,
-      "loss": 2.8895,
+      "loss": 2.844,
       "step": 236
     },
     {
       "epoch": 4.31,
       "learning_rate": 0.00023400000000000002,
-      "loss": 2.8636,
+      "loss": 2.7934,
       "step": 237
     },
     {
       "epoch": 4.33,
       "learning_rate": 0.000235,
-      "loss": 2.8683,
+      "loss": 2.7844,
       "step": 238
     },
     {
       "epoch": 4.35,
       "learning_rate": 0.000236,
-      "loss": 2.8383,
+      "loss": 2.7554,
       "step": 239
     },
     {
       "epoch": 4.36,
       "learning_rate": 0.000237,
-      "loss": 2.8606,
+      "loss": 2.7467,
       "step": 240
     },
     {
       "epoch": 4.38,
       "learning_rate": 0.00023799999999999998,
-      "loss": 2.8377,
+      "loss": 2.7069,
       "step": 241
     },
     {
       "epoch": 4.4,
       "learning_rate": 0.00023899999999999998,
-      "loss": 2.8447,
+      "loss": 2.8861,
       "step": 242
     },
     {
       "epoch": 4.42,
       "learning_rate": 0.00024,
-      "loss": 2.7794,
+      "loss": 2.6861,
       "step": 243
     },
     {
       "epoch": 4.44,
       "learning_rate": 0.000241,
-      "loss": 2.7507,
+      "loss": 2.7521,
       "step": 244
     },
     {
       "epoch": 4.45,
       "learning_rate": 0.000242,
-      "loss": 2.7547,
+      "loss": 2.7237,
       "step": 245
     },
     {
       "epoch": 4.47,
       "learning_rate": 0.000243,
-      "loss": 2.7685,
+      "loss": 2.7602,
       "step": 246
     },
     {
       "epoch": 4.49,
       "learning_rate": 0.000244,
-      "loss": 2.771,
+      "loss": 2.7474,
       "step": 247
     },
     {
       "epoch": 4.51,
       "learning_rate": 0.000245,
-      "loss": 2.7773,
+      "loss": 2.6487,
       "step": 248
     },
     {
       "epoch": 4.53,
       "learning_rate": 0.000246,
-      "loss": 2.7137,
+      "loss": 2.6627,
       "step": 249
     },
     {
       "epoch": 4.55,
       "learning_rate": 0.000247,
-      "loss": 2.6896,
+      "loss": 2.6139,
       "step": 250
     },
     {
       "epoch": 4.56,
       "learning_rate": 0.000248,
-      "loss": 2.6704,
+      "loss": 2.5797,
       "step": 251
     },
     {
       "epoch": 4.58,
       "learning_rate": 0.000249,
-      "loss": 2.694,
+      "loss": 2.5825,
       "step": 252
     },
     {
       "epoch": 4.6,
       "learning_rate": 0.00025,
-      "loss": 2.6537,
+      "loss": 2.5326,
       "step": 253
     },
     {
       "epoch": 4.62,
       "learning_rate": 0.00025100000000000003,
-      "loss": 2.626,
+      "loss": 2.5127,
       "step": 254
     },
     {
       "epoch": 4.64,
       "learning_rate": 0.000252,
-      "loss": 2.6568,
+      "loss": 2.5616,
       "step": 255
     },
     {
       "epoch": 4.65,
       "learning_rate": 0.000253,
-      "loss": 2.6394,
+      "loss": 2.4932,
       "step": 256
     },
     {
       "epoch": 4.67,
       "learning_rate": 0.000254,
-      "loss": 2.5723,
+      "loss": 2.3507,
       "step": 257
     },
     {
       "epoch": 4.69,
       "learning_rate": 0.000255,
-      "loss": 2.5843,
+      "loss": 2.4566,
       "step": 258
     },
     {
       "epoch": 4.71,
       "learning_rate": 0.000256,
-      "loss": 2.5655,
+      "loss": 2.3755,
       "step": 259
     },
     {
       "epoch": 4.73,
       "learning_rate": 0.000257,
-      "loss": 2.4741,
+      "loss": 2.3236,
       "step": 260
     },
     {
       "epoch": 4.75,
       "learning_rate": 0.00025800000000000004,
-      "loss": 2.4483,
+      "loss": 2.2666,
       "step": 261
     },
     {
       "epoch": 4.76,
       "learning_rate": 0.000259,
-      "loss": 2.4522,
+      "loss": 2.2723,
       "step": 262
     },
     {
       "epoch": 4.78,
       "learning_rate": 0.00026000000000000003,
-      "loss": 2.432,
+      "loss": 2.2324,
       "step": 263
     },
     {
       "epoch": 4.8,
       "learning_rate": 0.000261,
-      "loss": 2.397,
+      "loss": 2.2332,
       "step": 264
     },
     {
       "epoch": 4.82,
       "learning_rate": 0.000262,
-      "loss": 2.3505,
+      "loss": 2.1582,
       "step": 265
     },
     {
       "epoch": 4.84,
       "learning_rate": 0.000263,
-      "loss": 2.3409,
+      "loss": 2.1142,
       "step": 266
     },
     {
       "epoch": 4.85,
       "learning_rate": 0.000264,
-      "loss": 2.2642,
+      "loss": 2.0329,
       "step": 267
     },
     {
       "epoch": 4.87,
       "learning_rate": 0.00026500000000000004,
-      "loss": 2.275,
+      "loss": 2.0573,
       "step": 268
     },
     {
       "epoch": 4.89,
       "learning_rate": 0.000266,
-      "loss": 2.2057,
+      "loss": 2.038,
       "step": 269
     },
     {
       "epoch": 4.91,
       "learning_rate": 0.00026700000000000004,
-      "loss": 2.2152,
+      "loss": 2.0085,
       "step": 270
     },
     {
       "epoch": 4.93,
       "learning_rate": 0.000268,
-      "loss": 2.126,
+      "loss": 1.9054,
       "step": 271
     },
     {
       "epoch": 4.95,
       "learning_rate": 0.00026900000000000003,
-      "loss": 2.1146,
+      "loss": 1.8839,
       "step": 272
     },
     {
       "epoch": 4.96,
       "learning_rate": 0.00027,
-      "loss": 2.0546,
+      "loss": 1.8373,
       "step": 273
     },
     {
       "epoch": 4.98,
       "learning_rate": 0.00027100000000000003,
-      "loss": 2.0544,
+      "loss": 1.7761,
       "step": 274
     },
     {
       "epoch": 5.0,
       "learning_rate": 0.00027200000000000005,
-      "loss": 1.9963,
+      "loss": 1.86,
       "step": 275
     },
     {
       "epoch": 5.02,
       "learning_rate": 0.000273,
-      "loss": 1.9721,
+      "loss": 1.7408,
       "step": 276
     },
     {
       "epoch": 5.04,
       "learning_rate": 0.00027400000000000005,
-      "loss": 1.927,
+      "loss": 1.7459,
       "step": 277
     },
     {
       "epoch": 5.05,
       "learning_rate": 0.000275,
-      "loss": 1.8303,
+      "loss": 1.6199,
       "step": 278
     },
     {
       "epoch": 5.07,
       "learning_rate": 0.00027600000000000004,
-      "loss": 1.8781,
+      "loss": 1.7135,
       "step": 279
     },
     {
       "epoch": 5.09,
       "learning_rate": 0.000277,
-      "loss": 1.8205,
+      "loss": 1.6402,
       "step": 280
     },
     {
       "epoch": 5.11,
       "learning_rate": 0.00027800000000000004,
-      "loss": 1.784,
+      "loss": 1.5784,
       "step": 281
     },
     {
       "epoch": 5.13,
       "learning_rate": 0.000279,
-      "loss": 1.7186,
+      "loss": 1.6468,
       "step": 282
     },
     {
       "epoch": 5.15,
       "learning_rate": 0.00028000000000000003,
-      "loss": 1.6852,
+      "loss": 1.5326,
       "step": 283
     },
     {
       "epoch": 5.16,
       "learning_rate": 0.00028100000000000005,
-      "loss": 1.6617,
+      "loss": 1.5378,
       "step": 284
     },
     {
       "epoch": 5.18,
       "learning_rate": 0.00028199999999999997,
-      "loss": 1.7069,
+      "loss": 1.5159,
       "step": 285
     },
     {
       "epoch": 5.2,
       "learning_rate": 0.000283,
-      "loss": 1.6373,
+      "loss": 1.5036,
       "step": 286
     },
     {
       "epoch": 5.22,
       "learning_rate": 0.00028399999999999996,
-      "loss": 1.6009,
+      "loss": 1.5014,
       "step": 287
     },
     {
       "epoch": 5.24,
       "learning_rate": 0.000285,
-      "loss": 1.7024,
+      "loss": 1.6438,
       "step": 288
     },
     {
       "epoch": 5.25,
       "learning_rate": 0.00028599999999999996,
-      "loss": 1.5761,
+      "loss": 1.4459,
       "step": 289
     },
     {
       "epoch": 5.27,
       "learning_rate": 0.000287,
-      "loss": 1.5909,
+      "loss": 1.5018,
       "step": 290
     },
     {
       "epoch": 5.29,
       "learning_rate": 0.000288,
-      "loss": 1.5538,
+      "loss": 1.4682,
       "step": 291
     },
     {
       "epoch": 5.31,
       "learning_rate": 0.000289,
-      "loss": 1.5244,
+      "loss": 1.4767,
       "step": 292
     },
     {
       "epoch": 5.33,
       "learning_rate": 0.00029,
-      "loss": 1.619,
+      "loss": 1.5045,
       "step": 293
     },
     {
       "epoch": 5.35,
       "learning_rate": 0.00029099999999999997,
-      "loss": 1.6194,
+      "loss": 1.5072,
       "step": 294
     },
     {
       "epoch": 5.36,
       "learning_rate": 0.000292,
-      "loss": 1.5265,
+      "loss": 1.4073,
       "step": 295
     },
     {
       "epoch": 5.38,
       "learning_rate": 0.00029299999999999997,
-      "loss": 1.4398,
+      "loss": 1.3549,
       "step": 296
     },
     {
       "epoch": 5.4,
       "learning_rate": 0.000294,
-      "loss": 1.4835,
+      "loss": 1.3665,
       "step": 297
     },
     {
       "epoch": 5.42,
       "learning_rate": 0.000295,
-      "loss": 1.3887,
+      "loss": 1.319,
       "step": 298
     },
     {
       "epoch": 5.44,
       "learning_rate": 0.000296,
-      "loss": 1.4731,
+      "loss": 1.3656,
       "step": 299
     },
     {
       "epoch": 5.45,
       "learning_rate": 0.000297,
-      "loss": 1.4778,
+      "loss": 1.3657,
       "step": 300
     },
     {
       "epoch": 5.47,
       "learning_rate": 0.000298,
-      "loss": 1.4505,
+      "loss": 1.3522,
       "step": 301
     },
     {
       "epoch": 5.49,
       "learning_rate": 0.000299,
-      "loss": 1.3636,
+      "loss": 1.2801,
       "step": 302
     },
     {
       "epoch": 5.51,
       "learning_rate": 0.0003,
-      "loss": 1.375,
+      "loss": 1.2717,
       "step": 303
     },
     {
       "epoch": 5.53,
       "learning_rate": 0.000301,
-      "loss": 1.396,
+      "loss": 1.3205,
       "step": 304
     },
     {
       "epoch": 5.55,
       "learning_rate": 0.000302,
-      "loss": 1.3229,
+      "loss": 1.2476,
       "step": 305
     },
     {
       "epoch": 5.56,
       "learning_rate": 0.000303,
-      "loss": 1.4473,
+      "loss": 1.3382,
       "step": 306
     },
     {
       "epoch": 5.58,
       "learning_rate": 0.000304,
-      "loss": 1.4354,
+      "loss": 1.3966,
       "step": 307
     },
     {
       "epoch": 5.6,
       "learning_rate": 0.000305,
-      "loss": 1.3831,
+      "loss": 1.3101,
       "step": 308
     },
     {
       "epoch": 5.62,
       "learning_rate": 0.000306,
-      "loss": 1.3579,
+      "loss": 1.2761,
       "step": 309
     },
     {
       "epoch": 5.64,
       "learning_rate": 0.000307,
-      "loss": 1.3633,
+      "loss": 1.3056,
       "step": 310
     },
     {
       "epoch": 5.65,
       "learning_rate": 0.000308,
-      "loss": 1.4267,
+      "loss": 1.3277,
       "step": 311
     },
     {
       "epoch": 5.67,
       "learning_rate": 0.00030900000000000003,
-      "loss": 1.2759,
+      "loss": 1.2432,
       "step": 312
     },
     {
       "epoch": 5.69,
       "learning_rate": 0.00031,
-      "loss": 1.3246,
+      "loss": 1.237,
       "step": 313
     },
     {
       "epoch": 5.71,
       "learning_rate": 0.000311,
-      "loss": 1.3183,
+      "loss": 1.2074,
       "step": 314
     },
     {
       "epoch": 5.73,
       "learning_rate": 0.000312,
-      "loss": 1.2379,
+      "loss": 1.17,
       "step": 315
     },
     {
       "epoch": 5.75,
       "learning_rate": 0.000313,
-      "loss": 1.2722,
+      "loss": 1.1822,
       "step": 316
     },
     {
       "epoch": 5.76,
       "learning_rate": 0.000314,
-      "loss": 1.2942,
+      "loss": 1.2003,
       "step": 317
     },
     {
       "epoch": 5.78,
       "learning_rate": 0.000315,
-      "loss": 1.2907,
+      "loss": 1.2198,
       "step": 318
     },
     {
       "epoch": 5.8,
       "learning_rate": 0.000316,
-      "loss": 1.2964,
+      "loss": 1.2261,
       "step": 319
     },
     {
       "epoch": 5.82,
       "learning_rate": 0.000317,
-      "loss": 1.2643,
+      "loss": 1.2267,
       "step": 320
     },
     {
       "epoch": 5.84,
       "learning_rate": 0.00031800000000000003,
-      "loss": 1.2212,
+      "loss": 1.18,
       "step": 321
     },
     {
       "epoch": 5.85,
       "learning_rate": 0.000319,
-      "loss": 1.2463,
+      "loss": 1.1968,
       "step": 322
     },
     {
       "epoch": 5.87,
       "learning_rate": 0.00032,
-      "loss": 1.2733,
+      "loss": 1.2618,
       "step": 323
     },
     {
       "epoch": 5.89,
       "learning_rate": 0.000321,
-      "loss": 1.259,
+      "loss": 1.2039,
       "step": 324
     },
     {
       "epoch": 5.91,
       "learning_rate": 0.000322,
-      "loss": 1.2705,
+      "loss": 1.2139,
       "step": 325
     },
     {
       "epoch": 5.93,
       "learning_rate": 0.000323,
-      "loss": 1.2436,
+      "loss": 1.1935,
       "step": 326
     },
     {
       "epoch": 5.95,
       "learning_rate": 0.000324,
-      "loss": 1.1917,
+      "loss": 1.1772,
       "step": 327
     },
     {
       "epoch": 5.96,
       "learning_rate": 0.00032500000000000004,
-      "loss": 1.2366,
+      "loss": 1.1747,
       "step": 328
     },
     {
       "epoch": 5.98,
       "learning_rate": 0.000326,
-      "loss": 1.2528,
+      "loss": 1.2621,
       "step": 329
     },
     {
       "epoch": 6.0,
       "learning_rate": 0.00032700000000000003,
-      "loss": 1.3331,
+      "loss": 1.1954,
       "step": 330
     },
     {
       "epoch": 6.02,
       "learning_rate": 0.000328,
-      "loss": 1.1346,
+      "loss": 1.1302,
       "step": 331
     },
     {
       "epoch": 6.04,
       "learning_rate": 0.00032900000000000003,
-      "loss": 1.141,
+      "loss": 1.0611,
       "step": 332
     },
     {
       "epoch": 6.05,
       "learning_rate": 0.00033,
-      "loss": 1.1234,
+      "loss": 1.1015,
       "step": 333
     },
     {
       "epoch": 6.07,
       "learning_rate": 0.000331,
-      "loss": 1.1523,
+      "loss": 1.1412,
       "step": 334
     },
     {
       "epoch": 6.09,
       "learning_rate": 0.00033200000000000005,
-      "loss": 1.1621,
+      "loss": 1.122,
       "step": 335
     },
     {
       "epoch": 6.11,
       "learning_rate": 0.000333,
-      "loss": 1.1154,
+      "loss": 1.0705,
       "step": 336
     },
     {
       "epoch": 6.13,
       "learning_rate": 0.00033400000000000004,
-      "loss": 1.1856,
+      "loss": 1.1745,
       "step": 337
     },
     {
       "epoch": 6.15,
       "learning_rate": 0.000335,
-      "loss": 1.1252,
+      "loss": 1.0772,
       "step": 338
     },
     {
       "epoch": 6.16,
       "learning_rate": 0.00033600000000000004,
-      "loss": 1.0908,
+      "loss": 1.0405,
       "step": 339
     },
     {
       "epoch": 6.18,
       "learning_rate": 0.000337,
-      "loss": 1.0554,
+      "loss": 1.0319,
       "step": 340
     },
     {
       "epoch": 6.2,
       "learning_rate": 0.00033800000000000003,
-      "loss": 1.1727,
+      "loss": 1.0935,
       "step": 341
     },
     {
       "epoch": 6.22,
       "learning_rate": 0.00033900000000000005,
-      "loss": 1.0502,
+      "loss": 1.061,
       "step": 342
     },
     {
       "epoch": 6.24,
       "learning_rate": 0.00034,
-      "loss": 1.0986,
+      "loss": 1.0589,
       "step": 343
     },
     {
       "epoch": 6.25,
       "learning_rate": 0.00034100000000000005,
-      "loss": 1.1523,
+      "loss": 1.1206,
       "step": 344
     },
     {
       "epoch": 6.27,
       "learning_rate": 0.000342,
-      "loss": 1.0862,
+      "loss": 1.0117,
       "step": 345
     },
     {
       "epoch": 6.29,
       "learning_rate": 0.00034300000000000004,
-      "loss": 1.0383,
+      "loss": 0.9684,
       "step": 346
     },
     {
       "epoch": 6.31,
       "learning_rate": 0.00034399999999999996,
-      "loss": 1.0746,
+      "loss": 1.0159,
       "step": 347
     },
     {
       "epoch": 6.33,
       "learning_rate": 0.000345,
-      "loss": 1.0967,
+      "loss": 1.0606,
       "step": 348
     },
     {
       "epoch": 6.35,
       "learning_rate": 0.000346,
-      "loss": 1.1217,
+      "loss": 1.0541,
       "step": 349
     },
     {
       "epoch": 6.36,
       "learning_rate": 0.000347,
-      "loss": 1.132,
+      "loss": 1.0617,
       "step": 350
     },
     {
       "epoch": 6.38,
       "learning_rate": 0.000348,
-      "loss": 0.9944,
+      "loss": 0.972,
       "step": 351
     },
     {
       "epoch": 6.4,
       "learning_rate": 0.00034899999999999997,
-      "loss": 1.0508,
+      "loss": 1.0331,
       "step": 352
     },
     {
       "epoch": 6.42,
       "learning_rate": 0.00035,
-      "loss": 1.0427,
+      "loss": 1.0425,
       "step": 353
     },
     {
       "epoch": 6.44,
       "learning_rate": 0.00035099999999999997,
-      "loss": 1.0892,
+      "loss": 0.9986,
       "step": 354
     },
     {
       "epoch": 6.45,
       "learning_rate": 0.000352,
-      "loss": 1.0795,
+      "loss": 1.0247,
       "step": 355
     },
     {
       "epoch": 6.47,
       "learning_rate": 0.00035299999999999996,
-      "loss": 1.0638,
+      "loss": 1.035,
       "step": 356
     },
     {
       "epoch": 6.49,
       "learning_rate": 0.000354,
-      "loss": 1.0504,
+      "loss": 0.9932,
       "step": 357
     },
     {
       "epoch": 6.51,
       "learning_rate": 0.000355,
-      "loss": 1.1216,
+      "loss": 1.0521,
       "step": 358
     },
     {
       "epoch": 6.53,
       "learning_rate": 0.000356,
-      "loss": 1.1328,
+      "loss": 1.0885,
       "step": 359
     },
     {
       "epoch": 6.55,
       "learning_rate": 0.000357,
-      "loss": 1.127,
+      "loss": 1.0847,
       "step": 360
     },
     {
       "epoch": 6.56,
       "learning_rate": 0.000358,
-      "loss": 1.0966,
+      "loss": 1.0337,
       "step": 361
     },
     {
       "epoch": 6.58,
       "learning_rate": 0.000359,
-      "loss": 1.0784,
+      "loss": 1.052,
       "step": 362
     },
     {
       "epoch": 6.6,
       "learning_rate": 0.00035999999999999997,
-      "loss": 1.0739,
+      "loss": 1.0218,
       "step": 363
     },
     {
       "epoch": 6.62,
       "learning_rate": 0.000361,
-      "loss": 1.0411,
+      "loss": 1.0221,
       "step": 364
     },
     {
       "epoch": 6.64,
       "learning_rate": 0.000362,
-      "loss": 1.1095,
+      "loss": 1.0908,
       "step": 365
     },
     {
       "epoch": 6.65,
       "learning_rate": 0.000363,
-      "loss": 1.0733,
+      "loss": 0.9903,
       "step": 366
     },
     {
       "epoch": 6.67,
       "learning_rate": 0.000364,
-      "loss": 1.0737,
+      "loss": 1.0619,
       "step": 367
     },
     {
       "epoch": 6.69,
       "learning_rate": 0.000365,
-      "loss": 1.0375,
+      "loss": 0.9264,
       "step": 368
     },
     {
       "epoch": 6.71,
       "learning_rate": 0.000366,
-      "loss": 1.0831,
+      "loss": 1.0361,
       "step": 369
     },
     {
       "epoch": 6.73,
       "learning_rate": 0.000367,
-      "loss": 1.0394,
+      "loss": 1.0227,
       "step": 370
     },
     {
       "epoch": 6.75,
       "learning_rate": 0.000368,
-      "loss": 0.9889,
+      "loss": 0.9315,
       "step": 371
     },
     {
       "epoch": 6.76,
       "learning_rate": 0.000369,
-      "loss": 1.0911,
+      "loss": 1.0542,
       "step": 372
     },
     {
       "epoch": 6.78,
       "learning_rate": 0.00037,
-      "loss": 1.0856,
+      "loss": 1.037,
       "step": 373
     },
     {
       "epoch": 6.8,
       "learning_rate": 0.000371,
-      "loss": 1.0748,
+      "loss": 1.037,
       "step": 374
     },
     {
       "epoch": 6.82,
       "learning_rate": 0.000372,
-      "loss": 1.0622,
+      "loss": 1.0376,
       "step": 375
     },
     {
       "epoch": 6.84,
       "learning_rate": 0.000373,
-      "loss": 0.9811,
+      "loss": 0.9484,
       "step": 376
     },
     {
       "epoch": 6.85,
       "learning_rate": 0.000374,
-      "loss": 1.0323,
+      "loss": 0.9741,
       "step": 377
     },
     {
       "epoch": 6.87,
       "learning_rate": 0.000375,
-      "loss": 1.0578,
+      "loss": 1.0515,
       "step": 378
     },
     {
       "epoch": 6.89,
       "learning_rate": 0.00037600000000000003,
-      "loss": 1.0758,
+      "loss": 1.0496,
       "step": 379
     },
     {
       "epoch": 6.91,
       "learning_rate": 0.000377,
-      "loss": 1.1099,
+      "loss": 1.0859,
       "step": 380
     },
     {
       "epoch": 6.93,
       "learning_rate": 0.000378,
-      "loss": 1.0668,
+      "loss": 1.0638,
       "step": 381
     },
     {
       "epoch": 6.95,
       "learning_rate": 0.000379,
-      "loss": 0.9397,
+      "loss": 0.9455,
       "step": 382
     },
     {
       "epoch": 6.96,
       "learning_rate": 0.00038,
-      "loss": 1.0402,
+      "loss": 0.9733,
       "step": 383
     },
     {
       "epoch": 6.98,
       "learning_rate": 0.000381,
-      "loss": 1.0519,
+      "loss": 1.0388,
       "step": 384
     },
     {
       "epoch": 7.0,
       "learning_rate": 0.000382,
-      "loss": 1.0441,
+      "loss": 0.9567,
       "step": 385
     },
     {
       "epoch": 7.02,
       "learning_rate": 0.00038300000000000004,
-      "loss": 0.935,
+      "loss": 0.8843,
       "step": 386
     },
     {
       "epoch": 7.04,
       "learning_rate": 0.000384,
-      "loss": 0.9422,
+      "loss": 0.9399,
       "step": 387
     },
     {
       "epoch": 7.05,
       "learning_rate": 0.00038500000000000003,
-      "loss": 0.9731,
+      "loss": 0.9382,
       "step": 388
     },
     {
       "epoch": 7.07,
       "learning_rate": 0.000386,
-      "loss": 0.9508,
+      "loss": 0.9171,
       "step": 389
     },
     {
       "epoch": 7.09,
       "learning_rate": 0.00038700000000000003,
-      "loss": 1.0009,
+      "loss": 0.9416,
       "step": 390
     },
     {
       "epoch": 7.11,
       "learning_rate": 0.000388,
-      "loss": 0.9187,
+      "loss": 0.9034,
       "step": 391
     },
     {
       "epoch": 7.13,
       "learning_rate": 0.000389,
-      "loss": 1.0291,
+      "loss": 0.9505,
       "step": 392
     },
     {
       "epoch": 7.15,
       "learning_rate": 0.00039000000000000005,
-      "loss": 0.9495,
+      "loss": 0.9661,
       "step": 393
     },
     {
       "epoch": 7.16,
       "learning_rate": 0.000391,
-      "loss": 0.9221,
+      "loss": 0.9266,
       "step": 394
     },
     {
       "epoch": 7.18,
       "learning_rate": 0.00039200000000000004,
-      "loss": 0.901,
+      "loss": 0.8559,
       "step": 395
     },
     {
       "epoch": 7.2,
       "learning_rate": 0.000393,
-      "loss": 0.9298,
+      "loss": 0.8785,
       "step": 396
     },
     {
       "epoch": 7.22,
       "learning_rate": 0.00039400000000000004,
-      "loss": 0.9699,
+      "loss": 0.8587,
       "step": 397
     },
     {
       "epoch": 7.24,
       "learning_rate": 0.000395,
-      "loss": 1.1264,
+      "loss": 1.0522,
       "step": 398
     },
     {
       "epoch": 7.25,
       "learning_rate": 0.00039600000000000003,
-      "loss": 0.9658,
+      "loss": 0.9128,
       "step": 399
     },
     {
       "epoch": 7.27,
       "learning_rate": 0.00039700000000000005,
-      "loss": 0.8804,
+      "loss": 0.8792,
       "step": 400
     },
     {
       "epoch": 7.29,
       "learning_rate": 0.000398,
-      "loss": 0.9635,
+      "loss": 0.9411,
       "step": 401
     },
     {
       "epoch": 7.31,
       "learning_rate": 0.00039900000000000005,
-      "loss": 0.9579,
+      "loss": 0.9392,
       "step": 402
     },
     {
       "epoch": 7.33,
       "learning_rate": 0.0004,
-      "loss": 0.9887,
+      "loss": 0.9501,
       "step": 403
     },
     {
       "epoch": 7.35,
       "learning_rate": 0.00040100000000000004,
-      "loss": 0.9273,
+      "loss": 0.8803,
       "step": 404
     },
     {
       "epoch": 7.36,
       "learning_rate": 0.000402,
-      "loss": 1.0134,
+      "loss": 0.9914,
       "step": 405
     },
     {
       "epoch": 7.38,
       "learning_rate": 0.00040300000000000004,
-      "loss": 0.925,
+      "loss": 0.8893,
       "step": 406
     },
     {
       "epoch": 7.4,
       "learning_rate": 0.000404,
-      "loss": 0.9501,
+      "loss": 0.9668,
       "step": 407
     },
     {
       "epoch": 7.42,
       "learning_rate": 0.00040500000000000003,
-      "loss": 0.9671,
+      "loss": 0.9053,
       "step": 408
     },
     {
       "epoch": 7.44,
       "learning_rate": 0.00040600000000000006,
-      "loss": 0.9789,
+      "loss": 0.9762,
       "step": 409
     },
     {
       "epoch": 7.45,
       "learning_rate": 0.00040699999999999997,
-      "loss": 0.9534,
+      "loss": 0.9441,
       "step": 410
     },
     {
       "epoch": 7.47,
       "learning_rate": 0.000408,
-      "loss": 0.9791,
+      "loss": 0.9539,
       "step": 411
     },
     {
       "epoch": 7.49,
       "learning_rate": 0.00040899999999999997,
-      "loss": 0.9946,
+      "loss": 0.953,
       "step": 412
     },
     {
       "epoch": 7.51,
       "learning_rate": 0.00041,
-      "loss": 0.9503,
+      "loss": 0.8984,
       "step": 413
     },
     {
       "epoch": 7.53,
       "learning_rate": 0.00041099999999999996,
-      "loss": 0.9521,
+      "loss": 0.9295,
       "step": 414
     },
     {
       "epoch": 7.55,
       "learning_rate": 0.000412,
-      "loss": 0.8911,
+      "loss": 0.8109,
       "step": 415
     },
     {
       "epoch": 7.56,
       "learning_rate": 0.000413,
-      "loss": 0.9995,
+      "loss": 0.9197,
       "step": 416
     },
     {
       "epoch": 7.58,
       "learning_rate": 0.000414,
-      "loss": 0.8599,
+      "loss": 0.8734,
       "step": 417
     },
     {
       "epoch": 7.6,
       "learning_rate": 0.000415,
-      "loss": 0.9807,
+      "loss": 0.9178,
       "step": 418
     },
     {
       "epoch": 7.62,
       "learning_rate": 0.000416,
-      "loss": 0.9763,
+      "loss": 0.8988,
       "step": 419
     },
     {
       "epoch": 7.64,
       "learning_rate": 0.000417,
-      "loss": 0.9907,
+      "loss": 0.9398,
       "step": 420
     },
     {
       "epoch": 7.65,
       "learning_rate": 0.00041799999999999997,
-      "loss": 0.9902,
+      "loss": 0.9222,
       "step": 421
     },
     {
       "epoch": 7.67,
       "learning_rate": 0.000419,
-      "loss": 0.9357,
+      "loss": 0.9112,
       "step": 422
     },
     {
       "epoch": 7.69,
       "learning_rate": 0.00042,
-      "loss": 0.9726,
+      "loss": 0.921,
       "step": 423
     },
     {
       "epoch": 7.71,
       "learning_rate": 0.000421,
-      "loss": 0.8948,
+      "loss": 0.8122,
       "step": 424
     },
     {
       "epoch": 7.73,
       "learning_rate": 0.000422,
-      "loss": 0.8388,
+      "loss": 0.8219,
       "step": 425
     },
     {
       "epoch": 7.75,
       "learning_rate": 0.000423,
-      "loss": 0.9411,
+      "loss": 0.8916,
       "step": 426
     },
     {
       "epoch": 7.76,
       "learning_rate": 0.000424,
-      "loss": 0.9462,
+      "loss": 0.877,
       "step": 427
     },
     {
       "epoch": 7.78,
       "learning_rate": 0.000425,
-      "loss": 0.9624,
+      "loss": 0.9434,
       "step": 428
     },
     {
       "epoch": 7.8,
       "learning_rate": 0.000426,
-      "loss": 0.9028,
+      "loss": 0.9098,
       "step": 429
     },
     {
       "epoch": 7.82,
       "learning_rate": 0.000427,
-      "loss": 0.9561,
+      "loss": 0.8793,
       "step": 430
     },
     {
       "epoch": 7.84,
       "learning_rate": 0.000428,
-      "loss": 0.9102,
+      "loss": 0.8746,
       "step": 431
     },
     {
       "epoch": 7.85,
       "learning_rate": 0.000429,
-      "loss": 0.9069,
+      "loss": 0.8313,
       "step": 432
     },
     {
       "epoch": 7.87,
       "learning_rate": 0.00043,
-      "loss": 0.9072,
+      "loss": 0.8046,
       "step": 433
     },
     {
       "epoch": 7.89,
       "learning_rate": 0.000431,
-      "loss": 1.0516,
+      "loss": 1.0071,
       "step": 434
     },
     {
       "epoch": 7.91,
       "learning_rate": 0.000432,
-      "loss": 1.0194,
+      "loss": 0.9535,
       "step": 435
     },
     {
       "epoch": 7.93,
       "learning_rate": 0.000433,
-      "loss": 0.9752,
+      "loss": 0.8777,
       "step": 436
     },
     {
       "epoch": 7.95,
       "learning_rate": 0.00043400000000000003,
-      "loss": 1.0167,
+      "loss": 0.9838,
       "step": 437
     },
     {
       "epoch": 7.96,
       "learning_rate": 0.000435,
-      "loss": 0.9783,
+      "loss": 0.9856,
       "step": 438
     },
     {
       "epoch": 7.98,
       "learning_rate": 0.000436,
-      "loss": 0.9635,
+      "loss": 0.9119,
       "step": 439
     },
     {
       "epoch": 8.0,
       "learning_rate": 0.000437,
-      "loss": 1.0439,
+      "loss": 1.0307,
       "step": 440
     },
     {
       "epoch": 8.02,
       "learning_rate": 0.000438,
-      "loss": 0.9118,
+      "loss": 0.9632,
       "step": 441
     },
     {
       "epoch": 8.04,
       "learning_rate": 0.000439,
-      "loss": 0.9312,
+      "loss": 0.9094,
       "step": 442
     },
     {
       "epoch": 8.05,
       "learning_rate": 0.00044,
-      "loss": 0.8243,
+      "loss": 0.7813,
       "step": 443
     },
     {
       "epoch": 8.07,
       "learning_rate": 0.000441,
-      "loss": 0.9456,
+      "loss": 0.9449,
       "step": 444
     },
     {
       "epoch": 8.09,
       "learning_rate": 0.000442,
-      "loss": 0.8935,
+      "loss": 0.8776,
       "step": 445
     },
     {
       "epoch": 8.11,
       "learning_rate": 0.00044300000000000003,
-      "loss": 0.8975,
+      "loss": 0.8658,
       "step": 446
     },
     {
       "epoch": 8.13,
       "learning_rate": 0.000444,
-      "loss": 0.93,
+      "loss": 0.9551,
       "step": 447
     },
     {
       "epoch": 8.15,
       "learning_rate": 0.00044500000000000003,
-      "loss": 0.8504,
+      "loss": 0.847,
       "step": 448
     },
     {
       "epoch": 8.16,
       "learning_rate": 0.000446,
-      "loss": 0.7908,
+      "loss": 0.7381,
       "step": 449
     },
     {
       "epoch": 8.18,
       "learning_rate": 0.000447,
-      "loss": 1.0138,
+      "loss": 0.9999,
       "step": 450
     },
     {
       "epoch": 8.2,
       "learning_rate": 0.000448,
-      "loss": 0.8623,
+      "loss": 0.9006,
       "step": 451
     },
     {
       "epoch": 8.22,
       "learning_rate": 0.000449,
-      "loss": 0.9869,
+      "loss": 0.948,
       "step": 452
     },
     {
       "epoch": 8.24,
       "learning_rate": 0.00045000000000000004,
-      "loss": 0.8589,
+      "loss": 0.8423,
       "step": 453
     },
     {
       "epoch": 8.25,
       "learning_rate": 0.000451,
-      "loss": 0.9284,
+      "loss": 0.937,
       "step": 454
     },
     {
       "epoch": 8.27,
       "learning_rate": 0.00045200000000000004,
-      "loss": 0.8189,
+      "loss": 0.8166,
       "step": 455
     },
     {
       "epoch": 8.29,
       "learning_rate": 0.000453,
-      "loss": 0.8521,
+      "loss": 0.8076,
       "step": 456
     },
     {
       "epoch": 8.31,
       "learning_rate": 0.00045400000000000003,
-      "loss": 0.8163,
+      "loss": 0.8071,
       "step": 457
     },
     {
       "epoch": 8.33,
       "learning_rate": 0.000455,
-      "loss": 0.9288,
+      "loss": 0.9313,
       "step": 458
     },
     {
       "epoch": 8.35,
       "learning_rate": 0.000456,
-      "loss": 0.8635,
+      "loss": 0.7923,
       "step": 459
     },
     {
       "epoch": 8.36,
       "learning_rate": 0.00045700000000000005,
-      "loss": 0.966,
+      "loss": 0.8642,
       "step": 460
     },
     {
       "epoch": 8.38,
       "learning_rate": 0.000458,
-      "loss": 0.8354,
+      "loss": 0.7654,
       "step": 461
     },
     {
       "epoch": 8.4,
       "learning_rate": 0.00045900000000000004,
-      "loss": 0.8589,
+      "loss": 0.8239,
       "step": 462
     },
     {
       "epoch": 8.42,
       "learning_rate": 0.00046,
-      "loss": 0.905,
+      "loss": 0.861,
       "step": 463
     },
     {
       "epoch": 8.44,
       "learning_rate": 0.00046100000000000004,
-      "loss": 0.8545,
+      "loss": 0.8208,
       "step": 464
     },
     {
       "epoch": 8.45,
       "learning_rate": 0.000462,
-      "loss": 0.8957,
+      "loss": 0.908,
       "step": 465
     },
     {
       "epoch": 8.47,
       "learning_rate": 0.00046300000000000003,
-      "loss": 0.8613,
+      "loss": 0.8197,
       "step": 466
     },
     {
       "epoch": 8.49,
       "learning_rate": 0.00046400000000000006,
-      "loss": 0.7599,
+      "loss": 0.8069,
       "step": 467
     },
     {
       "epoch": 8.51,
       "learning_rate": 0.000465,
-      "loss": 0.857,
+      "loss": 0.8605,
       "step": 468
     },
     {
       "epoch": 8.53,
       "learning_rate": 0.00046600000000000005,
-      "loss": 0.8865,
+      "loss": 0.8243,
       "step": 469
     },
     {
       "epoch": 8.55,
       "learning_rate": 0.000467,
-      "loss": 0.8625,
+      "loss": 0.8503,
       "step": 470
     },
     {
       "epoch": 8.56,
       "learning_rate": 0.00046800000000000005,
-      "loss": 0.8662,
+      "loss": 0.7969,
       "step": 471
     },
     {
       "epoch": 8.58,
       "learning_rate": 0.00046899999999999996,
-      "loss": 0.89,
+      "loss": 0.854,
       "step": 472
     },
     {
       "epoch": 8.6,
       "learning_rate": 0.00047,
-      "loss": 0.8352,
+      "loss": 0.777,
       "step": 473
     },
     {
       "epoch": 8.62,
       "learning_rate": 0.000471,
-      "loss": 0.8122,
+      "loss": 0.7614,
       "step": 474
     },
     {
       "epoch": 8.64,
       "learning_rate": 0.000472,
-      "loss": 0.8122,
+      "loss": 0.8055,
       "step": 475
     },
     {
       "epoch": 8.65,
       "learning_rate": 0.000473,
-      "loss": 0.8249,
+      "loss": 0.8254,
       "step": 476
     },
     {
       "epoch": 8.67,
       "learning_rate": 0.000474,
-      "loss": 0.8663,
+      "loss": 0.852,
       "step": 477
     },
     {
       "epoch": 8.69,
       "learning_rate": 0.000475,
-      "loss": 0.9019,
+      "loss": 0.8888,
       "step": 478
     },
     {
       "epoch": 8.71,
       "learning_rate": 0.00047599999999999997,
-      "loss": 0.8454,
+      "loss": 0.8364,
       "step": 479
     },
     {
       "epoch": 8.73,
       "learning_rate": 0.000477,
-      "loss": 0.811,
+      "loss": 0.7961,
       "step": 480
     },
     {
       "epoch": 8.75,
       "learning_rate": 0.00047799999999999996,
-      "loss": 0.9226,
+      "loss": 0.8963,
       "step": 481
     },
     {
       "epoch": 8.76,
       "learning_rate": 0.000479,
-      "loss": 0.9817,
+      "loss": 0.9474,
       "step": 482
     },
     {
       "epoch": 8.78,
       "learning_rate": 0.00048,
-      "loss": 0.8788,
+      "loss": 0.8477,
       "step": 483
     },
     {
       "epoch": 8.8,
       "learning_rate": 0.000481,
-      "loss": 0.8805,
+      "loss": 0.8735,
       "step": 484
     },
     {
       "epoch": 8.82,
       "learning_rate": 0.000482,
-      "loss": 0.8645,
+      "loss": 0.8844,
       "step": 485
     },
     {
       "epoch": 8.84,
       "learning_rate": 0.000483,
-      "loss": 0.7519,
+      "loss": 0.7343,
       "step": 486
     },
     {
       "epoch": 8.85,
       "learning_rate": 0.000484,
-      "loss": 0.814,
+      "loss": 0.7789,
       "step": 487
     },
     {
       "epoch": 8.87,
       "learning_rate": 0.00048499999999999997,
-      "loss": 0.8516,
+      "loss": 0.8413,
       "step": 488
     },
     {
       "epoch": 8.89,
       "learning_rate": 0.000486,
-      "loss": 0.9208,
+      "loss": 0.8938,
       "step": 489
     },
     {
       "epoch": 8.91,
       "learning_rate": 0.000487,
-      "loss": 0.8605,
+      "loss": 0.8012,
       "step": 490
     },
     {
       "epoch": 8.93,
       "learning_rate": 0.000488,
-      "loss": 0.8389,
+      "loss": 0.8074,
       "step": 491
     },
     {
       "epoch": 8.95,
       "learning_rate": 0.000489,
-      "loss": 0.8912,
+      "loss": 0.859,
       "step": 492
     },
     {
       "epoch": 8.96,
       "learning_rate": 0.00049,
-      "loss": 0.8081,
+      "loss": 0.8264,
       "step": 493
     },
     {
       "epoch": 8.98,
       "learning_rate": 0.000491,
-      "loss": 0.8162,
+      "loss": 0.8048,
       "step": 494
     },
     {
       "epoch": 9.0,
       "learning_rate": 0.000492,
-      "loss": 0.8001,
+      "loss": 0.8173,
       "step": 495
     },
     {
       "epoch": 9.02,
       "learning_rate": 0.0004930000000000001,
-      "loss": 0.898,
+      "loss": 0.8447,
       "step": 496
     },
     {
       "epoch": 9.04,
       "learning_rate": 0.000494,
-      "loss": 0.7777,
+      "loss": 0.7551,
       "step": 497
     },
     {
       "epoch": 9.05,
       "learning_rate": 0.000495,
-      "loss": 0.7375,
+      "loss": 0.7038,
       "step": 498
     },
     {
       "epoch": 9.07,
       "learning_rate": 0.000496,
-      "loss": 0.8138,
+      "loss": 0.795,
       "step": 499
     },
     {
       "epoch": 9.09,
       "learning_rate": 0.000497,
-      "loss": 0.7897,
+      "loss": 0.7638,
       "step": 500
     },
     {
       "epoch": 9.09,
-      "eval_cer": 0.14910268519767914,
-      "eval_loss": 0.48732584714889526,
-      "eval_runtime": 14.2949,
-      "eval_samples_per_second": 115.216,
-      "eval_steps_per_second": 1.819,
-      "eval_wer": 0.6005515269124706,
+      "eval_cer": 0.13325537114499467,
+      "eval_loss": 0.47626373171806335,
+      "eval_runtime": 14.4634,
+      "eval_samples_per_second": 113.874,
+      "eval_steps_per_second": 1.798,
+      "eval_wer": 0.5313042590133796,
       "step": 500
     },
     {
       "epoch": 9.11,
       "learning_rate": 0.000498,
-      "loss": 0.8201,
+      "loss": 0.7801,
       "step": 501
     },
     {
       "epoch": 9.13,
       "learning_rate": 0.000499,
-      "loss": 0.8858,
+      "loss": 0.8516,
       "step": 502
     },
     {
       "epoch": 9.15,
       "learning_rate": 0.0005,
-      "loss": 0.7948,
+      "loss": 0.7815,
       "step": 503
     },
     {
       "epoch": 9.16,
-      "learning_rate": 0.0004997777777777778,
-      "loss": 0.8177,
+      "learning_rate": 0.0004998620689655172,
+      "loss": 0.83,
       "step": 504
     },
     {
       "epoch": 9.18,
-      "learning_rate": 0.0004995555555555555,
-      "loss": 0.8828,
+      "learning_rate": 0.0004997241379310345,
+      "loss": 0.8215,
       "step": 505
     },
     {
       "epoch": 9.2,
-      "learning_rate": 0.0004993333333333334,
-      "loss": 0.7545,
+      "learning_rate": 0.0004995862068965518,
+      "loss": 0.7472,
       "step": 506
     },
     {
       "epoch": 9.22,
-      "learning_rate": 0.0004991111111111111,
-      "loss": 0.7634,
+      "learning_rate": 0.000499448275862069,
+      "loss": 0.8104,
       "step": 507
     },
     {
       "epoch": 9.24,
-      "learning_rate": 0.0004988888888888889,
-      "loss": 0.8756,
+      "learning_rate": 0.0004993103448275862,
+      "loss": 0.8891,
       "step": 508
     },
     {
       "epoch": 9.25,
-      "learning_rate": 0.0004986666666666667,
-      "loss": 0.7918,
+      "learning_rate": 0.0004991724137931035,
+      "loss": 0.7957,
       "step": 509
     },
     {
       "epoch": 9.27,
-      "learning_rate": 0.0004984444444444445,
-      "loss": 0.7582,
+      "learning_rate": 0.0004990344827586207,
+      "loss": 0.7722,
       "step": 510
     },
     {
       "epoch": 9.29,
-      "learning_rate": 0.0004982222222222223,
-      "loss": 0.9094,
+      "learning_rate": 0.0004988965517241379,
+      "loss": 0.8653,
       "step": 511
     },
     {
       "epoch": 9.31,
-      "learning_rate": 0.000498,
-      "loss": 0.8351,
+      "learning_rate": 0.0004987586206896551,
+      "loss": 0.7846,
       "step": 512
     },
     {
       "epoch": 9.33,
-      "learning_rate": 0.0004977777777777778,
-      "loss": 0.874,
+      "learning_rate": 0.0004986206896551725,
+      "loss": 0.8281,
       "step": 513
     },
     {
       "epoch": 9.35,
-      "learning_rate": 0.0004975555555555556,
-      "loss": 0.8472,
+      "learning_rate": 0.0004984827586206897,
+      "loss": 0.7657,
       "step": 514
     },
     {
       "epoch": 9.36,
-      "learning_rate": 0.0004973333333333334,
-      "loss": 0.8376,
+      "learning_rate": 0.0004983448275862069,
+      "loss": 0.7913,
       "step": 515
     },
     {
       "epoch": 9.38,
-      "learning_rate": 0.0004971111111111111,
-      "loss": 0.7541,
+      "learning_rate": 0.0004982068965517241,
+      "loss": 0.7778,
       "step": 516
     },
     {
       "epoch": 9.4,
-      "learning_rate": 0.0004968888888888889,
-      "loss": 0.7622,
+      "learning_rate": 0.0004980689655172414,
+      "loss": 0.7558,
       "step": 517
     },
     {
       "epoch": 9.42,
-      "learning_rate": 0.0004966666666666666,
-      "loss": 0.8904,
+      "learning_rate": 0.0004979310344827586,
+      "loss": 0.8889,
       "step": 518
     },
     {
       "epoch": 9.44,
-      "learning_rate": 0.0004964444444444444,
-      "loss": 0.7626,
+      "learning_rate": 0.0004977931034482758,
+      "loss": 0.8146,
       "step": 519
     },
     {
       "epoch": 9.45,
-      "learning_rate": 0.0004962222222222222,
-      "loss": 0.8221,
+      "learning_rate": 0.0004976551724137932,
+      "loss": 0.8435,
       "step": 520
     },
     {
       "epoch": 9.47,
-      "learning_rate": 0.000496,
-      "loss": 0.7658,
+      "learning_rate": 0.0004975172413793104,
+      "loss": 0.7772,
       "step": 521
     },
     {
       "epoch": 9.49,
-      "learning_rate": 0.0004957777777777778,
-      "loss": 0.7445,
+      "learning_rate": 0.0004973793103448276,
+      "loss": 0.7214,
       "step": 522
     },
     {
       "epoch": 9.51,
-      "learning_rate": 0.0004955555555555556,
-      "loss": 0.7052,
+      "learning_rate": 0.0004972413793103449,
+      "loss": 0.684,
       "step": 523
     },
     {
       "epoch": 9.53,
-      "learning_rate": 0.0004953333333333334,
-      "loss": 0.8351,
+      "learning_rate": 0.0004971034482758621,
+      "loss": 0.8127,
       "step": 524
     },
     {
       "epoch": 9.55,
-      "learning_rate": 0.0004951111111111111,
-      "loss": 0.812,
+      "learning_rate": 0.0004969655172413793,
+      "loss": 0.7809,
       "step": 525
     },
     {
       "epoch": 9.56,
-      "learning_rate": 0.0004948888888888889,
-      "loss": 0.761,
+      "learning_rate": 0.0004968275862068965,
+      "loss": 0.7843,
       "step": 526
     },
     {
       "epoch": 9.58,
-      "learning_rate": 0.0004946666666666667,
-      "loss": 0.8359,
+      "learning_rate": 0.0004966896551724139,
+      "loss": 0.8201,
       "step": 527
     },
     {
       "epoch": 9.6,
-      "learning_rate": 0.0004944444444444445,
-      "loss": 0.8409,
+      "learning_rate": 0.0004965517241379311,
+      "loss": 0.8629,
       "step": 528
     },
     {
       "epoch": 9.62,
-      "learning_rate": 0.0004942222222222223,
-      "loss": 0.8239,
+      "learning_rate": 0.0004964137931034483,
+      "loss": 0.7867,
       "step": 529
     },
     {
       "epoch": 9.64,
-      "learning_rate": 0.000494,
-      "loss": 0.7098,
+      "learning_rate": 0.0004962758620689655,
+      "loss": 0.6753,
       "step": 530
     },
     {
       "epoch": 9.65,
-      "learning_rate": 0.0004937777777777778,
-      "loss": 0.8208,
+      "learning_rate": 0.0004961379310344828,
+      "loss": 0.7942,
       "step": 531
     },
     {
       "epoch": 9.67,
-      "learning_rate": 0.0004935555555555556,
-      "loss": 0.7843,
+      "learning_rate": 0.000496,
+      "loss": 0.7328,
       "step": 532
     },
     {
       "epoch": 9.69,
-      "learning_rate": 0.0004933333333333334,
-      "loss": 0.8878,
+      "learning_rate": 0.0004958620689655172,
+      "loss": 0.8218,
       "step": 533
     },
     {
       "epoch": 9.71,
-      "learning_rate": 0.0004931111111111112,
-      "loss": 0.8092,
+      "learning_rate": 0.0004957241379310344,
+      "loss": 0.7961,
       "step": 534
     },
     {
       "epoch": 9.73,
-      "learning_rate": 0.0004928888888888888,
-      "loss": 0.8474,
+      "learning_rate": 0.0004955862068965518,
+      "loss": 0.8286,
       "step": 535
     },
     {
       "epoch": 9.75,
-      "learning_rate": 0.0004926666666666666,
-      "loss": 0.9214,
+      "learning_rate": 0.000495448275862069,
+      "loss": 0.8324,
       "step": 536
     },
     {
       "epoch": 9.76,
-      "learning_rate": 0.0004924444444444444,
-      "loss": 0.8417,
+      "learning_rate": 0.0004953103448275862,
+      "loss": 0.7896,
       "step": 537
     },
     {
       "epoch": 9.78,
-      "learning_rate": 0.0004922222222222222,
-      "loss": 0.7836,
+      "learning_rate": 0.0004951724137931035,
+      "loss": 0.734,
       "step": 538
     },
     {
       "epoch": 9.8,
-      "learning_rate": 0.000492,
-      "loss": 0.8103,
+      "learning_rate": 0.0004950344827586207,
+      "loss": 0.7828,
       "step": 539
     },
     {
       "epoch": 9.82,
-      "learning_rate": 0.0004917777777777777,
-      "loss": 0.8528,
+      "learning_rate": 0.0004948965517241379,
+      "loss": 0.8202,
       "step": 540
     },
     {
       "epoch": 9.84,
-      "learning_rate": 0.0004915555555555556,
-      "loss": 0.803,
+      "learning_rate": 0.0004947586206896551,
+      "loss": 0.75,
       "step": 541
     },
     {
       "epoch": 9.85,
-      "learning_rate": 0.0004913333333333333,
-      "loss": 0.743,
+      "learning_rate": 0.0004946206896551725,
+      "loss": 0.7352,
       "step": 542
     },
     {
       "epoch": 9.87,
-      "learning_rate": 0.0004911111111111111,
-      "loss": 0.8193,
+      "learning_rate": 0.0004944827586206897,
+      "loss": 0.7797,
       "step": 543
     },
     {
       "epoch": 9.89,
-      "learning_rate": 0.0004908888888888889,
-      "loss": 0.9015,
+      "learning_rate": 0.0004943448275862069,
+      "loss": 0.8053,
       "step": 544
     },
     {
       "epoch": 9.91,
-      "learning_rate": 0.0004906666666666667,
-      "loss": 0.8194,
+      "learning_rate": 0.0004942068965517241,
+      "loss": 0.8175,
       "step": 545
     },
     {
       "epoch": 9.93,
-      "learning_rate": 0.0004904444444444445,
-      "loss": 0.7959,
+      "learning_rate": 0.0004940689655172414,
+      "loss": 0.763,
       "step": 546
     },
     {
       "epoch": 9.95,
-      "learning_rate": 0.0004902222222222222,
-      "loss": 0.7617,
+      "learning_rate": 0.0004939310344827586,
+      "loss": 0.7418,
       "step": 547
     },
     {
       "epoch": 9.96,
-      "learning_rate": 0.00049,
-      "loss": 0.8079,
+      "learning_rate": 0.0004937931034482758,
+      "loss": 0.816,
       "step": 548
     },
     {
       "epoch": 9.98,
-      "learning_rate": 0.0004897777777777778,
-      "loss": 0.8178,
+      "learning_rate": 0.0004936551724137931,
+      "loss": 0.7919,
       "step": 549
     },
     {
       "epoch": 10.0,
-      "learning_rate": 0.0004895555555555556,
-      "loss": 0.823,
+      "learning_rate": 0.0004935172413793104,
+      "loss": 0.7819,
       "step": 550
     },
     {
       "epoch": 10.02,
-      "learning_rate": 0.0004893333333333334,
-      "loss": 0.8018,
+      "learning_rate": 0.0004933793103448276,
+      "loss": 0.7706,
       "step": 551
     },
     {
       "epoch": 10.04,
-      "learning_rate": 0.0004891111111111112,
-      "loss": 0.705,
+      "learning_rate": 0.0004932413793103448,
+      "loss": 0.6985,
       "step": 552
     },
     {
       "epoch": 10.05,
-      "learning_rate": 0.0004888888888888889,
-      "loss": 0.7259,
+      "learning_rate": 0.000493103448275862,
+      "loss": 0.6919,
       "step": 553
     },
     {
       "epoch": 10.07,
-      "learning_rate": 0.0004886666666666667,
-      "loss": 0.6615,
+      "learning_rate": 0.0004929655172413793,
+      "loss": 0.6819,
       "step": 554
     },
     {
       "epoch": 10.09,
-      "learning_rate": 0.0004884444444444445,
-      "loss": 0.7275,
+      "learning_rate": 0.0004928275862068965,
+      "loss": 0.6997,
       "step": 555
     },
     {
       "epoch": 10.11,
-      "learning_rate": 0.00048822222222222225,
-      "loss": 0.6734,
+      "learning_rate": 0.0004926896551724139,
+      "loss": 0.6376,
       "step": 556
     },
     {
       "epoch": 10.13,
-      "learning_rate": 0.000488,
-      "loss": 0.7628,
+      "learning_rate": 0.0004925517241379311,
+      "loss": 0.7473,
       "step": 557
     },
     {
       "epoch": 10.15,
-      "learning_rate": 0.0004877777777777778,
-      "loss": 0.7396,
+      "learning_rate": 0.0004924137931034483,
+      "loss": 0.677,
       "step": 558
     },
     {
       "epoch": 10.16,
-      "learning_rate": 0.0004875555555555556,
-      "loss": 0.872,
+      "learning_rate": 0.0004922758620689655,
+      "loss": 0.7951,
       "step": 559
     },
     {
       "epoch": 10.18,
-      "learning_rate": 0.0004873333333333333,
-      "loss": 0.7527,
+      "learning_rate": 0.0004921379310344828,
+      "loss": 0.7393,
       "step": 560
     },
     {
       "epoch": 10.2,
-      "learning_rate": 0.0004871111111111111,
-      "loss": 0.8061,
+      "learning_rate": 0.000492,
+      "loss": 0.8045,
       "step": 561
     },
     {
       "epoch": 10.22,
-      "learning_rate": 0.00048688888888888886,
-      "loss": 0.7303,
+      "learning_rate": 0.0004918620689655172,
+      "loss": 0.7415,
       "step": 562
     },
     {
       "epoch": 10.24,
-      "learning_rate": 0.0004866666666666667,
-      "loss": 0.7925,
+      "learning_rate": 0.0004917241379310345,
+      "loss": 0.8347,
       "step": 563
     },
     {
       "epoch": 10.25,
-      "learning_rate": 0.00048644444444444446,
-      "loss": 0.7377,
+      "learning_rate": 0.0004915862068965518,
+      "loss": 0.7404,
       "step": 564
     },
     {
       "epoch": 10.27,
-      "learning_rate": 0.00048622222222222225,
-      "loss": 0.6792,
+      "learning_rate": 0.000491448275862069,
+      "loss": 0.6586,
       "step": 565
     },
     {
       "epoch": 10.29,
-      "learning_rate": 0.000486,
-      "loss": 0.7468,
+      "learning_rate": 0.0004913103448275862,
+      "loss": 0.745,
       "step": 566
     },
     {
       "epoch": 10.31,
-      "learning_rate": 0.00048577777777777774,
-      "loss": 0.7418,
+      "learning_rate": 0.0004911724137931034,
+      "loss": 0.7126,
       "step": 567
     },
     {
       "epoch": 10.33,
-      "learning_rate": 0.0004855555555555556,
-      "loss": 0.7246,
+      "learning_rate": 0.0004910344827586207,
+      "loss": 0.7129,
       "step": 568
     },
     {
       "epoch": 10.35,
-      "learning_rate": 0.00048533333333333333,
-      "loss": 0.7183,
+      "learning_rate": 0.0004908965517241379,
+      "loss": 0.7776,
       "step": 569
     },
     {
       "epoch": 10.36,
-      "learning_rate": 0.0004851111111111111,
-      "loss": 0.7317,
+      "learning_rate": 0.0004907586206896552,
+      "loss": 0.7537,
       "step": 570
     },
     {
       "epoch": 10.38,
-      "learning_rate": 0.00048488888888888887,
-      "loss": 0.767,
+      "learning_rate": 0.0004906206896551725,
+      "loss": 0.7965,
       "step": 571
     },
     {
       "epoch": 10.4,
-      "learning_rate": 0.0004846666666666667,
-      "loss": 0.7288,
+      "learning_rate": 0.0004904827586206897,
+      "loss": 0.7261,
       "step": 572
     },
     {
       "epoch": 10.42,
-      "learning_rate": 0.00048444444444444446,
-      "loss": 0.712,
+      "learning_rate": 0.0004903448275862069,
+      "loss": 0.7281,
       "step": 573
     },
     {
       "epoch": 10.44,
-      "learning_rate": 0.0004842222222222222,
-      "loss": 0.6407,
+      "learning_rate": 0.0004902068965517241,
+      "loss": 0.6622,
       "step": 574
     },
     {
       "epoch": 10.45,
-      "learning_rate": 0.000484,
-      "loss": 0.7519,
+      "learning_rate": 0.0004900689655172414,
+      "loss": 0.7809,
       "step": 575
     },
     {
       "epoch": 10.47,
-      "learning_rate": 0.00048377777777777774,
-      "loss": 0.7959,
+      "learning_rate": 0.0004899310344827586,
+      "loss": 0.8246,
       "step": 576
     },
     {
       "epoch": 10.49,
-      "learning_rate": 0.0004835555555555556,
-      "loss": 0.7328,
+      "learning_rate": 0.0004897931034482759,
+      "loss": 0.7526,
       "step": 577
     },
     {
       "epoch": 10.51,
-      "learning_rate": 0.00048333333333333334,
-      "loss": 0.6959,
+      "learning_rate": 0.0004896551724137931,
+      "loss": 0.6914,
       "step": 578
     },
     {
       "epoch": 10.53,
-      "learning_rate": 0.00048311111111111113,
-      "loss": 0.7278,
+      "learning_rate": 0.0004895172413793104,
+      "loss": 0.7453,
       "step": 579
     },
     {
       "epoch": 10.55,
-      "learning_rate": 0.0004828888888888889,
-      "loss": 0.711,
+      "learning_rate": 0.0004893793103448276,
+      "loss": 0.787,
       "step": 580
     },
     {
       "epoch": 10.56,
-      "learning_rate": 0.00048266666666666667,
-      "loss": 0.7566,
+      "learning_rate": 0.0004892413793103448,
+      "loss": 0.8064,
       "step": 581
     },
     {
       "epoch": 10.58,
-      "learning_rate": 0.00048244444444444447,
-      "loss": 0.788,
+      "learning_rate": 0.000489103448275862,
+      "loss": 0.7988,
       "step": 582
     },
     {
       "epoch": 10.6,
-      "learning_rate": 0.0004822222222222222,
-      "loss": 0.7082,
+      "learning_rate": 0.0004889655172413793,
+      "loss": 0.738,
       "step": 583
     },
     {
       "epoch": 10.62,
-      "learning_rate": 0.000482,
-      "loss": 0.7159,
+      "learning_rate": 0.0004888275862068966,
+      "loss": 0.737,
       "step": 584
     },
     {
       "epoch": 10.64,
-      "learning_rate": 0.0004817777777777778,
-      "loss": 0.7187,
+      "learning_rate": 0.0004886896551724138,
+      "loss": 0.7292,
       "step": 585
     },
     {
       "epoch": 10.65,
-      "learning_rate": 0.0004815555555555556,
-      "loss": 0.735,
+      "learning_rate": 0.000488551724137931,
+      "loss": 0.7983,
       "step": 586
     },
     {
       "epoch": 10.67,
-      "learning_rate": 0.00048133333333333334,
-      "loss": 0.7821,
+      "learning_rate": 0.0004884137931034483,
+      "loss": 0.7801,
       "step": 587
     },
     {
       "epoch": 10.69,
-      "learning_rate": 0.0004811111111111111,
-      "loss": 0.6459,
+      "learning_rate": 0.0004882758620689655,
+      "loss": 0.6709,
       "step": 588
     },
     {
       "epoch": 10.71,
-      "learning_rate": 0.0004808888888888889,
-      "loss": 0.7005,
+      "learning_rate": 0.0004881379310344828,
+      "loss": 0.7017,
       "step": 589
     },
     {
       "epoch": 10.73,
-      "learning_rate": 0.0004806666666666667,
-      "loss": 0.7725,
+      "learning_rate": 0.000488,
+      "loss": 0.7789,
       "step": 590
     },
     {
       "epoch": 10.75,
-      "learning_rate": 0.0004804444444444445,
-      "loss": 0.7134,
+      "learning_rate": 0.00048786206896551726,
+      "loss": 0.7244,
       "step": 591
     },
     {
       "epoch": 10.76,
-      "learning_rate": 0.0004802222222222222,
-      "loss": 0.7256,
+      "learning_rate": 0.00048772413793103447,
+      "loss": 0.747,
       "step": 592
     },
     {
       "epoch": 10.78,
-      "learning_rate": 0.00048,
-      "loss": 0.7379,
+      "learning_rate": 0.00048758620689655173,
+      "loss": 0.7144,
       "step": 593
     },
     {
       "epoch": 10.8,
-      "learning_rate": 0.0004797777777777778,
-      "loss": 0.7506,
+      "learning_rate": 0.00048744827586206895,
+      "loss": 0.8007,
       "step": 594
     },
     {
       "epoch": 10.82,
-      "learning_rate": 0.0004795555555555556,
-      "loss": 0.685,
+      "learning_rate": 0.0004873103448275862,
+      "loss": 0.6917,
       "step": 595
     },
     {
       "epoch": 10.84,
-      "learning_rate": 0.00047933333333333335,
-      "loss": 0.6432,
+      "learning_rate": 0.0004871724137931035,
+      "loss": 0.652,
       "step": 596
     },
     {
       "epoch": 10.85,
-      "learning_rate": 0.0004791111111111111,
-      "loss": 0.7538,
+      "learning_rate": 0.00048703448275862074,
+      "loss": 0.7099,
       "step": 597
     },
     {
       "epoch": 10.87,
-      "learning_rate": 0.0004788888888888889,
-      "loss": 0.8764,
+      "learning_rate": 0.00048689655172413796,
+      "loss": 0.8715,
       "step": 598
     },
     {
       "epoch": 10.89,
-      "learning_rate": 0.0004786666666666667,
-      "loss": 0.7448,
+      "learning_rate": 0.0004867586206896552,
+      "loss": 0.7094,
       "step": 599
     },
     {
       "epoch": 10.91,
-      "learning_rate": 0.0004784444444444445,
-      "loss": 0.771,
+      "learning_rate": 0.00048662068965517243,
+      "loss": 0.7733,
       "step": 600
     },
     {
       "epoch": 10.93,
-      "learning_rate": 0.0004782222222222222,
-      "loss": 0.7709,
+      "learning_rate": 0.0004864827586206897,
+      "loss": 0.7492,
       "step": 601
     },
     {
       "epoch": 10.95,
-      "learning_rate": 0.00047799999999999996,
-      "loss": 0.8086,
+      "learning_rate": 0.0004863448275862069,
+      "loss": 0.7635,
       "step": 602
     },
     {
       "epoch": 10.96,
-      "learning_rate": 0.0004777777777777778,
-      "loss": 0.7049,
+      "learning_rate": 0.0004862068965517241,
+      "loss": 0.7207,
       "step": 603
     },
     {
       "epoch": 10.98,
-      "learning_rate": 0.00047755555555555556,
-      "loss": 0.7623,
+      "learning_rate": 0.0004860689655172414,
+      "loss": 0.8461,
       "step": 604
     },
     {
       "epoch": 11.0,
-      "learning_rate": 0.00047733333333333335,
-      "loss": 0.6455,
+      "learning_rate": 0.0004859310344827586,
+      "loss": 0.7251,
       "step": 605
     },
     {
       "epoch": 11.02,
-      "learning_rate": 0.0004771111111111111,
-      "loss": 0.6705,
+      "learning_rate": 0.00048579310344827587,
+      "loss": 0.6876,
       "step": 606
     },
     {
       "epoch": 11.04,
-      "learning_rate": 0.0004768888888888889,
-      "loss": 0.6187,
+      "learning_rate": 0.0004856551724137931,
+      "loss": 0.6459,
       "step": 607
     },
     {
       "epoch": 11.05,
-      "learning_rate": 0.0004766666666666667,
-      "loss": 0.7443,
+      "learning_rate": 0.00048551724137931035,
+      "loss": 0.6729,
       "step": 608
     },
     {
       "epoch": 11.07,
-      "learning_rate": 0.0004764444444444445,
-      "loss": 0.6763,
+      "learning_rate": 0.00048537931034482756,
+      "loss": 0.6974,
       "step": 609
     },
     {
       "epoch": 11.09,
-      "learning_rate": 0.0004762222222222222,
-      "loss": 0.7158,
+      "learning_rate": 0.0004852413793103448,
+      "loss": 0.7437,
       "step": 610
     },
     {
       "epoch": 11.11,
-      "learning_rate": 0.00047599999999999997,
-      "loss": 0.6664,
+      "learning_rate": 0.00048510344827586204,
+      "loss": 0.6952,
       "step": 611
     },
     {
       "epoch": 11.13,
-      "learning_rate": 0.0004757777777777778,
-      "loss": 0.6999,
+      "learning_rate": 0.0004849655172413793,
+      "loss": 0.7107,
       "step": 612
     },
     {
       "epoch": 11.15,
-      "learning_rate": 0.00047555555555555556,
-      "loss": 0.731,
+      "learning_rate": 0.00048482758620689657,
+      "loss": 0.6813,
       "step": 613
     },
     {
       "epoch": 11.16,
-      "learning_rate": 0.00047533333333333336,
-      "loss": 0.7094,
+      "learning_rate": 0.00048468965517241383,
+      "loss": 0.7323,
       "step": 614
     },
     {
       "epoch": 11.18,
-      "learning_rate": 0.0004751111111111111,
-      "loss": 0.7106,
+      "learning_rate": 0.00048455172413793105,
+      "loss": 0.6802,
       "step": 615
     },
     {
       "epoch": 11.2,
-      "learning_rate": 0.00047488888888888895,
-      "loss": 0.7095,
+      "learning_rate": 0.0004844137931034483,
+      "loss": 0.6268,
       "step": 616
     },
     {
       "epoch": 11.22,
-      "learning_rate": 0.0004746666666666667,
-      "loss": 0.7114,
+      "learning_rate": 0.0004842758620689655,
+      "loss": 0.6579,
       "step": 617
     },
     {
       "epoch": 11.24,
-      "learning_rate": 0.00047444444444444444,
-      "loss": 0.727,
+      "learning_rate": 0.0004841379310344828,
+      "loss": 0.7563,
       "step": 618
     },
     {
       "epoch": 11.25,
-      "learning_rate": 0.00047422222222222223,
-      "loss": 0.7587,
+      "learning_rate": 0.000484,
+      "loss": 0.7744,
       "step": 619
     },
     {
       "epoch": 11.27,
-      "learning_rate": 0.000474,
-      "loss": 0.6688,
+      "learning_rate": 0.00048386206896551727,
+      "loss": 0.6484,
       "step": 620
     },
     {
       "epoch": 11.29,
-      "learning_rate": 0.0004737777777777778,
-      "loss": 0.6934,
+      "learning_rate": 0.0004837241379310345,
+      "loss": 0.7153,
       "step": 621
     },
     {
       "epoch": 11.31,
-      "learning_rate": 0.00047355555555555557,
-      "loss": 0.681,
+      "learning_rate": 0.00048358620689655174,
+      "loss": 0.6888,
       "step": 622
     },
     {
       "epoch": 11.33,
-      "learning_rate": 0.00047333333333333336,
-      "loss": 0.6717,
+      "learning_rate": 0.00048344827586206896,
+      "loss": 0.6684,
       "step": 623
     },
     {
       "epoch": 11.35,
-      "learning_rate": 0.0004731111111111111,
-      "loss": 0.7529,
+      "learning_rate": 0.0004833103448275862,
+      "loss": 0.7362,
       "step": 624
     },
     {
       "epoch": 11.36,
-      "learning_rate": 0.0004728888888888889,
-      "loss": 0.6886,
+      "learning_rate": 0.00048317241379310343,
+      "loss": 0.6911,
       "step": 625
     },
     {
       "epoch": 11.38,
-      "learning_rate": 0.0004726666666666667,
-      "loss": 0.669,
+      "learning_rate": 0.0004830344827586207,
+      "loss": 0.654,
       "step": 626
     },
     {
       "epoch": 11.4,
-      "learning_rate": 0.00047244444444444444,
-      "loss": 0.6686,
+      "learning_rate": 0.0004828965517241379,
+      "loss": 0.6659,
       "step": 627
     },
     {
       "epoch": 11.42,
-      "learning_rate": 0.00047222222222222224,
-      "loss": 0.7577,
+      "learning_rate": 0.0004827586206896552,
+      "loss": 0.7531,
       "step": 628
     },
     {
       "epoch": 11.44,
-      "learning_rate": 0.000472,
-      "loss": 0.7586,
+      "learning_rate": 0.00048262068965517244,
+      "loss": 0.7129,
       "step": 629
     },
     {
       "epoch": 11.45,
-      "learning_rate": 0.00047177777777777783,
-      "loss": 0.7331,
+      "learning_rate": 0.0004824827586206897,
+      "loss": 0.7172,
       "step": 630
     },
     {
       "epoch": 11.47,
-      "learning_rate": 0.0004715555555555556,
-      "loss": 0.6973,
+      "learning_rate": 0.0004823448275862069,
+      "loss": 0.7162,
       "step": 631
     },
     {
       "epoch": 11.49,
-      "learning_rate": 0.0004713333333333333,
-      "loss": 0.6729,
+      "learning_rate": 0.0004822068965517242,
+      "loss": 0.7096,
       "step": 632
     },
     {
       "epoch": 11.51,
-      "learning_rate": 0.0004711111111111111,
-      "loss": 0.6857,
+      "learning_rate": 0.0004820689655172414,
+      "loss": 0.7202,
       "step": 633
     },
     {
       "epoch": 11.53,
-      "learning_rate": 0.0004708888888888889,
-      "loss": 0.6912,
+      "learning_rate": 0.00048193103448275867,
+      "loss": 0.6899,
       "step": 634
     },
     {
       "epoch": 11.55,
-      "learning_rate": 0.0004706666666666667,
-      "loss": 0.7032,
+      "learning_rate": 0.0004817931034482759,
+      "loss": 0.663,
       "step": 635
     },
     {
       "epoch": 11.56,
-      "learning_rate": 0.00047044444444444445,
-      "loss": 0.74,
+      "learning_rate": 0.00048165517241379314,
+      "loss": 0.7462,
       "step": 636
     },
     {
       "epoch": 11.58,
-      "learning_rate": 0.00047022222222222224,
-      "loss": 0.6253,
+      "learning_rate": 0.00048151724137931036,
+      "loss": 0.6865,
       "step": 637
     },
     {
       "epoch": 11.6,
-      "learning_rate": 0.00047,
-      "loss": 0.6703,
+      "learning_rate": 0.00048137931034482757,
+      "loss": 0.6928,
       "step": 638
     },
     {
       "epoch": 11.62,
-      "learning_rate": 0.0004697777777777778,
-      "loss": 0.6352,
+      "learning_rate": 0.00048124137931034483,
+      "loss": 0.6587,
       "step": 639
     },
     {
       "epoch": 11.64,
-      "learning_rate": 0.0004695555555555556,
-      "loss": 0.8026,
+      "learning_rate": 0.00048110344827586205,
+      "loss": 0.7528,
       "step": 640
     },
     {
       "epoch": 11.65,
-      "learning_rate": 0.0004693333333333333,
-      "loss": 0.6589,
+      "learning_rate": 0.0004809655172413793,
+      "loss": 0.6303,
       "step": 641
     },
     {
       "epoch": 11.67,
-      "learning_rate": 0.0004691111111111111,
-      "loss": 0.7327,
+      "learning_rate": 0.0004808275862068965,
+      "loss": 0.7701,
       "step": 642
     },
     {
       "epoch": 11.69,
-      "learning_rate": 0.0004688888888888889,
-      "loss": 0.8019,
+      "learning_rate": 0.0004806896551724138,
+      "loss": 0.7225,
       "step": 643
     },
     {
       "epoch": 11.71,
-      "learning_rate": 0.0004686666666666667,
-      "loss": 0.6998,
+      "learning_rate": 0.000480551724137931,
+      "loss": 0.6614,
       "step": 644
     },
     {
       "epoch": 11.73,
-      "learning_rate": 0.00046844444444444445,
-      "loss": 0.6045,
+      "learning_rate": 0.00048041379310344827,
+      "loss": 0.6298,
       "step": 645
     },
     {
       "epoch": 11.75,
-      "learning_rate": 0.0004682222222222222,
-      "loss": 0.6516,
+      "learning_rate": 0.00048027586206896553,
+      "loss": 0.6412,
       "step": 646
     },
     {
       "epoch": 11.76,
-      "learning_rate": 0.00046800000000000005,
-      "loss": 0.7027,
+      "learning_rate": 0.0004801379310344828,
+      "loss": 0.6971,
       "step": 647
     },
     {
       "epoch": 11.78,
-      "learning_rate": 0.0004677777777777778,
-      "loss": 0.7088,
+      "learning_rate": 0.00048,
+      "loss": 0.7207,
       "step": 648
     },
     {
       "epoch": 11.8,
-      "learning_rate": 0.0004675555555555556,
-      "loss": 0.6559,
+      "learning_rate": 0.0004798620689655173,
+      "loss": 0.6844,
       "step": 649
     },
     {
       "epoch": 11.82,
-      "learning_rate": 0.0004673333333333333,
-      "loss": 0.6327,
+      "learning_rate": 0.0004797241379310345,
+      "loss": 0.6321,
       "step": 650
     },
     {
       "epoch": 11.84,
-      "learning_rate": 0.0004671111111111111,
-      "loss": 0.721,
+      "learning_rate": 0.00047958620689655176,
+      "loss": 0.7481,
       "step": 651
     },
     {
       "epoch": 11.85,
-      "learning_rate": 0.0004668888888888889,
-      "loss": 0.6213,
+      "learning_rate": 0.00047944827586206897,
+      "loss": 0.6611,
       "step": 652
     },
     {
       "epoch": 11.87,
-      "learning_rate": 0.00046666666666666666,
-      "loss": 0.7424,
+      "learning_rate": 0.00047931034482758623,
+      "loss": 0.7427,
       "step": 653
     },
     {
       "epoch": 11.89,
-      "learning_rate": 0.00046644444444444446,
-      "loss": 0.6099,
+      "learning_rate": 0.00047917241379310345,
+      "loss": 0.6243,
       "step": 654
     },
     {
       "epoch": 11.91,
-      "learning_rate": 0.0004662222222222222,
-      "loss": 0.6282,
+      "learning_rate": 0.0004790344827586207,
+      "loss": 0.6658,
       "step": 655
     },
     {
       "epoch": 11.93,
-      "learning_rate": 0.00046600000000000005,
-      "loss": 0.7802,
+      "learning_rate": 0.0004788965517241379,
+      "loss": 0.8307,
       "step": 656
     },
     {
       "epoch": 11.95,
-      "learning_rate": 0.0004657777777777778,
-      "loss": 0.6357,
+      "learning_rate": 0.0004787586206896552,
+      "loss": 0.6315,
       "step": 657
     },
     {
       "epoch": 11.96,
-      "learning_rate": 0.0004655555555555556,
-      "loss": 0.729,
+      "learning_rate": 0.0004786206896551724,
+      "loss": 0.7214,
       "step": 658
     },
     {
       "epoch": 11.98,
-      "learning_rate": 0.00046533333333333333,
-      "loss": 0.7149,
+      "learning_rate": 0.00047848275862068967,
+      "loss": 0.6948,
       "step": 659
     },
     {
       "epoch": 12.0,
-      "learning_rate": 0.0004651111111111111,
-      "loss": 0.8023,
+      "learning_rate": 0.0004783448275862069,
+      "loss": 0.9703,
       "step": 660
     },
     {
       "epoch": 12.02,
-      "learning_rate": 0.0004648888888888889,
-      "loss": 0.6035,
+      "learning_rate": 0.00047820689655172415,
+      "loss": 0.6023,
       "step": 661
     },
     {
       "epoch": 12.04,
-      "learning_rate": 0.00046466666666666667,
-      "loss": 0.6833,
+      "learning_rate": 0.00047806896551724136,
+      "loss": 0.69,
       "step": 662
     },
     {
       "epoch": 12.05,
-      "learning_rate": 0.00046444444444444446,
-      "loss": 0.67,
+      "learning_rate": 0.0004779310344827587,
+      "loss": 0.6857,
       "step": 663
     },
     {
       "epoch": 12.07,
-      "learning_rate": 0.0004642222222222222,
-      "loss": 0.664,
+      "learning_rate": 0.0004777931034482759,
+      "loss": 0.6994,
       "step": 664
     },
     {
       "epoch": 12.09,
-      "learning_rate": 0.00046400000000000006,
-      "loss": 0.6857,
+      "learning_rate": 0.00047765517241379316,
+      "loss": 0.7019,
       "step": 665
     },
     {
       "epoch": 12.11,
-      "learning_rate": 0.0004637777777777778,
-      "loss": 0.6698,
+      "learning_rate": 0.00047751724137931037,
+      "loss": 0.6845,
       "step": 666
     },
     {
       "epoch": 12.13,
-      "learning_rate": 0.00046355555555555554,
-      "loss": 0.7026,
+      "learning_rate": 0.00047737931034482763,
+      "loss": 0.7169,
       "step": 667
     },
     {
       "epoch": 12.15,
-      "learning_rate": 0.00046333333333333334,
-      "loss": 0.6725,
+      "learning_rate": 0.00047724137931034485,
+      "loss": 0.7118,
       "step": 668
     },
     {
       "epoch": 12.16,
-      "learning_rate": 0.0004631111111111111,
-      "loss": 0.6116,
+      "learning_rate": 0.0004771034482758621,
+      "loss": 0.6348,
       "step": 669
     },
     {
       "epoch": 12.18,
-      "learning_rate": 0.00046288888888888893,
-      "loss": 0.6356,
+      "learning_rate": 0.0004769655172413793,
+      "loss": 0.6615,
       "step": 670
     },
     {
       "epoch": 12.2,
-      "learning_rate": 0.0004626666666666667,
-      "loss": 0.6505,
+      "learning_rate": 0.00047682758620689654,
+      "loss": 0.6377,
       "step": 671
     },
     {
       "epoch": 12.22,
-      "learning_rate": 0.00046244444444444447,
-      "loss": 0.624,
+      "learning_rate": 0.0004766896551724138,
+      "loss": 0.6143,
       "step": 672
     },
     {
       "epoch": 12.24,
-      "learning_rate": 0.0004622222222222222,
-      "loss": 0.7153,
+      "learning_rate": 0.000476551724137931,
+      "loss": 0.6777,
       "step": 673
     },
     {
       "epoch": 12.25,
-      "learning_rate": 0.000462,
-      "loss": 0.661,
+      "learning_rate": 0.0004764137931034483,
+      "loss": 0.6743,
       "step": 674
     },
     {
       "epoch": 12.27,
-      "learning_rate": 0.0004617777777777778,
-      "loss": 0.5991,
+      "learning_rate": 0.0004762758620689655,
+      "loss": 0.6016,
       "step": 675
     },
     {
       "epoch": 12.29,
-      "learning_rate": 0.00046155555555555555,
-      "loss": 0.5946,
+      "learning_rate": 0.00047613793103448276,
+      "loss": 0.5865,
       "step": 676
     },
     {
       "epoch": 12.31,
-      "learning_rate": 0.00046133333333333334,
-      "loss": 0.6661,
+      "learning_rate": 0.00047599999999999997,
+      "loss": 0.6779,
       "step": 677
     },
     {
       "epoch": 12.33,
-      "learning_rate": 0.00046111111111111114,
-      "loss": 0.6633,
+      "learning_rate": 0.00047586206896551724,
+      "loss": 0.6526,
       "step": 678
     },
     {
       "epoch": 12.35,
-      "learning_rate": 0.00046088888888888894,
-      "loss": 0.6911,
+      "learning_rate": 0.00047572413793103445,
+      "loss": 0.7,
       "step": 679
     },
     {
       "epoch": 12.36,
-      "learning_rate": 0.0004606666666666667,
-      "loss": 0.7059,
+      "learning_rate": 0.00047558620689655177,
+      "loss": 0.7099,
       "step": 680
     },
     {
       "epoch": 12.38,
-      "learning_rate": 0.0004604444444444444,
-      "loss": 0.6463,
+      "learning_rate": 0.000475448275862069,
+      "loss": 0.653,
       "step": 681
     },
     {
       "epoch": 12.4,
-      "learning_rate": 0.0004602222222222222,
-      "loss": 0.6703,
+      "learning_rate": 0.00047531034482758625,
+      "loss": 0.6358,
       "step": 682
     },
     {
       "epoch": 12.42,
-      "learning_rate": 0.00046,
-      "loss": 0.636,
+      "learning_rate": 0.00047517241379310346,
+      "loss": 0.6181,
       "step": 683
     },
     {
       "epoch": 12.44,
-      "learning_rate": 0.0004597777777777778,
-      "loss": 0.6628,
+      "learning_rate": 0.0004750344827586207,
+      "loss": 0.6407,
       "step": 684
     },
     {
       "epoch": 12.45,
-      "learning_rate": 0.00045955555555555555,
-      "loss": 0.6968,
+      "learning_rate": 0.00047489655172413794,
+      "loss": 0.7019,
       "step": 685
     },
     {
       "epoch": 12.47,
-      "learning_rate": 0.00045933333333333335,
-      "loss": 0.6474,
+      "learning_rate": 0.0004747586206896552,
+      "loss": 0.6495,
       "step": 686
     },
     {
       "epoch": 12.49,
-      "learning_rate": 0.00045911111111111115,
-      "loss": 0.6292,
+      "learning_rate": 0.0004746206896551724,
+      "loss": 0.6642,
       "step": 687
     },
     {
       "epoch": 12.51,
-      "learning_rate": 0.0004588888888888889,
-      "loss": 0.6825,
+      "learning_rate": 0.0004744827586206897,
+      "loss": 0.6705,
       "step": 688
     },
     {
       "epoch": 12.53,
-      "learning_rate": 0.0004586666666666667,
-      "loss": 0.6821,
+      "learning_rate": 0.0004743448275862069,
+      "loss": 0.7008,
       "step": 689
     },
     {
       "epoch": 12.55,
-      "learning_rate": 0.0004584444444444444,
-      "loss": 0.6025,
+      "learning_rate": 0.00047420689655172416,
+      "loss": 0.6046,
       "step": 690
     },
     {
       "epoch": 12.56,
-      "learning_rate": 0.0004582222222222222,
-      "loss": 0.5728,
+      "learning_rate": 0.00047406896551724137,
+      "loss": 0.6035,
       "step": 691
     },
     {
       "epoch": 12.58,
-      "learning_rate": 0.000458,
-      "loss": 0.6316,
+      "learning_rate": 0.00047393103448275863,
+      "loss": 0.6552,
       "step": 692
     },
     {
       "epoch": 12.6,
-      "learning_rate": 0.0004577777777777778,
-      "loss": 0.6035,
+      "learning_rate": 0.00047379310344827585,
+      "loss": 0.6251,
       "step": 693
     },
     {
       "epoch": 12.62,
-      "learning_rate": 0.00045755555555555556,
-      "loss": 0.5604,
+      "learning_rate": 0.0004736551724137931,
+      "loss": 0.6279,
       "step": 694
     },
     {
       "epoch": 12.64,
-      "learning_rate": 0.0004573333333333333,
-      "loss": 0.5972,
+      "learning_rate": 0.0004735172413793103,
+      "loss": 0.6245,
       "step": 695
     },
     {
       "epoch": 12.65,
-      "learning_rate": 0.00045711111111111115,
-      "loss": 0.6067,
+      "learning_rate": 0.00047337931034482764,
+      "loss": 0.6813,
       "step": 696
     },
     {
       "epoch": 12.67,
-      "learning_rate": 0.0004568888888888889,
-      "loss": 0.6844,
+      "learning_rate": 0.00047324137931034486,
+      "loss": 0.6891,
       "step": 697
     },
     {
       "epoch": 12.69,
-      "learning_rate": 0.0004566666666666667,
-      "loss": 0.6154,
+      "learning_rate": 0.0004731034482758621,
+      "loss": 0.6255,
       "step": 698
     },
     {
       "epoch": 12.71,
-      "learning_rate": 0.00045644444444444443,
-      "loss": 0.7001,
+      "learning_rate": 0.00047296551724137933,
+      "loss": 0.6605,
       "step": 699
     },
     {
       "epoch": 12.73,
-      "learning_rate": 0.00045622222222222223,
-      "loss": 0.5963,
+      "learning_rate": 0.0004728275862068966,
+      "loss": 0.6135,
       "step": 700
     },
     {
       "epoch": 12.75,
-      "learning_rate": 0.000456,
-      "loss": 0.6186,
+      "learning_rate": 0.0004726896551724138,
+      "loss": 0.646,
       "step": 701
     },
     {
       "epoch": 12.76,
-      "learning_rate": 0.00045577777777777777,
-      "loss": 0.6799,
+      "learning_rate": 0.0004725517241379311,
+      "loss": 0.693,
       "step": 702
     },
     {
       "epoch": 12.78,
-      "learning_rate": 0.00045555555555555556,
-      "loss": 0.6808,
+      "learning_rate": 0.0004724137931034483,
+      "loss": 0.677,
       "step": 703
     },
     {
       "epoch": 12.8,
-      "learning_rate": 0.0004553333333333333,
-      "loss": 0.6356,
+      "learning_rate": 0.0004722758620689655,
+      "loss": 0.664,
       "step": 704
     },
     {
       "epoch": 12.82,
-      "learning_rate": 0.00045511111111111116,
-      "loss": 0.6298,
+      "learning_rate": 0.00047213793103448277,
+      "loss": 0.649,
       "step": 705
     },
     {
       "epoch": 12.84,
-      "learning_rate": 0.0004548888888888889,
-      "loss": 0.5776,
+      "learning_rate": 0.000472,
+      "loss": 0.6367,
       "step": 706
     },
     {
       "epoch": 12.85,
-      "learning_rate": 0.0004546666666666667,
-      "loss": 0.5919,
+      "learning_rate": 0.00047186206896551725,
+      "loss": 0.5992,
       "step": 707
     },
     {
       "epoch": 12.87,
-      "learning_rate": 0.00045444444444444444,
-      "loss": 0.642,
+      "learning_rate": 0.00047172413793103446,
+      "loss": 0.6556,
       "step": 708
     },
     {
       "epoch": 12.89,
-      "learning_rate": 0.00045422222222222223,
-      "loss": 0.6844,
+      "learning_rate": 0.0004715862068965517,
+      "loss": 0.7585,
       "step": 709
     },
     {
       "epoch": 12.91,
-      "learning_rate": 0.00045400000000000003,
-      "loss": 0.6694,
+      "learning_rate": 0.00047144827586206894,
+      "loss": 0.7504,
       "step": 710
     },
     {
       "epoch": 12.93,
-      "learning_rate": 0.0004537777777777778,
-      "loss": 0.6476,
+      "learning_rate": 0.0004713103448275862,
+      "loss": 0.7082,
       "step": 711
     },
     {
       "epoch": 12.95,
-      "learning_rate": 0.00045355555555555557,
-      "loss": 0.6265,
+      "learning_rate": 0.0004711724137931034,
+      "loss": 0.6501,
       "step": 712
     },
     {
       "epoch": 12.96,
-      "learning_rate": 0.0004533333333333333,
-      "loss": 0.6331,
+      "learning_rate": 0.00047103448275862073,
+      "loss": 0.6471,
       "step": 713
     },
     {
       "epoch": 12.98,
-      "learning_rate": 0.00045311111111111116,
-      "loss": 0.6545,
+      "learning_rate": 0.00047089655172413795,
+      "loss": 0.6969,
       "step": 714
     },
     {
       "epoch": 13.0,
-      "learning_rate": 0.0004528888888888889,
-      "loss": 0.6465,
+      "learning_rate": 0.0004707586206896552,
+      "loss": 0.6008,
       "step": 715
     },
     {
       "epoch": 13.02,
-      "learning_rate": 0.00045266666666666665,
-      "loss": 0.6095,
+      "learning_rate": 0.0004706206896551724,
+      "loss": 0.6735,
       "step": 716
     },
     {
       "epoch": 13.04,
-      "learning_rate": 0.00045244444444444444,
-      "loss": 0.6206,
+      "learning_rate": 0.0004704827586206897,
+      "loss": 0.6573,
       "step": 717
     },
     {
       "epoch": 13.05,
-      "learning_rate": 0.00045222222222222224,
+      "learning_rate": 0.0004703448275862069,
       "loss": 0.621,
       "step": 718
     },
     {
       "epoch": 13.07,
-      "learning_rate": 0.00045200000000000004,
-      "loss": 0.5383,
+      "learning_rate": 0.00047020689655172417,
+      "loss": 0.5912,
       "step": 719
     },
     {
       "epoch": 13.09,
-      "learning_rate": 0.0004517777777777778,
-      "loss": 0.649,
+      "learning_rate": 0.0004700689655172414,
+      "loss": 0.6931,
       "step": 720
     },
     {
       "epoch": 13.11,
-      "learning_rate": 0.0004515555555555556,
-      "loss": 0.6406,
+      "learning_rate": 0.00046993103448275865,
+      "loss": 0.6792,
       "step": 721
     },
     {
       "epoch": 13.13,
-      "learning_rate": 0.0004513333333333333,
-      "loss": 0.5254,
+      "learning_rate": 0.00046979310344827586,
+      "loss": 0.5878,
       "step": 722
     },
     {
       "epoch": 13.15,
-      "learning_rate": 0.0004511111111111111,
-      "loss": 0.5951,
+      "learning_rate": 0.0004696551724137931,
+      "loss": 0.5868,
       "step": 723
     },
     {
       "epoch": 13.16,
-      "learning_rate": 0.0004508888888888889,
-      "loss": 0.5065,
+      "learning_rate": 0.00046951724137931034,
+      "loss": 0.5467,
       "step": 724
     },
     {
       "epoch": 13.18,
-      "learning_rate": 0.00045066666666666665,
-      "loss": 0.5604,
+      "learning_rate": 0.0004693793103448276,
+      "loss": 0.5911,
       "step": 725
     },
     {
       "epoch": 13.2,
-      "learning_rate": 0.00045044444444444445,
-      "loss": 0.5576,
+      "learning_rate": 0.0004692413793103448,
+      "loss": 0.6288,
       "step": 726
     },
     {
       "epoch": 13.22,
-      "learning_rate": 0.00045022222222222225,
-      "loss": 0.692,
+      "learning_rate": 0.0004691034482758621,
+      "loss": 0.7275,
       "step": 727
     },
     {
       "epoch": 13.24,
-      "learning_rate": 0.00045000000000000004,
-      "loss": 0.557,
+      "learning_rate": 0.0004689655172413793,
+      "loss": 0.5899,
       "step": 728
     },
     {
       "epoch": 13.25,
-      "learning_rate": 0.0004497777777777778,
-      "loss": 0.6161,
+      "learning_rate": 0.00046882758620689656,
+      "loss": 0.6233,
       "step": 729
     },
     {
       "epoch": 13.27,
-      "learning_rate": 0.0004495555555555555,
-      "loss": 0.5495,
+      "learning_rate": 0.0004686896551724138,
+      "loss": 0.6017,
       "step": 730
     },
     {
       "epoch": 13.29,
-      "learning_rate": 0.0004493333333333333,
-      "loss": 0.6714,
+      "learning_rate": 0.0004685517241379311,
+      "loss": 0.6839,
       "step": 731
     },
     {
       "epoch": 13.31,
-      "learning_rate": 0.0004491111111111111,
-      "loss": 0.6002,
+      "learning_rate": 0.0004684137931034483,
+      "loss": 0.5871,
       "step": 732
     },
     {
       "epoch": 13.33,
-      "learning_rate": 0.0004488888888888889,
-      "loss": 0.5709,
+      "learning_rate": 0.00046827586206896557,
+      "loss": 0.6135,
       "step": 733
     },
     {
       "epoch": 13.35,
-      "learning_rate": 0.00044866666666666666,
-      "loss": 0.4946,
+      "learning_rate": 0.0004681379310344828,
+      "loss": 0.5171,
       "step": 734
     },
     {
       "epoch": 13.36,
-      "learning_rate": 0.00044844444444444445,
-      "loss": 0.5833,
+      "learning_rate": 0.00046800000000000005,
+      "loss": 0.6092,
       "step": 735
     },
     {
       "epoch": 13.38,
-      "learning_rate": 0.00044822222222222225,
-      "loss": 0.6129,
+      "learning_rate": 0.00046786206896551726,
+      "loss": 0.6416,
       "step": 736
     },
     {
       "epoch": 13.4,
-      "learning_rate": 0.000448,
-      "loss": 0.6539,
+      "learning_rate": 0.00046772413793103447,
+      "loss": 0.7042,
       "step": 737
     },
     {
       "epoch": 13.42,
-      "learning_rate": 0.0004477777777777778,
-      "loss": 0.5902,
+      "learning_rate": 0.00046758620689655174,
+      "loss": 0.6731,
       "step": 738
     },
     {
       "epoch": 13.44,
-      "learning_rate": 0.00044755555555555553,
-      "loss": 0.5748,
+      "learning_rate": 0.00046744827586206895,
+      "loss": 0.5918,
       "step": 739
     },
     {
       "epoch": 13.45,
-      "learning_rate": 0.0004473333333333334,
-      "loss": 0.6714,
+      "learning_rate": 0.0004673103448275862,
+      "loss": 0.6271,
       "step": 740
     },
     {
       "epoch": 13.47,
-      "learning_rate": 0.0004471111111111111,
-      "loss": 0.6687,
+      "learning_rate": 0.0004671724137931034,
+      "loss": 0.7144,
       "step": 741
     },
     {
       "epoch": 13.49,
-      "learning_rate": 0.0004468888888888889,
-      "loss": 0.572,
+      "learning_rate": 0.0004670344827586207,
+      "loss": 0.6077,
       "step": 742
     },
     {
       "epoch": 13.51,
-      "learning_rate": 0.00044666666666666666,
-      "loss": 0.5925,
+      "learning_rate": 0.0004668965517241379,
+      "loss": 0.6652,
       "step": 743
     },
     {
       "epoch": 13.53,
-      "learning_rate": 0.0004464444444444444,
-      "loss": 0.638,
+      "learning_rate": 0.00046675862068965517,
+      "loss": 0.672,
       "step": 744
     },
     {
       "epoch": 13.55,
-      "learning_rate": 0.00044622222222222226,
-      "loss": 0.6534,
+      "learning_rate": 0.0004666206896551724,
+      "loss": 0.6835,
       "step": 745
     },
     {
       "epoch": 13.56,
-      "learning_rate": 0.000446,
-      "loss": 0.627,
+      "learning_rate": 0.00046648275862068965,
+      "loss": 0.6764,
       "step": 746
     },
     {
       "epoch": 13.58,
-      "learning_rate": 0.0004457777777777778,
-      "loss": 0.6565,
+      "learning_rate": 0.0004663448275862069,
+      "loss": 0.7506,
       "step": 747
     },
     {
       "epoch": 13.6,
-      "learning_rate": 0.00044555555555555554,
-      "loss": 0.6471,
+      "learning_rate": 0.0004662068965517242,
+      "loss": 0.718,
       "step": 748
     },
     {
       "epoch": 13.62,
-      "learning_rate": 0.0004453333333333334,
-      "loss": 0.6168,
+      "learning_rate": 0.0004660689655172414,
+      "loss": 0.6546,
       "step": 749
     },
     {
       "epoch": 13.64,
-      "learning_rate": 0.00044511111111111113,
-      "loss": 0.5954,
+      "learning_rate": 0.00046593103448275866,
+      "loss": 0.6182,
       "step": 750
     },
     {
       "epoch": 13.65,
-      "learning_rate": 0.00044488888888888887,
-      "loss": 0.6284,
+      "learning_rate": 0.00046579310344827587,
+      "loss": 0.588,
       "step": 751
     },
     {
       "epoch": 13.67,
-      "learning_rate": 0.00044466666666666667,
-      "loss": 0.611,
+      "learning_rate": 0.00046565517241379314,
+      "loss": 0.6933,
       "step": 752
     },
     {
       "epoch": 13.69,
-      "learning_rate": 0.0004444444444444444,
-      "loss": 0.6371,
+      "learning_rate": 0.00046551724137931035,
+      "loss": 0.66,
       "step": 753
     },
     {
       "epoch": 13.71,
-      "learning_rate": 0.00044422222222222226,
-      "loss": 0.6462,
+      "learning_rate": 0.0004653793103448276,
+      "loss": 0.6533,
       "step": 754
     },
     {
       "epoch": 13.73,
-      "learning_rate": 0.000444,
-      "loss": 0.6047,
+      "learning_rate": 0.0004652413793103448,
+      "loss": 0.6253,
       "step": 755
     },
     {
       "epoch": 13.75,
-      "learning_rate": 0.0004437777777777778,
-      "loss": 0.5854,
+      "learning_rate": 0.0004651034482758621,
+      "loss": 0.5984,
       "step": 756
     },
     {
       "epoch": 13.76,
-      "learning_rate": 0.00044355555555555554,
-      "loss": 0.6698,
+      "learning_rate": 0.0004649655172413793,
+      "loss": 0.6198,
       "step": 757
     },
     {
       "epoch": 13.78,
-      "learning_rate": 0.00044333333333333334,
-      "loss": 0.6904,
+      "learning_rate": 0.00046482758620689657,
+      "loss": 0.702,
       "step": 758
     },
     {
       "epoch": 13.8,
-      "learning_rate": 0.00044311111111111114,
-      "loss": 0.6389,
+      "learning_rate": 0.0004646896551724138,
+      "loss": 0.6844,
       "step": 759
     },
     {
       "epoch": 13.82,
-      "learning_rate": 0.0004428888888888889,
-      "loss": 0.6498,
+      "learning_rate": 0.00046455172413793105,
+      "loss": 0.6591,
       "step": 760
     },
     {
       "epoch": 13.84,
-      "learning_rate": 0.0004426666666666667,
-      "loss": 0.5574,
+      "learning_rate": 0.00046441379310344826,
+      "loss": 0.5993,
       "step": 761
     },
     {
       "epoch": 13.85,
-      "learning_rate": 0.0004424444444444444,
-      "loss": 0.6191,
+      "learning_rate": 0.0004642758620689655,
+      "loss": 0.6624,
       "step": 762
     },
     {
       "epoch": 13.87,
-      "learning_rate": 0.00044222222222222227,
-      "loss": 0.6843,
+      "learning_rate": 0.00046413793103448274,
+      "loss": 0.7294,
       "step": 763
     },
     {
       "epoch": 13.89,
-      "learning_rate": 0.000442,
-      "loss": 0.6415,
+      "learning_rate": 0.00046400000000000006,
+      "loss": 0.6712,
       "step": 764
     },
     {
       "epoch": 13.91,
-      "learning_rate": 0.00044177777777777775,
-      "loss": 0.6928,
+      "learning_rate": 0.00046386206896551727,
+      "loss": 0.7342,
       "step": 765
     },
     {
       "epoch": 13.93,
-      "learning_rate": 0.00044155555555555555,
-      "loss": 0.6799,
+      "learning_rate": 0.00046372413793103453,
+      "loss": 0.701,
       "step": 766
     },
     {
       "epoch": 13.95,
-      "learning_rate": 0.00044133333333333335,
-      "loss": 0.598,
+      "learning_rate": 0.00046358620689655175,
+      "loss": 0.5794,
       "step": 767
     },
     {
       "epoch": 13.96,
-      "learning_rate": 0.00044111111111111114,
-      "loss": 0.6123,
+      "learning_rate": 0.000463448275862069,
+      "loss": 0.6221,
       "step": 768
     },
     {
       "epoch": 13.98,
-      "learning_rate": 0.0004408888888888889,
-      "loss": 0.652,
+      "learning_rate": 0.0004633103448275862,
+      "loss": 0.6421,
       "step": 769
     },
     {
       "epoch": 14.0,
-      "learning_rate": 0.0004406666666666667,
-      "loss": 0.6272,
+      "learning_rate": 0.00046317241379310344,
+      "loss": 0.5978,
       "step": 770
     },
     {
       "epoch": 14.02,
-      "learning_rate": 0.0004404444444444445,
-      "loss": 0.6057,
+      "learning_rate": 0.0004630344827586207,
+      "loss": 0.5918,
       "step": 771
     },
     {
       "epoch": 14.04,
-      "learning_rate": 0.0004402222222222222,
-      "loss": 0.5851,
+      "learning_rate": 0.0004628965517241379,
+      "loss": 0.6381,
       "step": 772
     },
     {
       "epoch": 14.05,
-      "learning_rate": 0.00044,
-      "loss": 0.5667,
+      "learning_rate": 0.0004627586206896552,
+      "loss": 0.5944,
       "step": 773
     },
     {
       "epoch": 14.07,
-      "learning_rate": 0.00043977777777777776,
-      "loss": 0.6351,
+      "learning_rate": 0.0004626206896551724,
+      "loss": 0.6683,
       "step": 774
     },
     {
       "epoch": 14.09,
-      "learning_rate": 0.00043955555555555555,
-      "loss": 0.5428,
+      "learning_rate": 0.00046248275862068966,
+      "loss": 0.5465,
       "step": 775
     },
     {
       "epoch": 14.11,
-      "learning_rate": 0.00043933333333333335,
-      "loss": 0.5702,
+      "learning_rate": 0.00046234482758620687,
+      "loss": 0.5925,
       "step": 776
     },
     {
       "epoch": 14.13,
-      "learning_rate": 0.00043911111111111115,
-      "loss": 0.5919,
+      "learning_rate": 0.00046220689655172414,
+      "loss": 0.6196,
       "step": 777
     },
     {
       "epoch": 14.15,
-      "learning_rate": 0.0004388888888888889,
-      "loss": 0.5973,
+      "learning_rate": 0.00046206896551724135,
+      "loss": 0.6231,
       "step": 778
     },
     {
       "epoch": 14.16,
-      "learning_rate": 0.00043866666666666663,
-      "loss": 0.6029,
+      "learning_rate": 0.0004619310344827586,
+      "loss": 0.6169,
       "step": 779
     },
     {
       "epoch": 14.18,
-      "learning_rate": 0.0004384444444444445,
-      "loss": 0.6025,
+      "learning_rate": 0.0004617931034482758,
+      "loss": 0.6051,
       "step": 780
     },
     {
       "epoch": 14.2,
-      "learning_rate": 0.0004382222222222222,
-      "loss": 0.4788,
+      "learning_rate": 0.00046165517241379315,
+      "loss": 0.4676,
       "step": 781
     },
     {
       "epoch": 14.22,
-      "learning_rate": 0.000438,
-      "loss": 0.6229,
+      "learning_rate": 0.00046151724137931036,
+      "loss": 0.6829,
       "step": 782
     },
     {
       "epoch": 14.24,
-      "learning_rate": 0.00043777777777777776,
-      "loss": 0.5742,
+      "learning_rate": 0.0004613793103448276,
+      "loss": 0.628,
       "step": 783
     },
     {
       "epoch": 14.25,
-      "learning_rate": 0.00043755555555555556,
-      "loss": 0.6463,
+      "learning_rate": 0.00046124137931034484,
+      "loss": 0.6666,
       "step": 784
     },
     {
       "epoch": 14.27,
-      "learning_rate": 0.00043733333333333336,
-      "loss": 0.6106,
+      "learning_rate": 0.0004611034482758621,
+      "loss": 0.6691,
       "step": 785
     },
     {
       "epoch": 14.29,
-      "learning_rate": 0.00043711111111111115,
-      "loss": 0.5753,
+      "learning_rate": 0.0004609655172413793,
+      "loss": 0.6434,
       "step": 786
     },
     {
       "epoch": 14.31,
-      "learning_rate": 0.0004368888888888889,
-      "loss": 0.6625,
+      "learning_rate": 0.0004608275862068966,
+      "loss": 0.6765,
       "step": 787
     },
     {
       "epoch": 14.33,
-      "learning_rate": 0.00043666666666666664,
-      "loss": 0.5985,
+      "learning_rate": 0.0004606896551724138,
+      "loss": 0.6261,
       "step": 788
     },
     {
       "epoch": 14.35,
-      "learning_rate": 0.0004364444444444445,
-      "loss": 0.5407,
+      "learning_rate": 0.00046055172413793106,
+      "loss": 0.5736,
       "step": 789
     },
     {
       "epoch": 14.36,
-      "learning_rate": 0.00043622222222222223,
-      "loss": 0.5693,
+      "learning_rate": 0.00046041379310344827,
+      "loss": 0.6094,
       "step": 790
     },
     {
       "epoch": 14.38,
-      "learning_rate": 0.000436,
-      "loss": 0.5174,
+      "learning_rate": 0.00046027586206896554,
+      "loss": 0.5199,
       "step": 791
     },
     {
       "epoch": 14.4,
-      "learning_rate": 0.00043577777777777777,
-      "loss": 0.6024,
+      "learning_rate": 0.00046013793103448275,
+      "loss": 0.6319,
       "step": 792
     },
     {
       "epoch": 14.42,
-      "learning_rate": 0.0004355555555555555,
-      "loss": 0.5694,
+      "learning_rate": 0.00046,
+      "loss": 0.6125,
       "step": 793
     },
     {
       "epoch": 14.44,
-      "learning_rate": 0.00043533333333333336,
-      "loss": 0.6071,
+      "learning_rate": 0.0004598620689655172,
+      "loss": 0.6381,
       "step": 794
     },
     {
       "epoch": 14.45,
-      "learning_rate": 0.0004351111111111111,
-      "loss": 0.6489,
+      "learning_rate": 0.0004597241379310345,
+      "loss": 0.6826,
       "step": 795
     },
     {
       "epoch": 14.47,
-      "learning_rate": 0.0004348888888888889,
-      "loss": 0.6062,
+      "learning_rate": 0.0004595862068965517,
+      "loss": 0.6353,
       "step": 796
     },
     {
       "epoch": 14.49,
-      "learning_rate": 0.00043466666666666664,
-      "loss": 0.5583,
+      "learning_rate": 0.000459448275862069,
+      "loss": 0.6447,
       "step": 797
     },
     {
       "epoch": 14.51,
-      "learning_rate": 0.0004344444444444445,
-      "loss": 0.5436,
+      "learning_rate": 0.00045931034482758624,
+      "loss": 0.5607,
       "step": 798
     },
     {
       "epoch": 14.53,
-      "learning_rate": 0.00043422222222222224,
-      "loss": 0.6021,
+      "learning_rate": 0.0004591724137931035,
+      "loss": 0.6413,
       "step": 799
     },
     {
       "epoch": 14.55,
-      "learning_rate": 0.00043400000000000003,
-      "loss": 0.684,
+      "learning_rate": 0.0004590344827586207,
+      "loss": 0.7449,
       "step": 800
     },
     {
       "epoch": 14.56,
-      "learning_rate": 0.0004337777777777778,
-      "loss": 0.5714,
+      "learning_rate": 0.000458896551724138,
+      "loss": 0.6148,
       "step": 801
     },
     {
       "epoch": 14.58,
-      "learning_rate": 0.00043355555555555557,
-      "loss": 0.6462,
+      "learning_rate": 0.0004587586206896552,
+      "loss": 0.6804,
       "step": 802
     },
     {
       "epoch": 14.6,
-      "learning_rate": 0.00043333333333333337,
-      "loss": 0.5609,
+      "learning_rate": 0.0004586206896551724,
+      "loss": 0.529,
       "step": 803
     },
     {
       "epoch": 14.62,
-      "learning_rate": 0.0004331111111111111,
-      "loss": 0.5304,
+      "learning_rate": 0.00045848275862068967,
+      "loss": 0.5408,
       "step": 804
     },
     {
       "epoch": 14.64,
-      "learning_rate": 0.0004328888888888889,
-      "loss": 0.592,
+      "learning_rate": 0.0004583448275862069,
+      "loss": 0.6076,
       "step": 805
     },
     {
       "epoch": 14.65,
-      "learning_rate": 0.00043266666666666665,
-      "loss": 0.6386,
+      "learning_rate": 0.00045820689655172415,
+      "loss": 0.6229,
       "step": 806
     },
     {
       "epoch": 14.67,
-      "learning_rate": 0.0004324444444444445,
-      "loss": 0.6101,
+      "learning_rate": 0.00045806896551724136,
+      "loss": 0.5519,
       "step": 807
     },
     {
       "epoch": 14.69,
-      "learning_rate": 0.00043222222222222224,
-      "loss": 0.6201,
+      "learning_rate": 0.0004579310344827586,
+      "loss": 0.6327,
       "step": 808
     },
     {
       "epoch": 14.71,
-      "learning_rate": 0.000432,
-      "loss": 0.67,
+      "learning_rate": 0.00045779310344827584,
+      "loss": 0.6735,
       "step": 809
     },
     {
       "epoch": 14.73,
-      "learning_rate": 0.0004317777777777778,
-      "loss": 0.5052,
+      "learning_rate": 0.0004576551724137931,
+      "loss": 0.4891,
       "step": 810
     },
     {
       "epoch": 14.75,
-      "learning_rate": 0.0004315555555555556,
-      "loss": 0.6369,
+      "learning_rate": 0.0004575172413793103,
+      "loss": 0.6443,
       "step": 811
     },
     {
       "epoch": 14.76,
-      "learning_rate": 0.0004313333333333334,
-      "loss": 0.5996,
+      "learning_rate": 0.0004573793103448276,
+      "loss": 0.633,
       "step": 812
     },
     {
       "epoch": 14.78,
-      "learning_rate": 0.0004311111111111111,
-      "loss": 0.6324,
+      "learning_rate": 0.0004572413793103448,
+      "loss": 0.6362,
       "step": 813
     },
     {
       "epoch": 14.8,
-      "learning_rate": 0.0004308888888888889,
-      "loss": 0.5701,
+      "learning_rate": 0.0004571034482758621,
+      "loss": 0.5718,
       "step": 814
     },
     {
       "epoch": 14.82,
-      "learning_rate": 0.00043066666666666665,
-      "loss": 0.5735,
+      "learning_rate": 0.0004569655172413793,
+      "loss": 0.5672,
       "step": 815
     },
     {
       "epoch": 14.84,
-      "learning_rate": 0.00043044444444444445,
-      "loss": 0.6025,
+      "learning_rate": 0.0004568275862068966,
+      "loss": 0.6087,
       "step": 816
     },
     {
       "epoch": 14.85,
-      "learning_rate": 0.00043022222222222225,
-      "loss": 0.568,
+      "learning_rate": 0.0004566896551724138,
+      "loss": 0.5933,
       "step": 817
     },
     {
       "epoch": 14.87,
-      "learning_rate": 0.00043,
-      "loss": 0.6047,
+      "learning_rate": 0.00045655172413793107,
+      "loss": 0.6285,
       "step": 818
     },
     {
       "epoch": 14.89,
-      "learning_rate": 0.0004297777777777778,
-      "loss": 0.6053,
+      "learning_rate": 0.0004564137931034483,
+      "loss": 0.5913,
       "step": 819
     },
     {
       "epoch": 14.91,
-      "learning_rate": 0.0004295555555555556,
-      "loss": 0.5319,
+      "learning_rate": 0.00045627586206896555,
+      "loss": 0.5557,
       "step": 820
     },
     {
       "epoch": 14.93,
-      "learning_rate": 0.0004293333333333334,
-      "loss": 0.6231,
+      "learning_rate": 0.00045613793103448276,
+      "loss": 0.5956,
       "step": 821
     },
     {
       "epoch": 14.95,
-      "learning_rate": 0.0004291111111111111,
-      "loss": 0.5501,
+      "learning_rate": 0.000456,
+      "loss": 0.5676,
       "step": 822
     },
     {
       "epoch": 14.96,
-      "learning_rate": 0.00042888888888888886,
-      "loss": 0.5773,
+      "learning_rate": 0.00045586206896551724,
+      "loss": 0.595,
       "step": 823
     },
     {
       "epoch": 14.98,
-      "learning_rate": 0.00042866666666666666,
-      "loss": 0.6606,
+      "learning_rate": 0.0004557241379310345,
+      "loss": 0.7178,
       "step": 824
     },
     {
       "epoch": 15.0,
-      "learning_rate": 0.00042844444444444446,
-      "loss": 0.7816,
+      "learning_rate": 0.0004555862068965517,
+      "loss": 0.7597,
       "step": 825
     },
     {
       "epoch": 15.02,
-      "learning_rate": 0.00042822222222222225,
-      "loss": 0.56,
+      "learning_rate": 0.000455448275862069,
+      "loss": 0.6407,
       "step": 826
     },
     {
       "epoch": 15.04,
-      "learning_rate": 0.000428,
-      "loss": 0.5568,
+      "learning_rate": 0.0004553103448275862,
+      "loss": 0.6098,
       "step": 827
     },
     {
       "epoch": 15.05,
-      "learning_rate": 0.0004277777777777778,
-      "loss": 0.6125,
+      "learning_rate": 0.00045517241379310346,
+      "loss": 0.6375,
       "step": 828
     },
     {
       "epoch": 15.07,
-      "learning_rate": 0.0004275555555555556,
-      "loss": 0.5656,
+      "learning_rate": 0.00045503448275862067,
+      "loss": 0.5365,
       "step": 829
     },
     {
       "epoch": 15.09,
-      "learning_rate": 0.00042733333333333333,
-      "loss": 0.57,
+      "learning_rate": 0.00045489655172413794,
+      "loss": 0.6089,
       "step": 830
     },
     {
       "epoch": 15.11,
-      "learning_rate": 0.0004271111111111111,
-      "loss": 0.5728,
+      "learning_rate": 0.0004547586206896552,
+      "loss": 0.6511,
       "step": 831
     },
     {
       "epoch": 15.13,
-      "learning_rate": 0.00042688888888888887,
-      "loss": 0.5561,
+      "learning_rate": 0.00045462068965517247,
+      "loss": 0.5907,
       "step": 832
     },
     {
       "epoch": 15.15,
-      "learning_rate": 0.0004266666666666667,
-      "loss": 0.543,
+      "learning_rate": 0.0004544827586206897,
+      "loss": 0.5691,
       "step": 833
     },
     {
       "epoch": 15.16,
-      "learning_rate": 0.00042644444444444446,
-      "loss": 0.4992,
+      "learning_rate": 0.00045434482758620695,
+      "loss": 0.5476,
       "step": 834
     },
     {
       "epoch": 15.18,
-      "learning_rate": 0.00042622222222222226,
-      "loss": 0.5099,
+      "learning_rate": 0.00045420689655172416,
+      "loss": 0.5421,
       "step": 835
     },
     {
       "epoch": 15.2,
-      "learning_rate": 0.000426,
-      "loss": 0.5632,
+      "learning_rate": 0.00045406896551724137,
+      "loss": 0.5773,
       "step": 836
     },
     {
       "epoch": 15.22,
-      "learning_rate": 0.00042577777777777774,
-      "loss": 0.5319,
+      "learning_rate": 0.00045393103448275864,
+      "loss": 0.5617,
       "step": 837
     },
     {
       "epoch": 15.24,
-      "learning_rate": 0.0004255555555555556,
-      "loss": 0.5443,
+      "learning_rate": 0.00045379310344827585,
+      "loss": 0.5879,
       "step": 838
     },
     {
       "epoch": 15.25,
-      "learning_rate": 0.00042533333333333334,
-      "loss": 0.5557,
+      "learning_rate": 0.0004536551724137931,
+      "loss": 0.6017,
       "step": 839
     },
     {
       "epoch": 15.27,
-      "learning_rate": 0.00042511111111111113,
-      "loss": 0.584,
+      "learning_rate": 0.0004535172413793103,
+      "loss": 0.5918,
       "step": 840
     },
     {
       "epoch": 15.29,
-      "learning_rate": 0.0004248888888888889,
-      "loss": 0.6011,
+      "learning_rate": 0.0004533793103448276,
+      "loss": 0.6308,
       "step": 841
     },
     {
       "epoch": 15.31,
-      "learning_rate": 0.0004246666666666667,
-      "loss": 0.5167,
+      "learning_rate": 0.0004532413793103448,
+      "loss": 0.5687,
       "step": 842
     },
     {
       "epoch": 15.33,
-      "learning_rate": 0.00042444444444444447,
-      "loss": 0.5975,
+      "learning_rate": 0.00045310344827586207,
+      "loss": 0.6096,
       "step": 843
     },
     {
       "epoch": 15.35,
-      "learning_rate": 0.0004242222222222222,
-      "loss": 0.4988,
+      "learning_rate": 0.0004529655172413793,
+      "loss": 0.5367,
       "step": 844
     },
     {
       "epoch": 15.36,
-      "learning_rate": 0.000424,
-      "loss": 0.5726,
+      "learning_rate": 0.00045282758620689655,
+      "loss": 0.6234,
       "step": 845
     },
     {
       "epoch": 15.38,
-      "learning_rate": 0.00042377777777777775,
-      "loss": 0.4889,
+      "learning_rate": 0.00045268965517241376,
+      "loss": 0.4986,
       "step": 846
     },
     {
       "epoch": 15.4,
-      "learning_rate": 0.0004235555555555556,
-      "loss": 0.5144,
+      "learning_rate": 0.0004525517241379311,
+      "loss": 0.5207,
       "step": 847
     },
     {
       "epoch": 15.42,
-      "learning_rate": 0.00042333333333333334,
-      "loss": 0.548,
+      "learning_rate": 0.0004524137931034483,
+      "loss": 0.5578,
       "step": 848
     },
     {
       "epoch": 15.44,
-      "learning_rate": 0.00042311111111111114,
-      "loss": 0.5654,
+      "learning_rate": 0.00045227586206896556,
+      "loss": 0.6055,
       "step": 849
     },
     {
       "epoch": 15.45,
-      "learning_rate": 0.0004228888888888889,
-      "loss": 0.6105,
+      "learning_rate": 0.00045213793103448277,
+      "loss": 0.6478,
       "step": 850
     },
     {
       "epoch": 15.47,
-      "learning_rate": 0.0004226666666666667,
-      "loss": 0.5706,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 0.6282,
       "step": 851
     },
     {
       "epoch": 15.49,
-      "learning_rate": 0.0004224444444444445,
-      "loss": 0.5437,
+      "learning_rate": 0.00045186206896551725,
+      "loss": 0.5815,
       "step": 852
     },
     {
       "epoch": 15.51,
-      "learning_rate": 0.0004222222222222222,
-      "loss": 0.5863,
+      "learning_rate": 0.0004517241379310345,
+      "loss": 0.5706,
       "step": 853
     },
     {
       "epoch": 15.53,
-      "learning_rate": 0.000422,
-      "loss": 0.5807,
+      "learning_rate": 0.0004515862068965517,
+      "loss": 0.6395,
       "step": 854
     },
     {
       "epoch": 15.55,
-      "learning_rate": 0.00042177777777777775,
-      "loss": 0.4987,
+      "learning_rate": 0.000451448275862069,
+      "loss": 0.5337,
       "step": 855
     },
     {
       "epoch": 15.56,
-      "learning_rate": 0.0004215555555555556,
-      "loss": 0.6011,
+      "learning_rate": 0.0004513103448275862,
+      "loss": 0.5714,
       "step": 856
     },
     {
       "epoch": 15.58,
-      "learning_rate": 0.00042133333333333335,
-      "loss": 0.591,
+      "learning_rate": 0.00045117241379310347,
+      "loss": 0.6334,
       "step": 857
     },
     {
       "epoch": 15.6,
-      "learning_rate": 0.0004211111111111111,
-      "loss": 0.5857,
+      "learning_rate": 0.0004510344827586207,
+      "loss": 0.5975,
       "step": 858
     },
     {
       "epoch": 15.62,
-      "learning_rate": 0.0004208888888888889,
-      "loss": 0.5069,
+      "learning_rate": 0.00045089655172413795,
+      "loss": 0.5459,
       "step": 859
     },
     {
       "epoch": 15.64,
-      "learning_rate": 0.0004206666666666667,
-      "loss": 0.5889,
+      "learning_rate": 0.00045075862068965516,
+      "loss": 0.5802,
       "step": 860
     },
     {
       "epoch": 15.65,
-      "learning_rate": 0.0004204444444444445,
-      "loss": 0.5987,
+      "learning_rate": 0.0004506206896551724,
+      "loss": 0.621,
       "step": 861
     },
     {
       "epoch": 15.67,
-      "learning_rate": 0.0004202222222222222,
-      "loss": 0.7092,
+      "learning_rate": 0.00045048275862068964,
+      "loss": 0.6998,
       "step": 862
     },
     {
       "epoch": 15.69,
-      "learning_rate": 0.00042,
-      "loss": 0.6158,
+      "learning_rate": 0.0004503448275862069,
+      "loss": 0.6369,
       "step": 863
     },
     {
       "epoch": 15.71,
-      "learning_rate": 0.0004197777777777778,
-      "loss": 0.5925,
+      "learning_rate": 0.00045020689655172417,
+      "loss": 0.6038,
       "step": 864
     },
     {
       "epoch": 15.73,
-      "learning_rate": 0.00041955555555555556,
-      "loss": 0.5392,
+      "learning_rate": 0.00045006896551724144,
+      "loss": 0.5643,
       "step": 865
     },
     {
       "epoch": 15.75,
-      "learning_rate": 0.00041933333333333335,
-      "loss": 0.555,
+      "learning_rate": 0.00044993103448275865,
+      "loss": 0.5496,
       "step": 866
     },
     {
       "epoch": 15.76,
-      "learning_rate": 0.0004191111111111111,
-      "loss": 0.5727,
+      "learning_rate": 0.0004497931034482759,
+      "loss": 0.5554,
       "step": 867
     },
     {
       "epoch": 15.78,
-      "learning_rate": 0.0004188888888888889,
-      "loss": 0.5844,
+      "learning_rate": 0.0004496551724137931,
+      "loss": 0.5895,
       "step": 868
     },
     {
       "epoch": 15.8,
-      "learning_rate": 0.0004186666666666667,
-      "loss": 0.6179,
+      "learning_rate": 0.00044951724137931034,
+      "loss": 0.7032,
       "step": 869
     },
     {
       "epoch": 15.82,
-      "learning_rate": 0.0004184444444444445,
-      "loss": 0.5568,
+      "learning_rate": 0.0004493793103448276,
+      "loss": 0.6088,
       "step": 870
     },
     {
       "epoch": 15.84,
-      "learning_rate": 0.0004182222222222222,
-      "loss": 0.4909,
+      "learning_rate": 0.0004492413793103448,
+      "loss": 0.4917,
       "step": 871
     },
     {
       "epoch": 15.85,
-      "learning_rate": 0.00041799999999999997,
-      "loss": 0.6172,
+      "learning_rate": 0.0004491034482758621,
+      "loss": 0.6624,
       "step": 872
     },
     {
       "epoch": 15.87,
-      "learning_rate": 0.0004177777777777778,
-      "loss": 0.5807,
+      "learning_rate": 0.0004489655172413793,
+      "loss": 0.5987,
       "step": 873
     },
     {
       "epoch": 15.89,
-      "learning_rate": 0.00041755555555555556,
-      "loss": 0.5564,
+      "learning_rate": 0.00044882758620689656,
+      "loss": 0.5801,
       "step": 874
     },
     {
       "epoch": 15.91,
-      "learning_rate": 0.00041733333333333336,
-      "loss": 0.5781,
+      "learning_rate": 0.00044868965517241377,
+      "loss": 0.5636,
       "step": 875
     },
     {
       "epoch": 15.93,
-      "learning_rate": 0.0004171111111111111,
-      "loss": 0.6187,
+      "learning_rate": 0.00044855172413793104,
+      "loss": 0.6308,
       "step": 876
     },
     {
       "epoch": 15.95,
-      "learning_rate": 0.0004168888888888889,
-      "loss": 0.5653,
+      "learning_rate": 0.00044841379310344825,
+      "loss": 0.5508,
       "step": 877
     },
     {
       "epoch": 15.96,
-      "learning_rate": 0.0004166666666666667,
-      "loss": 0.5564,
+      "learning_rate": 0.0004482758620689655,
+      "loss": 0.5722,
       "step": 878
     },
     {
       "epoch": 15.98,
-      "learning_rate": 0.00041644444444444444,
-      "loss": 0.6487,
+      "learning_rate": 0.00044813793103448273,
+      "loss": 0.619,
       "step": 879
     },
     {
       "epoch": 16.0,
-      "learning_rate": 0.00041622222222222223,
-      "loss": 0.4856,
+      "learning_rate": 0.000448,
+      "loss": 0.5317,
       "step": 880
     },
     {
       "epoch": 16.02,
-      "learning_rate": 0.000416,
-      "loss": 0.5642,
+      "learning_rate": 0.00044786206896551726,
+      "loss": 0.56,
       "step": 881
     },
     {
       "epoch": 16.04,
-      "learning_rate": 0.0004157777777777778,
-      "loss": 0.4826,
+      "learning_rate": 0.0004477241379310345,
+      "loss": 0.4917,
       "step": 882
     },
     {
       "epoch": 16.05,
-      "learning_rate": 0.00041555555555555557,
-      "loss": 0.5251,
+      "learning_rate": 0.00044758620689655174,
+      "loss": 0.5261,
       "step": 883
     },
     {
       "epoch": 16.07,
-      "learning_rate": 0.00041533333333333336,
-      "loss": 0.583,
+      "learning_rate": 0.000447448275862069,
+      "loss": 0.5664,
       "step": 884
     },
     {
       "epoch": 16.09,
-      "learning_rate": 0.0004151111111111111,
-      "loss": 0.6653,
+      "learning_rate": 0.0004473103448275862,
+      "loss": 0.6484,
       "step": 885
     },
     {
       "epoch": 16.11,
-      "learning_rate": 0.00041488888888888885,
-      "loss": 0.5377,
+      "learning_rate": 0.0004471724137931035,
+      "loss": 0.5191,
       "step": 886
     },
     {
       "epoch": 16.13,
-      "learning_rate": 0.0004146666666666667,
-      "loss": 0.622,
+      "learning_rate": 0.0004470344827586207,
+      "loss": 0.6142,
       "step": 887
     },
     {
       "epoch": 16.15,
-      "learning_rate": 0.00041444444444444444,
-      "loss": 0.5287,
+      "learning_rate": 0.00044689655172413796,
+      "loss": 0.5351,
       "step": 888
     },
     {
       "epoch": 16.16,
-      "learning_rate": 0.00041422222222222224,
-      "loss": 0.5145,
+      "learning_rate": 0.00044675862068965517,
+      "loss": 0.503,
       "step": 889
     },
     {
       "epoch": 16.18,
-      "learning_rate": 0.000414,
-      "loss": 0.5364,
+      "learning_rate": 0.00044662068965517244,
+      "loss": 0.5623,
       "step": 890
     },
     {
       "epoch": 16.2,
-      "learning_rate": 0.00041377777777777783,
-      "loss": 0.513,
+      "learning_rate": 0.00044648275862068965,
+      "loss": 0.5451,
       "step": 891
     },
     {
       "epoch": 16.22,
-      "learning_rate": 0.0004135555555555556,
-      "loss": 0.5481,
+      "learning_rate": 0.0004463448275862069,
+      "loss": 0.5471,
       "step": 892
     },
     {
       "epoch": 16.24,
-      "learning_rate": 0.0004133333333333333,
-      "loss": 0.5787,
+      "learning_rate": 0.0004462068965517241,
+      "loss": 0.611,
       "step": 893
     },
     {
       "epoch": 16.25,
-      "learning_rate": 0.0004131111111111111,
-      "loss": 0.4984,
+      "learning_rate": 0.0004460689655172414,
+      "loss": 0.5363,
       "step": 894
     },
     {
       "epoch": 16.27,
-      "learning_rate": 0.0004128888888888889,
-      "loss": 0.5416,
+      "learning_rate": 0.0004459310344827586,
+      "loss": 0.5472,
       "step": 895
     },
     {
       "epoch": 16.29,
-      "learning_rate": 0.0004126666666666667,
-      "loss": 0.5183,
+      "learning_rate": 0.00044579310344827587,
+      "loss": 0.5656,
       "step": 896
     },
     {
       "epoch": 16.31,
-      "learning_rate": 0.00041244444444444445,
-      "loss": 0.556,
+      "learning_rate": 0.0004456551724137931,
+      "loss": 0.5489,
       "step": 897
     },
     {
       "epoch": 16.33,
-      "learning_rate": 0.00041222222222222224,
-      "loss": 0.529,
+      "learning_rate": 0.0004455172413793104,
+      "loss": 0.5483,
       "step": 898
     },
     {
       "epoch": 16.35,
-      "learning_rate": 0.000412,
-      "loss": 0.4577,
+      "learning_rate": 0.0004453793103448276,
+      "loss": 0.5266,
       "step": 899
     },
     {
       "epoch": 16.36,
-      "learning_rate": 0.0004117777777777778,
-      "loss": 0.514,
+      "learning_rate": 0.0004452413793103449,
+      "loss": 0.5539,
       "step": 900
     },
     {
       "epoch": 16.38,
-      "learning_rate": 0.0004115555555555556,
-      "loss": 0.5532,
+      "learning_rate": 0.0004451034482758621,
+      "loss": 0.5132,
       "step": 901
     },
     {
       "epoch": 16.4,
-      "learning_rate": 0.0004113333333333333,
-      "loss": 0.5362,
+      "learning_rate": 0.00044496551724137936,
+      "loss": 0.5524,
       "step": 902
     },
     {
       "epoch": 16.42,
-      "learning_rate": 0.0004111111111111111,
-      "loss": 0.5474,
+      "learning_rate": 0.00044482758620689657,
+      "loss": 0.5619,
       "step": 903
     },
     {
       "epoch": 16.44,
-      "learning_rate": 0.0004108888888888889,
-      "loss": 0.5389,
+      "learning_rate": 0.0004446896551724138,
+      "loss": 0.5865,
       "step": 904
     },
     {
       "epoch": 16.45,
-      "learning_rate": 0.0004106666666666667,
-      "loss": 0.521,
+      "learning_rate": 0.00044455172413793105,
+      "loss": 0.5711,
       "step": 905
     },
     {
       "epoch": 16.47,
-      "learning_rate": 0.00041044444444444445,
-      "loss": 0.49,
+      "learning_rate": 0.00044441379310344826,
+      "loss": 0.5284,
       "step": 906
     },
     {
       "epoch": 16.49,
-      "learning_rate": 0.0004102222222222222,
-      "loss": 0.4473,
+      "learning_rate": 0.0004442758620689655,
+      "loss": 0.4792,
       "step": 907
     },
     {
       "epoch": 16.51,
-      "learning_rate": 0.00041,
-      "loss": 0.5312,
+      "learning_rate": 0.00044413793103448274,
+      "loss": 0.5166,
       "step": 908
     },
     {
       "epoch": 16.53,
-      "learning_rate": 0.0004097777777777778,
-      "loss": 0.4725,
+      "learning_rate": 0.000444,
+      "loss": 0.5325,
       "step": 909
     },
     {
       "epoch": 16.55,
-      "learning_rate": 0.0004095555555555556,
-      "loss": 0.5215,
+      "learning_rate": 0.0004438620689655172,
+      "loss": 0.5366,
       "step": 910
     },
     {
       "epoch": 16.56,
-      "learning_rate": 0.0004093333333333333,
-      "loss": 0.5377,
+      "learning_rate": 0.0004437241379310345,
+      "loss": 0.5202,
       "step": 911
     },
     {
       "epoch": 16.58,
-      "learning_rate": 0.0004091111111111111,
-      "loss": 0.5771,
+      "learning_rate": 0.0004435862068965517,
+      "loss": 0.6253,
       "step": 912
     },
     {
       "epoch": 16.6,
-      "learning_rate": 0.0004088888888888889,
-      "loss": 0.4745,
+      "learning_rate": 0.00044344827586206896,
+      "loss": 0.5242,
       "step": 913
     },
     {
       "epoch": 16.62,
-      "learning_rate": 0.00040866666666666666,
-      "loss": 0.5754,
+      "learning_rate": 0.00044331034482758617,
+      "loss": 0.5458,
       "step": 914
     },
     {
       "epoch": 16.64,
-      "learning_rate": 0.00040844444444444446,
-      "loss": 0.5553,
+      "learning_rate": 0.0004431724137931035,
+      "loss": 0.5792,
       "step": 915
     },
     {
       "epoch": 16.65,
-      "learning_rate": 0.0004082222222222222,
-      "loss": 0.6076,
+      "learning_rate": 0.0004430344827586207,
+      "loss": 0.5283,
       "step": 916
     },
     {
       "epoch": 16.67,
-      "learning_rate": 0.000408,
-      "loss": 0.5639,
+      "learning_rate": 0.00044289655172413797,
+      "loss": 0.6107,
       "step": 917
     },
     {
       "epoch": 16.69,
-      "learning_rate": 0.0004077777777777778,
-      "loss": 0.5597,
+      "learning_rate": 0.0004427586206896552,
+      "loss": 0.5579,
       "step": 918
     },
     {
       "epoch": 16.71,
-      "learning_rate": 0.0004075555555555556,
-      "loss": 0.533,
+      "learning_rate": 0.00044262068965517245,
+      "loss": 0.5361,
       "step": 919
     },
     {
       "epoch": 16.73,
-      "learning_rate": 0.00040733333333333333,
-      "loss": 0.4605,
+      "learning_rate": 0.00044248275862068966,
+      "loss": 0.4807,
       "step": 920
     },
     {
       "epoch": 16.75,
-      "learning_rate": 0.0004071111111111111,
-      "loss": 0.5368,
+      "learning_rate": 0.0004423448275862069,
+      "loss": 0.5383,
       "step": 921
     },
     {
       "epoch": 16.76,
-      "learning_rate": 0.0004068888888888889,
-      "loss": 0.4649,
+      "learning_rate": 0.00044220689655172414,
+      "loss": 0.5156,
       "step": 922
     },
     {
       "epoch": 16.78,
-      "learning_rate": 0.00040666666666666667,
-      "loss": 0.5969,
+      "learning_rate": 0.0004420689655172414,
+      "loss": 0.5918,
       "step": 923
     },
     {
       "epoch": 16.8,
-      "learning_rate": 0.00040644444444444446,
-      "loss": 0.5197,
+      "learning_rate": 0.0004419310344827586,
+      "loss": 0.5612,
       "step": 924
     },
     {
       "epoch": 16.82,
-      "learning_rate": 0.0004062222222222222,
-      "loss": 0.4939,
+      "learning_rate": 0.0004417931034482759,
+      "loss": 0.5269,
       "step": 925
     },
     {
       "epoch": 16.84,
-      "learning_rate": 0.00040600000000000006,
-      "loss": 0.5101,
+      "learning_rate": 0.0004416551724137931,
+      "loss": 0.5327,
       "step": 926
     },
     {
       "epoch": 16.85,
-      "learning_rate": 0.0004057777777777778,
-      "loss": 0.5102,
+      "learning_rate": 0.00044151724137931036,
+      "loss": 0.5367,
       "step": 927
     },
     {
       "epoch": 16.87,
-      "learning_rate": 0.00040555555555555554,
-      "loss": 0.5608,
+      "learning_rate": 0.00044137931034482757,
+      "loss": 0.5613,
       "step": 928
     },
     {
       "epoch": 16.89,
-      "learning_rate": 0.00040533333333333334,
-      "loss": 0.5667,
+      "learning_rate": 0.00044124137931034484,
+      "loss": 0.5723,
       "step": 929
     },
     {
       "epoch": 16.91,
-      "learning_rate": 0.0004051111111111111,
-      "loss": 0.5334,
+      "learning_rate": 0.00044110344827586205,
+      "loss": 0.5429,
       "step": 930
     },
     {
       "epoch": 16.93,
-      "learning_rate": 0.00040488888888888893,
-      "loss": 0.5235,
+      "learning_rate": 0.00044096551724137937,
+      "loss": 0.5491,
       "step": 931
     },
     {
       "epoch": 16.95,
-      "learning_rate": 0.00040466666666666667,
-      "loss": 0.5812,
+      "learning_rate": 0.0004408275862068966,
+      "loss": 0.5633,
       "step": 932
     },
     {
       "epoch": 16.96,
-      "learning_rate": 0.00040444444444444447,
-      "loss": 0.5131,
+      "learning_rate": 0.00044068965517241385,
+      "loss": 0.5177,
       "step": 933
     },
     {
       "epoch": 16.98,
-      "learning_rate": 0.0004042222222222222,
-      "loss": 0.5416,
+      "learning_rate": 0.00044055172413793106,
+      "loss": 0.5953,
       "step": 934
     },
     {
       "epoch": 17.0,
-      "learning_rate": 0.000404,
-      "loss": 0.4502,
+      "learning_rate": 0.0004404137931034483,
+      "loss": 0.4534,
       "step": 935
     },
     {
       "epoch": 17.02,
-      "learning_rate": 0.0004037777777777778,
-      "loss": 0.505,
+      "learning_rate": 0.00044027586206896554,
+      "loss": 0.5338,
       "step": 936
     },
     {
       "epoch": 17.04,
-      "learning_rate": 0.00040355555555555555,
-      "loss": 0.4935,
+      "learning_rate": 0.00044013793103448275,
+      "loss": 0.5404,
       "step": 937
     },
     {
       "epoch": 17.05,
-      "learning_rate": 0.00040333333333333334,
-      "loss": 0.4436,
+      "learning_rate": 0.00044,
+      "loss": 0.5012,
       "step": 938
     },
     {
       "epoch": 17.07,
-      "learning_rate": 0.0004031111111111111,
-      "loss": 0.5719,
+      "learning_rate": 0.00043986206896551723,
+      "loss": 0.6215,
       "step": 939
     },
     {
       "epoch": 17.09,
-      "learning_rate": 0.00040288888888888894,
-      "loss": 0.4852,
+      "learning_rate": 0.0004397241379310345,
+      "loss": 0.5329,
       "step": 940
     },
     {
       "epoch": 17.11,
-      "learning_rate": 0.0004026666666666667,
-      "loss": 0.5232,
+      "learning_rate": 0.0004395862068965517,
+      "loss": 0.5408,
       "step": 941
     },
     {
       "epoch": 17.13,
-      "learning_rate": 0.0004024444444444444,
-      "loss": 0.5754,
+      "learning_rate": 0.00043944827586206897,
+      "loss": 0.5721,
       "step": 942
     },
     {
       "epoch": 17.15,
-      "learning_rate": 0.0004022222222222222,
-      "loss": 0.5362,
+      "learning_rate": 0.0004393103448275862,
+      "loss": 0.5599,
       "step": 943
     },
     {
       "epoch": 17.16,
-      "learning_rate": 0.000402,
-      "loss": 0.4674,
+      "learning_rate": 0.00043917241379310345,
+      "loss": 0.5431,
       "step": 944
     },
     {
       "epoch": 17.18,
-      "learning_rate": 0.0004017777777777778,
-      "loss": 0.5048,
+      "learning_rate": 0.00043903448275862066,
+      "loss": 0.5591,
       "step": 945
     },
     {
       "epoch": 17.2,
-      "learning_rate": 0.00040155555555555555,
-      "loss": 0.5008,
+      "learning_rate": 0.00043889655172413793,
+      "loss": 0.5481,
       "step": 946
     },
     {
       "epoch": 17.22,
-      "learning_rate": 0.00040133333333333335,
-      "loss": 0.554,
+      "learning_rate": 0.00043875862068965514,
+      "loss": 0.5797,
       "step": 947
     },
     {
       "epoch": 17.24,
-      "learning_rate": 0.0004011111111111111,
-      "loss": 0.5233,
+      "learning_rate": 0.00043862068965517246,
+      "loss": 0.5658,
       "step": 948
     },
     {
       "epoch": 17.25,
-      "learning_rate": 0.0004008888888888889,
-      "loss": 0.5562,
+      "learning_rate": 0.00043848275862068967,
+      "loss": 0.5807,
       "step": 949
     },
     {
       "epoch": 17.27,
-      "learning_rate": 0.0004006666666666667,
-      "loss": 0.5097,
+      "learning_rate": 0.00043834482758620694,
+      "loss": 0.5408,
       "step": 950
     },
     {
       "epoch": 17.29,
-      "learning_rate": 0.0004004444444444444,
-      "loss": 0.5174,
+      "learning_rate": 0.00043820689655172415,
+      "loss": 0.5182,
       "step": 951
     },
     {
       "epoch": 17.31,
-      "learning_rate": 0.0004002222222222222,
-      "loss": 0.5286,
+      "learning_rate": 0.0004380689655172414,
+      "loss": 0.539,
       "step": 952
     },
     {
       "epoch": 17.33,
-      "learning_rate": 0.0004,
-      "loss": 0.548,
+      "learning_rate": 0.00043793103448275863,
+      "loss": 0.6095,
       "step": 953
     },
     {
       "epoch": 17.35,
-      "learning_rate": 0.0003997777777777778,
-      "loss": 0.5727,
+      "learning_rate": 0.0004377931034482759,
+      "loss": 0.5947,
       "step": 954
     },
     {
       "epoch": 17.36,
-      "learning_rate": 0.00039955555555555556,
-      "loss": 0.54,
+      "learning_rate": 0.0004376551724137931,
+      "loss": 0.568,
       "step": 955
     },
     {
       "epoch": 17.38,
-      "learning_rate": 0.0003993333333333333,
-      "loss": 0.4795,
+      "learning_rate": 0.00043751724137931037,
+      "loss": 0.4808,
       "step": 956
     },
     {
       "epoch": 17.4,
-      "learning_rate": 0.00039911111111111115,
-      "loss": 0.5275,
+      "learning_rate": 0.0004373793103448276,
+      "loss": 0.5277,
       "step": 957
     },
     {
       "epoch": 17.42,
-      "learning_rate": 0.0003988888888888889,
-      "loss": 0.4985,
+      "learning_rate": 0.00043724137931034485,
+      "loss": 0.5243,
       "step": 958
     },
     {
       "epoch": 17.44,
-      "learning_rate": 0.0003986666666666667,
-      "loss": 0.567,
+      "learning_rate": 0.00043710344827586206,
+      "loss": 0.5248,
       "step": 959
     },
     {
       "epoch": 17.45,
-      "learning_rate": 0.00039844444444444443,
-      "loss": 0.5116,
+      "learning_rate": 0.0004369655172413793,
+      "loss": 0.545,
       "step": 960
     },
     {
       "epoch": 17.47,
-      "learning_rate": 0.00039822222222222223,
-      "loss": 0.5216,
+      "learning_rate": 0.00043682758620689654,
+      "loss": 0.5298,
       "step": 961
     },
     {
       "epoch": 17.49,
-      "learning_rate": 0.000398,
-      "loss": 0.4937,
+      "learning_rate": 0.0004366896551724138,
+      "loss": 0.4961,
       "step": 962
     },
     {
       "epoch": 17.51,
-      "learning_rate": 0.00039777777777777777,
-      "loss": 0.4838,
+      "learning_rate": 0.000436551724137931,
+      "loss": 0.5142,
       "step": 963
     },
     {
       "epoch": 17.53,
-      "learning_rate": 0.00039755555555555556,
-      "loss": 0.5595,
+      "learning_rate": 0.0004364137931034483,
+      "loss": 0.5845,
       "step": 964
     },
     {
       "epoch": 17.55,
-      "learning_rate": 0.0003973333333333333,
-      "loss": 0.4846,
+      "learning_rate": 0.00043627586206896555,
+      "loss": 0.5243,
       "step": 965
     },
     {
       "epoch": 17.56,
-      "learning_rate": 0.00039711111111111116,
-      "loss": 0.5562,
+      "learning_rate": 0.0004361379310344828,
+      "loss": 0.565,
       "step": 966
     },
     {
       "epoch": 17.58,
-      "learning_rate": 0.0003968888888888889,
-      "loss": 0.5229,
+      "learning_rate": 0.000436,
+      "loss": 0.5144,
       "step": 967
     },
     {
       "epoch": 17.6,
-      "learning_rate": 0.0003966666666666667,
-      "loss": 0.5829,
+      "learning_rate": 0.0004358620689655173,
+      "loss": 0.5796,
       "step": 968
     },
     {
       "epoch": 17.62,
-      "learning_rate": 0.00039644444444444444,
-      "loss": 0.4936,
+      "learning_rate": 0.0004357241379310345,
+      "loss": 0.4966,
       "step": 969
     },
     {
       "epoch": 17.64,
-      "learning_rate": 0.0003962222222222222,
-      "loss": 0.5096,
+      "learning_rate": 0.0004355862068965517,
+      "loss": 0.5195,
       "step": 970
     },
     {
       "epoch": 17.65,
-      "learning_rate": 0.00039600000000000003,
-      "loss": 0.4996,
+      "learning_rate": 0.000435448275862069,
+      "loss": 0.5112,
       "step": 971
     },
     {
       "epoch": 17.67,
-      "learning_rate": 0.00039577777777777777,
-      "loss": 0.5556,
+      "learning_rate": 0.0004353103448275862,
+      "loss": 0.6046,
       "step": 972
     },
     {
       "epoch": 17.69,
-      "learning_rate": 0.00039555555555555557,
-      "loss": 0.5059,
+      "learning_rate": 0.00043517241379310346,
+      "loss": 0.5046,
       "step": 973
     },
     {
       "epoch": 17.71,
-      "learning_rate": 0.0003953333333333333,
-      "loss": 0.5602,
+      "learning_rate": 0.00043503448275862067,
+      "loss": 0.5956,
       "step": 974
     },
     {
       "epoch": 17.73,
-      "learning_rate": 0.00039511111111111116,
-      "loss": 0.5248,
+      "learning_rate": 0.00043489655172413794,
+      "loss": 0.5467,
       "step": 975
     },
     {
       "epoch": 17.75,
-      "learning_rate": 0.0003948888888888889,
-      "loss": 0.6092,
+      "learning_rate": 0.00043475862068965515,
+      "loss": 0.6128,
       "step": 976
     },
     {
       "epoch": 17.76,
-      "learning_rate": 0.0003946666666666667,
-      "loss": 0.5014,
+      "learning_rate": 0.0004346206896551724,
+      "loss": 0.5185,
       "step": 977
     },
     {
       "epoch": 17.78,
-      "learning_rate": 0.00039444444444444444,
-      "loss": 0.5061,
+      "learning_rate": 0.00043448275862068963,
+      "loss": 0.5493,
       "step": 978
     },
     {
       "epoch": 17.8,
-      "learning_rate": 0.0003942222222222222,
-      "loss": 0.4982,
+      "learning_rate": 0.0004343448275862069,
+      "loss": 0.501,
       "step": 979
     },
     {
       "epoch": 17.82,
-      "learning_rate": 0.00039400000000000004,
-      "loss": 0.5767,
+      "learning_rate": 0.0004342068965517241,
+      "loss": 0.5536,
       "step": 980
     },
     {
       "epoch": 17.84,
-      "learning_rate": 0.0003937777777777778,
-      "loss": 0.4801,
+      "learning_rate": 0.00043406896551724137,
+      "loss": 0.5018,
       "step": 981
     },
     {
       "epoch": 17.85,
-      "learning_rate": 0.0003935555555555556,
-      "loss": 0.5333,
+      "learning_rate": 0.00043393103448275864,
+      "loss": 0.5595,
       "step": 982
     },
     {
       "epoch": 17.87,
-      "learning_rate": 0.0003933333333333333,
-      "loss": 0.5817,
+      "learning_rate": 0.0004337931034482759,
+      "loss": 0.5863,
       "step": 983
     },
     {
       "epoch": 17.89,
-      "learning_rate": 0.00039311111111111117,
-      "loss": 0.6053,
+      "learning_rate": 0.0004336551724137931,
+      "loss": 0.6514,
       "step": 984
     },
     {
       "epoch": 17.91,
-      "learning_rate": 0.0003928888888888889,
-      "loss": 0.5194,
+      "learning_rate": 0.0004335172413793104,
+      "loss": 0.5924,
       "step": 985
     },
     {
       "epoch": 17.93,
-      "learning_rate": 0.00039266666666666665,
-      "loss": 0.5796,
+      "learning_rate": 0.0004333793103448276,
+      "loss": 0.5965,
       "step": 986
     },
     {
       "epoch": 17.95,
-      "learning_rate": 0.00039244444444444445,
-      "loss": 0.5251,
+      "learning_rate": 0.00043324137931034486,
+      "loss": 0.5293,
       "step": 987
     },
     {
       "epoch": 17.96,
-      "learning_rate": 0.00039222222222222225,
-      "loss": 0.6369,
+      "learning_rate": 0.00043310344827586207,
+      "loss": 0.606,
       "step": 988
     },
     {
       "epoch": 17.98,
-      "learning_rate": 0.00039200000000000004,
-      "loss": 0.594,
+      "learning_rate": 0.00043296551724137934,
+      "loss": 0.6079,
       "step": 989
     },
     {
       "epoch": 18.0,
-      "learning_rate": 0.0003917777777777778,
-      "loss": 0.4169,
+      "learning_rate": 0.00043282758620689655,
+      "loss": 0.4955,
       "step": 990
     },
     {
       "epoch": 18.02,
-      "learning_rate": 0.0003915555555555556,
-      "loss": 0.5526,
+      "learning_rate": 0.0004326896551724138,
+      "loss": 0.5906,
       "step": 991
     },
     {
       "epoch": 18.04,
-      "learning_rate": 0.0003913333333333333,
-      "loss": 0.5082,
+      "learning_rate": 0.00043255172413793103,
+      "loss": 0.5204,
       "step": 992
     },
     {
       "epoch": 18.05,
-      "learning_rate": 0.0003911111111111111,
-      "loss": 0.4657,
+      "learning_rate": 0.0004324137931034483,
+      "loss": 0.4622,
       "step": 993
     },
     {
       "epoch": 18.07,
-      "learning_rate": 0.0003908888888888889,
-      "loss": 0.4774,
+      "learning_rate": 0.0004322758620689655,
+      "loss": 0.5158,
       "step": 994
     },
     {
       "epoch": 18.09,
-      "learning_rate": 0.00039066666666666666,
-      "loss": 0.6214,
+      "learning_rate": 0.00043213793103448277,
+      "loss": 0.6422,
       "step": 995
     },
     {
       "epoch": 18.11,
-      "learning_rate": 0.00039044444444444445,
-      "loss": 0.4671,
+      "learning_rate": 0.000432,
+      "loss": 0.5438,
       "step": 996
     },
     {
       "epoch": 18.13,
-      "learning_rate": 0.00039022222222222225,
-      "loss": 0.5042,
+      "learning_rate": 0.00043186206896551725,
+      "loss": 0.5258,
       "step": 997
     },
     {
       "epoch": 18.15,
-      "learning_rate": 0.00039000000000000005,
-      "loss": 0.4834,
+      "learning_rate": 0.00043172413793103446,
+      "loss": 0.5167,
       "step": 998
     },
     {
       "epoch": 18.16,
-      "learning_rate": 0.0003897777777777778,
-      "loss": 0.4661,
+      "learning_rate": 0.0004315862068965518,
+      "loss": 0.5,
       "step": 999
     },
     {
       "epoch": 18.18,
-      "learning_rate": 0.00038955555555555553,
-      "loss": 0.5166,
+      "learning_rate": 0.000431448275862069,
+      "loss": 0.5739,
       "step": 1000
     },
     {
       "epoch": 18.18,
-      "eval_cer": 0.10703308895185835,
-      "eval_loss": 0.4058302044868469,
-      "eval_runtime": 13.7955,
-      "eval_samples_per_second": 119.387,
-      "eval_steps_per_second": 1.885,
-      "eval_wer": 0.42375651108160556,
+      "eval_cer": 0.10985172191487129,
+      "eval_loss": 0.40066882967948914,
+      "eval_runtime": 13.8129,
+      "eval_samples_per_second": 119.236,
+      "eval_steps_per_second": 1.882,
+      "eval_wer": 0.4357062608518027,
       "step": 1000
     },
     {
       "epoch": 18.2,
-      "learning_rate": 0.00038933333333333333,
-      "loss": 0.4833,
+      "learning_rate": 0.00043131034482758626,
+      "loss": 0.5327,
       "step": 1001
     },
     {
       "epoch": 18.22,
-      "learning_rate": 0.0003891111111111111,
-      "loss": 0.5381,
+      "learning_rate": 0.00043117241379310347,
+      "loss": 0.5624,
       "step": 1002
     },
     {
       "epoch": 18.24,
-      "learning_rate": 0.0003888888888888889,
-      "loss": 0.5217,
+      "learning_rate": 0.0004310344827586207,
+      "loss": 0.5754,
       "step": 1003
     },
     {
       "epoch": 18.25,
-      "learning_rate": 0.00038866666666666666,
-      "loss": 0.5053,
+      "learning_rate": 0.00043089655172413795,
+      "loss": 0.4967,
       "step": 1004
     },
     {
       "epoch": 18.27,
-      "learning_rate": 0.00038844444444444446,
-      "loss": 0.4945,
+      "learning_rate": 0.00043075862068965516,
+      "loss": 0.5091,
       "step": 1005
     },
     {
       "epoch": 18.29,
-      "learning_rate": 0.00038822222222222226,
-      "loss": 0.521,
+      "learning_rate": 0.00043062068965517243,
+      "loss": 0.5303,
       "step": 1006
     },
     {
       "epoch": 18.31,
-      "learning_rate": 0.000388,
-      "loss": 0.5191,
+      "learning_rate": 0.00043048275862068964,
+      "loss": 0.5443,
       "step": 1007
     },
     {
       "epoch": 18.33,
-      "learning_rate": 0.0003877777777777778,
-      "loss": 0.5179,
+      "learning_rate": 0.0004303448275862069,
+      "loss": 0.5319,
       "step": 1008
     },
     {
       "epoch": 18.35,
-      "learning_rate": 0.00038755555555555554,
-      "loss": 0.5311,
+      "learning_rate": 0.0004302068965517241,
+      "loss": 0.6022,
       "step": 1009
     },
     {
       "epoch": 18.36,
-      "learning_rate": 0.00038733333333333333,
+      "learning_rate": 0.0004300689655172414,
       "loss": 0.5273,
       "step": 1010
     },
     {
       "epoch": 18.38,
-      "learning_rate": 0.00038711111111111113,
-      "loss": 0.5278,
+      "learning_rate": 0.0004299310344827586,
+      "loss": 0.5585,
       "step": 1011
     },
     {
       "epoch": 18.4,
-      "learning_rate": 0.0003868888888888889,
-      "loss": 0.5245,
+      "learning_rate": 0.00042979310344827586,
+      "loss": 0.5734,
       "step": 1012
     },
     {
       "epoch": 18.42,
-      "learning_rate": 0.00038666666666666667,
-      "loss": 0.4747,
+      "learning_rate": 0.0004296551724137931,
+      "loss": 0.5343,
       "step": 1013
     },
     {
       "epoch": 18.44,
-      "learning_rate": 0.0003864444444444444,
-      "loss": 0.4757,
+      "learning_rate": 0.00042951724137931034,
+      "loss": 0.522,
       "step": 1014
     },
     {
       "epoch": 18.45,
-      "learning_rate": 0.00038622222222222226,
-      "loss": 0.5681,
+      "learning_rate": 0.0004293793103448276,
+      "loss": 0.5479,
       "step": 1015
     },
     {
       "epoch": 18.47,
-      "learning_rate": 0.000386,
-      "loss": 0.4909,
+      "learning_rate": 0.00042924137931034487,
+      "loss": 0.5165,
       "step": 1016
     },
     {
       "epoch": 18.49,
-      "learning_rate": 0.0003857777777777778,
-      "loss": 0.4773,
+      "learning_rate": 0.0004291034482758621,
+      "loss": 0.5253,
       "step": 1017
     },
     {
       "epoch": 18.51,
-      "learning_rate": 0.00038555555555555554,
-      "loss": 0.5559,
+      "learning_rate": 0.00042896551724137935,
+      "loss": 0.5882,
       "step": 1018
     },
     {
       "epoch": 18.53,
-      "learning_rate": 0.0003853333333333334,
-      "loss": 0.4634,
+      "learning_rate": 0.00042882758620689656,
+      "loss": 0.5055,
       "step": 1019
     },
     {
       "epoch": 18.55,
-      "learning_rate": 0.00038511111111111114,
-      "loss": 0.4778,
+      "learning_rate": 0.00042868965517241383,
+      "loss": 0.5334,
       "step": 1020
     },
     {
       "epoch": 18.56,
-      "learning_rate": 0.0003848888888888889,
-      "loss": 0.5013,
+      "learning_rate": 0.00042855172413793104,
+      "loss": 0.5703,
       "step": 1021
     },
     {
       "epoch": 18.58,
-      "learning_rate": 0.0003846666666666667,
-      "loss": 0.5821,
+      "learning_rate": 0.0004284137931034483,
+      "loss": 0.5804,
       "step": 1022
     },
     {
       "epoch": 18.6,
-      "learning_rate": 0.0003844444444444444,
-      "loss": 0.5666,
+      "learning_rate": 0.0004282758620689655,
+      "loss": 0.6351,
       "step": 1023
     },
     {
       "epoch": 18.62,
-      "learning_rate": 0.00038422222222222227,
-      "loss": 0.5348,
+      "learning_rate": 0.0004281379310344828,
+      "loss": 0.5542,
       "step": 1024
     },
     {
       "epoch": 18.64,
-      "learning_rate": 0.000384,
-      "loss": 0.482,
+      "learning_rate": 0.000428,
+      "loss": 0.5001,
       "step": 1025
     },
     {
       "epoch": 18.65,
-      "learning_rate": 0.0003837777777777778,
-      "loss": 0.5249,
+      "learning_rate": 0.00042786206896551726,
+      "loss": 0.5971,
       "step": 1026
     },
     {
       "epoch": 18.67,
-      "learning_rate": 0.00038355555555555555,
-      "loss": 0.5216,
+      "learning_rate": 0.0004277241379310345,
+      "loss": 0.5724,
       "step": 1027
     },
     {
       "epoch": 18.69,
-      "learning_rate": 0.00038333333333333334,
-      "loss": 0.5047,
+      "learning_rate": 0.00042758620689655174,
+      "loss": 0.6076,
       "step": 1028
     },
     {
       "epoch": 18.71,
-      "learning_rate": 0.00038311111111111114,
-      "loss": 0.5283,
+      "learning_rate": 0.00042744827586206895,
+      "loss": 0.586,
       "step": 1029
     },
     {
       "epoch": 18.73,
-      "learning_rate": 0.0003828888888888889,
-      "loss": 0.5301,
+      "learning_rate": 0.0004273103448275862,
+      "loss": 0.5554,
       "step": 1030
     },
     {
       "epoch": 18.75,
-      "learning_rate": 0.0003826666666666667,
-      "loss": 0.5303,
+      "learning_rate": 0.00042717241379310343,
+      "loss": 0.5398,
       "step": 1031
     },
     {
       "epoch": 18.76,
-      "learning_rate": 0.0003824444444444444,
-      "loss": 0.509,
+      "learning_rate": 0.00042703448275862075,
+      "loss": 0.5242,
       "step": 1032
     },
     {
       "epoch": 18.78,
-      "learning_rate": 0.0003822222222222223,
-      "loss": 0.525,
+      "learning_rate": 0.00042689655172413796,
+      "loss": 0.5747,
       "step": 1033
     },
     {
       "epoch": 18.8,
-      "learning_rate": 0.000382,
-      "loss": 0.5554,
+      "learning_rate": 0.0004267586206896552,
+      "loss": 0.6691,
       "step": 1034
     },
     {
       "epoch": 18.82,
-      "learning_rate": 0.00038177777777777776,
-      "loss": 0.4874,
+      "learning_rate": 0.00042662068965517244,
+      "loss": 0.5623,
       "step": 1035
     },
     {
       "epoch": 18.84,
-      "learning_rate": 0.00038155555555555555,
-      "loss": 0.4786,
+      "learning_rate": 0.00042648275862068965,
+      "loss": 0.5915,
       "step": 1036
     },
     {
       "epoch": 18.85,
-      "learning_rate": 0.00038133333333333335,
-      "loss": 0.4368,
+      "learning_rate": 0.0004263448275862069,
+      "loss": 0.529,
       "step": 1037
     },
     {
       "epoch": 18.87,
-      "learning_rate": 0.00038111111111111115,
-      "loss": 0.4799,
+      "learning_rate": 0.00042620689655172413,
+      "loss": 0.5927,
       "step": 1038
     },
     {
       "epoch": 18.89,
-      "learning_rate": 0.0003808888888888889,
-      "loss": 0.4745,
+      "learning_rate": 0.0004260689655172414,
+      "loss": 0.5346,
       "step": 1039
     },
     {
       "epoch": 18.91,
-      "learning_rate": 0.0003806666666666667,
-      "loss": 0.5432,
+      "learning_rate": 0.0004259310344827586,
+      "loss": 0.5835,
       "step": 1040
     },
     {
       "epoch": 18.93,
-      "learning_rate": 0.00038044444444444443,
-      "loss": 0.5946,
+      "learning_rate": 0.00042579310344827587,
+      "loss": 0.6416,
       "step": 1041
     },
     {
       "epoch": 18.95,
-      "learning_rate": 0.0003802222222222222,
-      "loss": 0.4719,
+      "learning_rate": 0.0004256551724137931,
+      "loss": 0.5284,
       "step": 1042
     },
     {
       "epoch": 18.96,
-      "learning_rate": 0.00038,
-      "loss": 0.5357,
+      "learning_rate": 0.00042551724137931035,
+      "loss": 0.5723,
       "step": 1043
     },
     {
       "epoch": 18.98,
-      "learning_rate": 0.00037977777777777776,
-      "loss": 0.5585,
+      "learning_rate": 0.00042537931034482756,
+      "loss": 0.5841,
       "step": 1044
     },
     {
       "epoch": 19.0,
-      "learning_rate": 0.00037955555555555556,
-      "loss": 0.5219,
+      "learning_rate": 0.00042524137931034483,
+      "loss": 0.5915,
       "step": 1045
     },
     {
       "epoch": 19.02,
-      "learning_rate": 0.00037933333333333336,
-      "loss": 0.5676,
+      "learning_rate": 0.00042510344827586204,
+      "loss": 0.6382,
       "step": 1046
     },
     {
       "epoch": 19.04,
-      "learning_rate": 0.00037911111111111115,
-      "loss": 0.4579,
+      "learning_rate": 0.0004249655172413793,
+      "loss": 0.5142,
       "step": 1047
     },
     {
       "epoch": 19.05,
-      "learning_rate": 0.0003788888888888889,
-      "loss": 0.5285,
+      "learning_rate": 0.0004248275862068965,
+      "loss": 0.5735,
       "step": 1048
     },
     {
       "epoch": 19.07,
-      "learning_rate": 0.00037866666666666664,
-      "loss": 0.432,
+      "learning_rate": 0.00042468965517241384,
+      "loss": 0.4635,
       "step": 1049
     },
     {
       "epoch": 19.09,
-      "learning_rate": 0.0003784444444444445,
-      "loss": 0.4974,
+      "learning_rate": 0.00042455172413793105,
+      "loss": 0.5044,
       "step": 1050
     },
     {
       "epoch": 19.11,
-      "learning_rate": 0.00037822222222222223,
-      "loss": 0.4795,
+      "learning_rate": 0.0004244137931034483,
+      "loss": 0.5302,
       "step": 1051
     },
     {
       "epoch": 19.13,
-      "learning_rate": 0.000378,
-      "loss": 0.5108,
+      "learning_rate": 0.00042427586206896553,
+      "loss": 0.5422,
       "step": 1052
     },
     {
       "epoch": 19.15,
-      "learning_rate": 0.00037777777777777777,
-      "loss": 0.583,
+      "learning_rate": 0.0004241379310344828,
+      "loss": 0.6127,
       "step": 1053
     },
     {
       "epoch": 19.16,
-      "learning_rate": 0.00037755555555555557,
-      "loss": 0.4707,
+      "learning_rate": 0.000424,
+      "loss": 0.5567,
       "step": 1054
     },
     {
       "epoch": 19.18,
-      "learning_rate": 0.00037733333333333336,
-      "loss": 0.4968,
+      "learning_rate": 0.00042386206896551727,
+      "loss": 0.5704,
       "step": 1055
     },
     {
       "epoch": 19.2,
-      "learning_rate": 0.0003771111111111111,
-      "loss": 0.5548,
+      "learning_rate": 0.0004237241379310345,
+      "loss": 0.6051,
       "step": 1056
     },
     {
       "epoch": 19.22,
-      "learning_rate": 0.0003768888888888889,
-      "loss": 0.4948,
+      "learning_rate": 0.00042358620689655175,
+      "loss": 0.5522,
       "step": 1057
     },
     {
       "epoch": 19.24,
-      "learning_rate": 0.00037666666666666664,
-      "loss": 0.412,
+      "learning_rate": 0.00042344827586206896,
+      "loss": 0.4735,
       "step": 1058
     },
     {
       "epoch": 19.25,
-      "learning_rate": 0.0003764444444444445,
-      "loss": 0.5057,
+      "learning_rate": 0.00042331034482758623,
+      "loss": 0.5915,
       "step": 1059
     },
     {
       "epoch": 19.27,
-      "learning_rate": 0.00037622222222222224,
-      "loss": 0.4255,
+      "learning_rate": 0.00042317241379310344,
+      "loss": 0.491,
       "step": 1060
     },
     {
       "epoch": 19.29,
-      "learning_rate": 0.00037600000000000003,
-      "loss": 0.489,
+      "learning_rate": 0.0004230344827586207,
+      "loss": 0.5032,
       "step": 1061
     },
     {
       "epoch": 19.31,
-      "learning_rate": 0.0003757777777777778,
-      "loss": 0.501,
+      "learning_rate": 0.0004228965517241379,
+      "loss": 0.5245,
       "step": 1062
     },
     {
       "epoch": 19.33,
-      "learning_rate": 0.0003755555555555555,
-      "loss": 0.5372,
+      "learning_rate": 0.0004227586206896552,
+      "loss": 0.5677,
       "step": 1063
     },
     {
       "epoch": 19.35,
-      "learning_rate": 0.00037533333333333337,
-      "loss": 0.4937,
+      "learning_rate": 0.0004226206896551724,
+      "loss": 0.548,
       "step": 1064
     },
     {
       "epoch": 19.36,
-      "learning_rate": 0.0003751111111111111,
-      "loss": 0.4534,
+      "learning_rate": 0.00042248275862068966,
+      "loss": 0.4895,
       "step": 1065
     },
     {
       "epoch": 19.38,
-      "learning_rate": 0.0003748888888888889,
-      "loss": 0.4997,
+      "learning_rate": 0.00042234482758620693,
+      "loss": 0.5199,
       "step": 1066
     },
     {
       "epoch": 19.4,
-      "learning_rate": 0.00037466666666666665,
-      "loss": 0.4557,
+      "learning_rate": 0.0004222068965517242,
+      "loss": 0.4835,
       "step": 1067
     },
     {
       "epoch": 19.42,
-      "learning_rate": 0.0003744444444444445,
-      "loss": 0.5096,
+      "learning_rate": 0.0004220689655172414,
+      "loss": 0.531,
       "step": 1068
     },
     {
       "epoch": 19.44,
-      "learning_rate": 0.00037422222222222224,
-      "loss": 0.4369,
+      "learning_rate": 0.0004219310344827586,
+      "loss": 0.5284,
       "step": 1069
     },
     {
       "epoch": 19.45,
-      "learning_rate": 0.000374,
-      "loss": 0.4648,
+      "learning_rate": 0.0004217931034482759,
+      "loss": 0.5462,
       "step": 1070
     },
     {
       "epoch": 19.47,
-      "learning_rate": 0.0003737777777777778,
-      "loss": 0.5215,
+      "learning_rate": 0.0004216551724137931,
+      "loss": 0.5309,
       "step": 1071
     },
     {
       "epoch": 19.49,
-      "learning_rate": 0.0003735555555555555,
-      "loss": 0.4686,
+      "learning_rate": 0.00042151724137931036,
+      "loss": 0.5191,
       "step": 1072
     },
     {
       "epoch": 19.51,
-      "learning_rate": 0.0003733333333333334,
-      "loss": 0.5067,
+      "learning_rate": 0.0004213793103448276,
+      "loss": 0.5693,
       "step": 1073
     },
     {
       "epoch": 19.53,
-      "learning_rate": 0.0003731111111111111,
-      "loss": 0.5113,
+      "learning_rate": 0.00042124137931034484,
+      "loss": 0.5757,
       "step": 1074
     },
     {
       "epoch": 19.55,
-      "learning_rate": 0.0003728888888888889,
-      "loss": 0.48,
+      "learning_rate": 0.00042110344827586205,
+      "loss": 0.5173,
       "step": 1075
     },
     {
       "epoch": 19.56,
-      "learning_rate": 0.00037266666666666665,
-      "loss": 0.5198,
+      "learning_rate": 0.0004209655172413793,
+      "loss": 0.5259,
       "step": 1076
     },
     {
       "epoch": 19.58,
-      "learning_rate": 0.00037244444444444445,
-      "loss": 0.5036,
+      "learning_rate": 0.00042082758620689653,
+      "loss": 0.5808,
       "step": 1077
     },
     {
       "epoch": 19.6,
-      "learning_rate": 0.00037222222222222225,
-      "loss": 0.4759,
+      "learning_rate": 0.0004206896551724138,
+      "loss": 0.5039,
       "step": 1078
     },
     {
       "epoch": 19.62,
-      "learning_rate": 0.000372,
-      "loss": 0.4328,
+      "learning_rate": 0.000420551724137931,
+      "loss": 0.4827,
       "step": 1079
     },
     {
       "epoch": 19.64,
-      "learning_rate": 0.0003717777777777778,
-      "loss": 0.5229,
+      "learning_rate": 0.0004204137931034483,
+      "loss": 0.5554,
       "step": 1080
     },
     {
       "epoch": 19.65,
-      "learning_rate": 0.00037155555555555553,
-      "loss": 0.596,
+      "learning_rate": 0.0004202758620689655,
+      "loss": 0.6478,
       "step": 1081
     },
     {
       "epoch": 19.67,
-      "learning_rate": 0.0003713333333333334,
-      "loss": 0.512,
+      "learning_rate": 0.0004201379310344828,
+      "loss": 0.5527,
       "step": 1082
     },
     {
       "epoch": 19.69,
-      "learning_rate": 0.0003711111111111111,
-      "loss": 0.6065,
+      "learning_rate": 0.00042,
+      "loss": 0.6819,
       "step": 1083
     },
     {
       "epoch": 19.71,
-      "learning_rate": 0.00037088888888888886,
-      "loss": 0.522,
+      "learning_rate": 0.0004198620689655173,
+      "loss": 0.5277,
       "step": 1084
     },
     {
       "epoch": 19.73,
-      "learning_rate": 0.00037066666666666666,
-      "loss": 0.5174,
+      "learning_rate": 0.0004197241379310345,
+      "loss": 0.5817,
       "step": 1085
     },
     {
       "epoch": 19.75,
-      "learning_rate": 0.00037044444444444446,
-      "loss": 0.5017,
+      "learning_rate": 0.00041958620689655176,
+      "loss": 0.5534,
       "step": 1086
     },
     {
       "epoch": 19.76,
-      "learning_rate": 0.00037022222222222225,
-      "loss": 0.4742,
+      "learning_rate": 0.000419448275862069,
+      "loss": 0.5513,
       "step": 1087
     },
     {
       "epoch": 19.78,
-      "learning_rate": 0.00037,
-      "loss": 0.4694,
+      "learning_rate": 0.00041931034482758624,
+      "loss": 0.514,
       "step": 1088
     },
     {
       "epoch": 19.8,
-      "learning_rate": 0.0003697777777777778,
-      "loss": 0.5003,
+      "learning_rate": 0.00041917241379310345,
+      "loss": 0.5571,
       "step": 1089
     },
     {
       "epoch": 19.82,
-      "learning_rate": 0.0003695555555555556,
-      "loss": 0.4718,
+      "learning_rate": 0.0004190344827586207,
+      "loss": 0.5365,
       "step": 1090
     },
     {
       "epoch": 19.84,
-      "learning_rate": 0.00036933333333333333,
-      "loss": 0.4761,
+      "learning_rate": 0.00041889655172413793,
+      "loss": 0.495,
       "step": 1091
     },
     {
       "epoch": 19.85,
-      "learning_rate": 0.0003691111111111111,
-      "loss": 0.4421,
+      "learning_rate": 0.0004187586206896552,
+      "loss": 0.521,
       "step": 1092
     },
     {
       "epoch": 19.87,
-      "learning_rate": 0.00036888888888888887,
-      "loss": 0.5108,
+      "learning_rate": 0.0004186206896551724,
+      "loss": 0.582,
       "step": 1093
     },
     {
       "epoch": 19.89,
-      "learning_rate": 0.00036866666666666667,
-      "loss": 0.4182,
+      "learning_rate": 0.0004184827586206897,
+      "loss": 0.4986,
       "step": 1094
     },
     {
       "epoch": 19.91,
-      "learning_rate": 0.00036844444444444446,
-      "loss": 0.5699,
+      "learning_rate": 0.0004183448275862069,
+      "loss": 0.5647,
       "step": 1095
     },
     {
       "epoch": 19.93,
-      "learning_rate": 0.00036822222222222226,
-      "loss": 0.5403,
+      "learning_rate": 0.00041820689655172415,
+      "loss": 0.6148,
       "step": 1096
     },
     {
       "epoch": 19.95,
-      "learning_rate": 0.000368,
-      "loss": 0.4065,
+      "learning_rate": 0.00041806896551724136,
+      "loss": 0.4507,
       "step": 1097
     },
     {
       "epoch": 19.96,
-      "learning_rate": 0.00036777777777777774,
-      "loss": 0.4928,
+      "learning_rate": 0.00041793103448275863,
+      "loss": 0.5244,
       "step": 1098
     },
     {
       "epoch": 19.98,
-      "learning_rate": 0.0003675555555555556,
-      "loss": 0.5326,
+      "learning_rate": 0.0004177931034482759,
+      "loss": 0.5439,
       "step": 1099
     },
     {
       "epoch": 20.0,
-      "learning_rate": 0.00036733333333333334,
-      "loss": 0.4815,
+      "learning_rate": 0.00041765517241379316,
+      "loss": 0.5507,
       "step": 1100
     },
     {
       "epoch": 20.02,
-      "learning_rate": 0.00036711111111111113,
-      "loss": 0.4911,
+      "learning_rate": 0.0004175172413793104,
+      "loss": 0.5894,
       "step": 1101
     },
     {
       "epoch": 20.04,
-      "learning_rate": 0.0003668888888888889,
-      "loss": 0.4394,
+      "learning_rate": 0.0004173793103448276,
+      "loss": 0.4877,
       "step": 1102
     },
     {
       "epoch": 20.05,
-      "learning_rate": 0.00036666666666666667,
-      "loss": 0.4046,
+      "learning_rate": 0.00041724137931034485,
+      "loss": 0.4787,
       "step": 1103
     },
     {
       "epoch": 20.07,
-      "learning_rate": 0.00036644444444444447,
-      "loss": 0.4674,
+      "learning_rate": 0.00041710344827586206,
+      "loss": 0.5647,
       "step": 1104
     },
     {
       "epoch": 20.09,
-      "learning_rate": 0.0003662222222222222,
-      "loss": 0.4655,
+      "learning_rate": 0.00041696551724137933,
+      "loss": 0.49,
       "step": 1105
     },
     {
       "epoch": 20.11,
-      "learning_rate": 0.000366,
-      "loss": 0.4087,
+      "learning_rate": 0.00041682758620689654,
+      "loss": 0.5022,
       "step": 1106
     },
     {
       "epoch": 20.13,
-      "learning_rate": 0.00036577777777777775,
-      "loss": 0.5156,
+      "learning_rate": 0.0004166896551724138,
+      "loss": 0.5834,
       "step": 1107
     },
     {
       "epoch": 20.15,
-      "learning_rate": 0.0003655555555555556,
-      "loss": 0.5056,
+      "learning_rate": 0.000416551724137931,
+      "loss": 0.5313,
       "step": 1108
     },
     {
       "epoch": 20.16,
-      "learning_rate": 0.00036533333333333334,
-      "loss": 0.5199,
+      "learning_rate": 0.0004164137931034483,
+      "loss": 0.5977,
       "step": 1109
     },
     {
       "epoch": 20.18,
-      "learning_rate": 0.00036511111111111114,
-      "loss": 0.4372,
+      "learning_rate": 0.0004162758620689655,
+      "loss": 0.5453,
       "step": 1110
     },
     {
       "epoch": 20.2,
-      "learning_rate": 0.0003648888888888889,
-      "loss": 0.4585,
+      "learning_rate": 0.00041613793103448276,
+      "loss": 0.5128,
       "step": 1111
     },
     {
       "epoch": 20.22,
-      "learning_rate": 0.0003646666666666666,
-      "loss": 0.4315,
+      "learning_rate": 0.000416,
+      "loss": 0.4704,
       "step": 1112
     },
     {
       "epoch": 20.24,
-      "learning_rate": 0.00036444444444444447,
-      "loss": 0.4665,
+      "learning_rate": 0.00041586206896551724,
+      "loss": 0.5382,
       "step": 1113
     },
     {
       "epoch": 20.25,
-      "learning_rate": 0.0003642222222222222,
-      "loss": 0.4536,
+      "learning_rate": 0.00041572413793103445,
+      "loss": 0.5255,
       "step": 1114
     },
     {
       "epoch": 20.27,
-      "learning_rate": 0.000364,
-      "loss": 0.4806,
+      "learning_rate": 0.0004155862068965517,
+      "loss": 0.501,
       "step": 1115
     },
     {
       "epoch": 20.29,
-      "learning_rate": 0.00036377777777777775,
-      "loss": 0.4873,
+      "learning_rate": 0.000415448275862069,
+      "loss": 0.5875,
       "step": 1116
     },
     {
       "epoch": 20.31,
-      "learning_rate": 0.0003635555555555556,
-      "loss": 0.4786,
+      "learning_rate": 0.00041531034482758625,
+      "loss": 0.5646,
       "step": 1117
     },
     {
       "epoch": 20.33,
-      "learning_rate": 0.00036333333333333335,
-      "loss": 0.5094,
+      "learning_rate": 0.00041517241379310346,
+      "loss": 0.5796,
       "step": 1118
     },
     {
       "epoch": 20.35,
-      "learning_rate": 0.0003631111111111111,
-      "loss": 0.4452,
+      "learning_rate": 0.00041503448275862073,
+      "loss": 0.5762,
       "step": 1119
     },
     {
       "epoch": 20.36,
-      "learning_rate": 0.0003628888888888889,
-      "loss": 0.4147,
+      "learning_rate": 0.00041489655172413794,
+      "loss": 0.4977,
       "step": 1120
     },
     {
       "epoch": 20.38,
-      "learning_rate": 0.0003626666666666667,
-      "loss": 0.4156,
+      "learning_rate": 0.0004147586206896552,
+      "loss": 0.5007,
       "step": 1121
     },
     {
       "epoch": 20.4,
-      "learning_rate": 0.0003624444444444445,
-      "loss": 0.3956,
+      "learning_rate": 0.0004146206896551724,
+      "loss": 0.4878,
       "step": 1122
     },
     {
       "epoch": 20.42,
-      "learning_rate": 0.0003622222222222222,
-      "loss": 0.5021,
+      "learning_rate": 0.0004144827586206897,
+      "loss": 0.5817,
       "step": 1123
     },
     {
       "epoch": 20.44,
-      "learning_rate": 0.000362,
-      "loss": 0.4532,
+      "learning_rate": 0.0004143448275862069,
+      "loss": 0.4926,
       "step": 1124
     },
     {
       "epoch": 20.45,
-      "learning_rate": 0.00036177777777777776,
-      "loss": 0.4266,
+      "learning_rate": 0.00041420689655172416,
+      "loss": 0.5488,
       "step": 1125
     },
     {
       "epoch": 20.47,
-      "learning_rate": 0.00036155555555555556,
-      "loss": 0.452,
+      "learning_rate": 0.0004140689655172414,
+      "loss": 0.4953,
       "step": 1126
     },
     {
       "epoch": 20.49,
-      "learning_rate": 0.00036133333333333335,
-      "loss": 0.4651,
+      "learning_rate": 0.00041393103448275864,
+      "loss": 0.5204,
       "step": 1127
     },
     {
       "epoch": 20.51,
-      "learning_rate": 0.0003611111111111111,
-      "loss": 0.4432,
+      "learning_rate": 0.00041379310344827585,
+      "loss": 0.532,
       "step": 1128
     },
     {
       "epoch": 20.53,
-      "learning_rate": 0.0003608888888888889,
-      "loss": 0.4777,
+      "learning_rate": 0.0004136551724137931,
+      "loss": 0.4874,
       "step": 1129
     },
     {
       "epoch": 20.55,
-      "learning_rate": 0.0003606666666666667,
-      "loss": 0.4296,
+      "learning_rate": 0.00041351724137931033,
+      "loss": 0.536,
       "step": 1130
     },
     {
       "epoch": 20.56,
-      "learning_rate": 0.0003604444444444445,
-      "loss": 0.4761,
+      "learning_rate": 0.0004133793103448276,
+      "loss": 0.5511,
       "step": 1131
     },
     {
       "epoch": 20.58,
-      "learning_rate": 0.0003602222222222222,
-      "loss": 0.483,
+      "learning_rate": 0.0004132413793103448,
+      "loss": 0.5303,
       "step": 1132
     },
     {
       "epoch": 20.6,
-      "learning_rate": 0.00035999999999999997,
-      "loss": 0.5602,
+      "learning_rate": 0.00041310344827586213,
+      "loss": 0.6272,
       "step": 1133
     },
     {
       "epoch": 20.62,
-      "learning_rate": 0.00035977777777777776,
-      "loss": 0.5059,
+      "learning_rate": 0.00041296551724137934,
+      "loss": 0.5749,
       "step": 1134
     },
     {
       "epoch": 20.64,
-      "learning_rate": 0.00035955555555555556,
-      "loss": 0.4605,
+      "learning_rate": 0.00041282758620689655,
+      "loss": 0.5201,
       "step": 1135
     },
     {
       "epoch": 20.65,
-      "learning_rate": 0.00035933333333333336,
-      "loss": 0.4876,
+      "learning_rate": 0.0004126896551724138,
+      "loss": 0.4935,
       "step": 1136
     },
     {
       "epoch": 20.67,
-      "learning_rate": 0.0003591111111111111,
-      "loss": 0.5087,
+      "learning_rate": 0.00041255172413793103,
+      "loss": 0.5972,
       "step": 1137
     },
     {
       "epoch": 20.69,
-      "learning_rate": 0.0003588888888888889,
-      "loss": 0.4759,
+      "learning_rate": 0.0004124137931034483,
+      "loss": 0.5502,
       "step": 1138
     },
     {
       "epoch": 20.71,
-      "learning_rate": 0.0003586666666666667,
-      "loss": 0.4742,
+      "learning_rate": 0.0004122758620689655,
+      "loss": 0.5323,
       "step": 1139
     },
     {
       "epoch": 20.73,
-      "learning_rate": 0.00035844444444444444,
-      "loss": 0.4653,
+      "learning_rate": 0.0004121379310344828,
+      "loss": 0.5229,
       "step": 1140
     },
     {
       "epoch": 20.75,
-      "learning_rate": 0.00035822222222222223,
-      "loss": 0.4744,
+      "learning_rate": 0.000412,
+      "loss": 0.5275,
       "step": 1141
     },
     {
       "epoch": 20.76,
-      "learning_rate": 0.000358,
-      "loss": 0.47,
+      "learning_rate": 0.00041186206896551725,
+      "loss": 0.5305,
       "step": 1142
     },
     {
       "epoch": 20.78,
-      "learning_rate": 0.00035777777777777777,
-      "loss": 0.4011,
+      "learning_rate": 0.00041172413793103446,
+      "loss": 0.5031,
       "step": 1143
     },
     {
       "epoch": 20.8,
-      "learning_rate": 0.00035755555555555557,
-      "loss": 0.533,
+      "learning_rate": 0.00041158620689655173,
+      "loss": 0.5797,
       "step": 1144
     },
     {
       "epoch": 20.82,
-      "learning_rate": 0.00035733333333333336,
-      "loss": 0.525,
+      "learning_rate": 0.00041144827586206894,
+      "loss": 0.5216,
       "step": 1145
     },
     {
       "epoch": 20.84,
-      "learning_rate": 0.0003571111111111111,
-      "loss": 0.5249,
+      "learning_rate": 0.0004113103448275862,
+      "loss": 0.5321,
       "step": 1146
     },
     {
       "epoch": 20.85,
-      "learning_rate": 0.00035688888888888885,
-      "loss": 0.4474,
+      "learning_rate": 0.0004111724137931034,
+      "loss": 0.4958,
       "step": 1147
     },
     {
       "epoch": 20.87,
-      "learning_rate": 0.0003566666666666667,
-      "loss": 0.4719,
+      "learning_rate": 0.0004110344827586207,
+      "loss": 0.5393,
       "step": 1148
     },
     {
       "epoch": 20.89,
-      "learning_rate": 0.00035644444444444444,
-      "loss": 0.5247,
+      "learning_rate": 0.0004108965517241379,
+      "loss": 0.5417,
       "step": 1149
     },
     {
       "epoch": 20.91,
-      "learning_rate": 0.00035622222222222224,
-      "loss": 0.4718,
+      "learning_rate": 0.0004107586206896552,
+      "loss": 0.5331,
       "step": 1150
     },
     {
       "epoch": 20.93,
-      "learning_rate": 0.000356,
-      "loss": 0.4837,
+      "learning_rate": 0.00041062068965517243,
+      "loss": 0.5355,
       "step": 1151
     },
     {
       "epoch": 20.95,
-      "learning_rate": 0.00035577777777777783,
-      "loss": 0.5016,
+      "learning_rate": 0.0004104827586206897,
+      "loss": 0.5461,
       "step": 1152
     },
     {
       "epoch": 20.96,
-      "learning_rate": 0.00035555555555555557,
-      "loss": 0.4625,
+      "learning_rate": 0.0004103448275862069,
+      "loss": 0.5036,
       "step": 1153
     },
     {
       "epoch": 20.98,
-      "learning_rate": 0.0003553333333333333,
-      "loss": 0.4693,
+      "learning_rate": 0.0004102068965517242,
+      "loss": 0.5037,
       "step": 1154
     },
     {
       "epoch": 21.0,
-      "learning_rate": 0.0003551111111111111,
-      "loss": 0.4338,
+      "learning_rate": 0.0004100689655172414,
+      "loss": 0.4756,
       "step": 1155
     },
     {
       "epoch": 21.02,
-      "learning_rate": 0.00035488888888888885,
-      "loss": 0.4439,
+      "learning_rate": 0.00040993103448275865,
+      "loss": 0.501,
       "step": 1156
     },
     {
       "epoch": 21.04,
-      "learning_rate": 0.0003546666666666667,
-      "loss": 0.4542,
+      "learning_rate": 0.00040979310344827586,
+      "loss": 0.4788,
       "step": 1157
     },
     {
       "epoch": 21.05,
-      "learning_rate": 0.00035444444444444445,
-      "loss": 0.4063,
+      "learning_rate": 0.00040965517241379313,
+      "loss": 0.4502,
       "step": 1158
     },
     {
       "epoch": 21.07,
-      "learning_rate": 0.00035422222222222224,
-      "loss": 0.4267,
+      "learning_rate": 0.00040951724137931034,
+      "loss": 0.4709,
       "step": 1159
     },
     {
       "epoch": 21.09,
-      "learning_rate": 0.000354,
-      "loss": 0.507,
+      "learning_rate": 0.0004093793103448276,
+      "loss": 0.5607,
       "step": 1160
     },
     {
       "epoch": 21.11,
-      "learning_rate": 0.00035377777777777784,
-      "loss": 0.4305,
+      "learning_rate": 0.0004092413793103448,
+      "loss": 0.4833,
       "step": 1161
     },
     {
       "epoch": 21.13,
-      "learning_rate": 0.0003535555555555556,
-      "loss": 0.4152,
+      "learning_rate": 0.0004091034482758621,
+      "loss": 0.5196,
       "step": 1162
     },
     {
       "epoch": 21.15,
-      "learning_rate": 0.0003533333333333333,
-      "loss": 0.388,
+      "learning_rate": 0.0004089655172413793,
+      "loss": 0.4413,
       "step": 1163
     },
     {
       "epoch": 21.16,
-      "learning_rate": 0.0003531111111111111,
-      "loss": 0.4326,
+      "learning_rate": 0.00040882758620689656,
+      "loss": 0.4714,
       "step": 1164
     },
     {
       "epoch": 21.18,
-      "learning_rate": 0.00035288888888888886,
-      "loss": 0.4594,
+      "learning_rate": 0.0004086896551724138,
+      "loss": 0.5406,
       "step": 1165
     },
     {
       "epoch": 21.2,
-      "learning_rate": 0.0003526666666666667,
-      "loss": 0.4416,
+      "learning_rate": 0.0004085517241379311,
+      "loss": 0.4918,
       "step": 1166
     },
     {
       "epoch": 21.22,
-      "learning_rate": 0.00035244444444444445,
-      "loss": 0.5016,
+      "learning_rate": 0.0004084137931034483,
+      "loss": 0.5404,
       "step": 1167
     },
     {
       "epoch": 21.24,
-      "learning_rate": 0.00035222222222222225,
-      "loss": 0.4614,
+      "learning_rate": 0.0004082758620689656,
+      "loss": 0.5304,
       "step": 1168
     },
     {
       "epoch": 21.25,
-      "learning_rate": 0.000352,
-      "loss": 0.5384,
+      "learning_rate": 0.0004081379310344828,
+      "loss": 0.5601,
       "step": 1169
     },
     {
       "epoch": 21.27,
-      "learning_rate": 0.0003517777777777778,
-      "loss": 0.4494,
+      "learning_rate": 0.000408,
+      "loss": 0.506,
       "step": 1170
     },
     {
       "epoch": 21.29,
-      "learning_rate": 0.0003515555555555556,
-      "loss": 0.4324,
+      "learning_rate": 0.00040786206896551726,
+      "loss": 0.4726,
       "step": 1171
     },
     {
       "epoch": 21.31,
-      "learning_rate": 0.0003513333333333333,
-      "loss": 0.4543,
+      "learning_rate": 0.0004077241379310345,
+      "loss": 0.5143,
       "step": 1172
     },
     {
       "epoch": 21.33,
-      "learning_rate": 0.0003511111111111111,
-      "loss": 0.4924,
+      "learning_rate": 0.00040758620689655174,
+      "loss": 0.5256,
       "step": 1173
     },
     {
       "epoch": 21.35,
-      "learning_rate": 0.00035088888888888886,
-      "loss": 0.5228,
+      "learning_rate": 0.00040744827586206895,
+      "loss": 0.5844,
       "step": 1174
     },
     {
       "epoch": 21.36,
-      "learning_rate": 0.0003506666666666667,
-      "loss": 0.4994,
+      "learning_rate": 0.0004073103448275862,
+      "loss": 0.5685,
       "step": 1175
     },
     {
       "epoch": 21.38,
-      "learning_rate": 0.00035044444444444446,
-      "loss": 0.4533,
+      "learning_rate": 0.00040717241379310343,
+      "loss": 0.5008,
       "step": 1176
     },
     {
       "epoch": 21.4,
-      "learning_rate": 0.0003502222222222222,
-      "loss": 0.4793,
+      "learning_rate": 0.0004070344827586207,
+      "loss": 0.5341,
       "step": 1177
     },
     {
       "epoch": 21.42,
-      "learning_rate": 0.00035,
-      "loss": 0.5031,
+      "learning_rate": 0.0004068965517241379,
+      "loss": 0.5564,
       "step": 1178
     },
     {
       "epoch": 21.44,
-      "learning_rate": 0.0003497777777777778,
-      "loss": 0.457,
+      "learning_rate": 0.0004067586206896552,
+      "loss": 0.5455,
       "step": 1179
     },
     {
       "epoch": 21.45,
-      "learning_rate": 0.0003495555555555556,
-      "loss": 0.4829,
+      "learning_rate": 0.0004066206896551724,
+      "loss": 0.5286,
       "step": 1180
     },
     {
       "epoch": 21.47,
-      "learning_rate": 0.00034933333333333333,
-      "loss": 0.4977,
+      "learning_rate": 0.00040648275862068965,
+      "loss": 0.5206,
       "step": 1181
     },
     {
       "epoch": 21.49,
-      "learning_rate": 0.00034911111111111113,
-      "loss": 0.4972,
+      "learning_rate": 0.00040634482758620686,
+      "loss": 0.5422,
       "step": 1182
     },
     {
       "epoch": 21.51,
-      "learning_rate": 0.0003488888888888889,
-      "loss": 0.5017,
+      "learning_rate": 0.0004062068965517242,
+      "loss": 0.5505,
       "step": 1183
     },
     {
       "epoch": 21.53,
-      "learning_rate": 0.00034866666666666667,
-      "loss": 0.4633,
+      "learning_rate": 0.0004060689655172414,
+      "loss": 0.5386,
       "step": 1184
     },
     {
       "epoch": 21.55,
-      "learning_rate": 0.00034844444444444446,
-      "loss": 0.3661,
+      "learning_rate": 0.00040593103448275866,
+      "loss": 0.4141,
       "step": 1185
     },
     {
       "epoch": 21.56,
-      "learning_rate": 0.0003482222222222222,
-      "loss": 0.5226,
+      "learning_rate": 0.0004057931034482759,
+      "loss": 0.5804,
       "step": 1186
     },
     {
       "epoch": 21.58,
-      "learning_rate": 0.000348,
-      "loss": 0.4927,
+      "learning_rate": 0.00040565517241379314,
+      "loss": 0.5341,
       "step": 1187
     },
     {
       "epoch": 21.6,
-      "learning_rate": 0.0003477777777777778,
-      "loss": 0.4527,
+      "learning_rate": 0.00040551724137931035,
+      "loss": 0.463,
       "step": 1188
     },
     {
       "epoch": 21.62,
-      "learning_rate": 0.0003475555555555556,
-      "loss": 0.4882,
+      "learning_rate": 0.0004053793103448276,
+      "loss": 0.5053,
       "step": 1189
     },
     {
       "epoch": 21.64,
-      "learning_rate": 0.00034733333333333334,
-      "loss": 0.4576,
+      "learning_rate": 0.00040524137931034483,
+      "loss": 0.5262,
       "step": 1190
     },
     {
       "epoch": 21.65,
-      "learning_rate": 0.0003471111111111111,
-      "loss": 0.4641,
+      "learning_rate": 0.0004051034482758621,
+      "loss": 0.4951,
       "step": 1191
     },
     {
       "epoch": 21.67,
-      "learning_rate": 0.00034688888888888893,
-      "loss": 0.4332,
+      "learning_rate": 0.0004049655172413793,
+      "loss": 0.5152,
       "step": 1192
     },
     {
       "epoch": 21.69,
-      "learning_rate": 0.00034666666666666667,
-      "loss": 0.4282,
+      "learning_rate": 0.0004048275862068966,
+      "loss": 0.466,
       "step": 1193
     },
     {
       "epoch": 21.71,
-      "learning_rate": 0.00034644444444444447,
-      "loss": 0.4349,
+      "learning_rate": 0.0004046896551724138,
+      "loss": 0.4735,
       "step": 1194
     },
     {
       "epoch": 21.73,
-      "learning_rate": 0.0003462222222222222,
-      "loss": 0.4382,
+      "learning_rate": 0.00040455172413793105,
+      "loss": 0.4712,
       "step": 1195
     },
     {
       "epoch": 21.75,
-      "learning_rate": 0.000346,
-      "loss": 0.4454,
+      "learning_rate": 0.00040441379310344826,
+      "loss": 0.5215,
       "step": 1196
     },
     {
       "epoch": 21.76,
-      "learning_rate": 0.0003457777777777778,
-      "loss": 0.5056,
+      "learning_rate": 0.00040427586206896553,
+      "loss": 0.5595,
       "step": 1197
     },
     {
       "epoch": 21.78,
-      "learning_rate": 0.00034555555555555555,
-      "loss": 0.4052,
+      "learning_rate": 0.00040413793103448274,
+      "loss": 0.491,
       "step": 1198
     },
     {
       "epoch": 21.8,
-      "learning_rate": 0.00034533333333333334,
-      "loss": 0.4607,
+      "learning_rate": 0.000404,
+      "loss": 0.5668,
       "step": 1199
     },
     {
       "epoch": 21.82,
-      "learning_rate": 0.0003451111111111111,
-      "loss": 0.5177,
+      "learning_rate": 0.0004038620689655173,
+      "loss": 0.5893,
       "step": 1200
     },
     {
       "epoch": 21.84,
-      "learning_rate": 0.00034488888888888894,
-      "loss": 0.4062,
+      "learning_rate": 0.00040372413793103454,
+      "loss": 0.4538,
       "step": 1201
     },
     {
       "epoch": 21.85,
-      "learning_rate": 0.0003446666666666667,
-      "loss": 0.4554,
+      "learning_rate": 0.00040358620689655175,
+      "loss": 0.4997,
       "step": 1202
     },
     {
       "epoch": 21.87,
-      "learning_rate": 0.0003444444444444445,
-      "loss": 0.5176,
+      "learning_rate": 0.00040344827586206896,
+      "loss": 0.5917,
       "step": 1203
     },
     {
       "epoch": 21.89,
-      "learning_rate": 0.0003442222222222222,
-      "loss": 0.4775,
+      "learning_rate": 0.00040331034482758623,
+      "loss": 0.5198,
       "step": 1204
     },
     {
       "epoch": 21.91,
-      "learning_rate": 0.00034399999999999996,
-      "loss": 0.5191,
+      "learning_rate": 0.00040317241379310344,
+      "loss": 0.5563,
       "step": 1205
     },
     {
       "epoch": 21.93,
-      "learning_rate": 0.0003437777777777778,
-      "loss": 0.5193,
+      "learning_rate": 0.0004030344827586207,
+      "loss": 0.5563,
       "step": 1206
     },
     {
       "epoch": 21.95,
-      "learning_rate": 0.00034355555555555555,
-      "loss": 0.4973,
+      "learning_rate": 0.0004028965517241379,
+      "loss": 0.5209,
       "step": 1207
     },
     {
       "epoch": 21.96,
-      "learning_rate": 0.00034333333333333335,
-      "loss": 0.4255,
+      "learning_rate": 0.0004027586206896552,
+      "loss": 0.4963,
       "step": 1208
     },
     {
       "epoch": 21.98,
-      "learning_rate": 0.0003431111111111111,
-      "loss": 0.5322,
+      "learning_rate": 0.0004026206896551724,
+      "loss": 0.5732,
       "step": 1209
     },
     {
       "epoch": 22.0,
-      "learning_rate": 0.00034288888888888894,
-      "loss": 0.4951,
+      "learning_rate": 0.00040248275862068966,
+      "loss": 0.5404,
       "step": 1210
     },
     {
       "epoch": 22.02,
-      "learning_rate": 0.0003426666666666667,
-      "loss": 0.5144,
+      "learning_rate": 0.0004023448275862069,
+      "loss": 0.5726,
       "step": 1211
     },
     {
       "epoch": 22.04,
-      "learning_rate": 0.0003424444444444444,
-      "loss": 0.4034,
+      "learning_rate": 0.00040220689655172414,
+      "loss": 0.4541,
       "step": 1212
     },
     {
       "epoch": 22.05,
-      "learning_rate": 0.0003422222222222222,
-      "loss": 0.4762,
+      "learning_rate": 0.00040206896551724135,
+      "loss": 0.5626,
       "step": 1213
     },
     {
       "epoch": 22.07,
-      "learning_rate": 0.000342,
-      "loss": 0.4259,
+      "learning_rate": 0.0004019310344827586,
+      "loss": 0.4904,
       "step": 1214
     },
     {
       "epoch": 22.09,
-      "learning_rate": 0.0003417777777777778,
-      "loss": 0.4817,
+      "learning_rate": 0.00040179310344827583,
+      "loss": 0.5129,
       "step": 1215
     },
     {
       "epoch": 22.11,
-      "learning_rate": 0.00034155555555555556,
-      "loss": 0.4129,
+      "learning_rate": 0.0004016551724137931,
+      "loss": 0.4477,
       "step": 1216
     },
     {
       "epoch": 22.13,
-      "learning_rate": 0.00034133333333333335,
-      "loss": 0.4539,
+      "learning_rate": 0.00040151724137931036,
+      "loss": 0.5024,
       "step": 1217
     },
     {
       "epoch": 22.15,
-      "learning_rate": 0.0003411111111111111,
-      "loss": 0.4899,
+      "learning_rate": 0.00040137931034482763,
+      "loss": 0.5229,
       "step": 1218
     },
     {
       "epoch": 22.16,
-      "learning_rate": 0.0003408888888888889,
-      "loss": 0.4624,
+      "learning_rate": 0.00040124137931034484,
+      "loss": 0.5138,
       "step": 1219
     },
     {
       "epoch": 22.18,
-      "learning_rate": 0.0003406666666666667,
-      "loss": 0.4065,
+      "learning_rate": 0.0004011034482758621,
+      "loss": 0.5155,
       "step": 1220
     },
     {
       "epoch": 22.2,
-      "learning_rate": 0.00034044444444444443,
-      "loss": 0.5008,
+      "learning_rate": 0.0004009655172413793,
+      "loss": 0.5727,
       "step": 1221
     },
     {
       "epoch": 22.22,
-      "learning_rate": 0.00034022222222222223,
-      "loss": 0.5189,
+      "learning_rate": 0.0004008275862068966,
+      "loss": 0.5983,
       "step": 1222
     },
     {
       "epoch": 22.24,
-      "learning_rate": 0.00034,
-      "loss": 0.4646,
+      "learning_rate": 0.0004006896551724138,
+      "loss": 0.5156,
       "step": 1223
     },
     {
       "epoch": 22.25,
-      "learning_rate": 0.0003397777777777778,
-      "loss": 0.4351,
+      "learning_rate": 0.00040055172413793106,
+      "loss": 0.4375,
       "step": 1224
     },
     {
       "epoch": 22.27,
-      "learning_rate": 0.00033955555555555556,
-      "loss": 0.4617,
+      "learning_rate": 0.0004004137931034483,
+      "loss": 0.5582,
       "step": 1225
     },
     {
       "epoch": 22.29,
-      "learning_rate": 0.0003393333333333333,
-      "loss": 0.4991,
+      "learning_rate": 0.00040027586206896554,
+      "loss": 0.5095,
       "step": 1226
     },
     {
       "epoch": 22.31,
-      "learning_rate": 0.0003391111111111111,
-      "loss": 0.5138,
+      "learning_rate": 0.00040013793103448275,
+      "loss": 0.5495,
       "step": 1227
     },
     {
       "epoch": 22.33,
-      "learning_rate": 0.0003388888888888889,
-      "loss": 0.4835,
+      "learning_rate": 0.0004,
+      "loss": 0.5048,
       "step": 1228
     },
     {
       "epoch": 22.35,
-      "learning_rate": 0.0003386666666666667,
-      "loss": 0.4848,
+      "learning_rate": 0.00039986206896551723,
+      "loss": 0.5907,
       "step": 1229
     },
     {
       "epoch": 22.36,
-      "learning_rate": 0.00033844444444444444,
-      "loss": 0.4352,
+      "learning_rate": 0.0003997241379310345,
+      "loss": 0.4948,
       "step": 1230
     },
     {
       "epoch": 22.38,
-      "learning_rate": 0.00033822222222222223,
-      "loss": 0.467,
+      "learning_rate": 0.0003995862068965517,
+      "loss": 0.5387,
       "step": 1231
     },
     {
       "epoch": 22.4,
-      "learning_rate": 0.00033800000000000003,
-      "loss": 0.5013,
+      "learning_rate": 0.000399448275862069,
+      "loss": 0.5338,
       "step": 1232
     },
     {
       "epoch": 22.42,
-      "learning_rate": 0.00033777777777777777,
-      "loss": 0.4979,
+      "learning_rate": 0.0003993103448275862,
+      "loss": 0.5853,
       "step": 1233
     },
     {
       "epoch": 22.44,
-      "learning_rate": 0.00033755555555555557,
-      "loss": 0.5048,
+      "learning_rate": 0.0003991724137931035,
+      "loss": 0.5536,
       "step": 1234
     },
     {
       "epoch": 22.45,
-      "learning_rate": 0.0003373333333333333,
-      "loss": 0.3969,
+      "learning_rate": 0.0003990344827586207,
+      "loss": 0.4713,
       "step": 1235
     },
     {
       "epoch": 22.47,
-      "learning_rate": 0.0003371111111111111,
-      "loss": 0.5369,
+      "learning_rate": 0.00039889655172413793,
+      "loss": 0.5651,
       "step": 1236
     },
     {
       "epoch": 22.49,
-      "learning_rate": 0.0003368888888888889,
-      "loss": 0.3876,
+      "learning_rate": 0.0003987586206896552,
+      "loss": 0.4469,
       "step": 1237
     },
     {
       "epoch": 22.51,
-      "learning_rate": 0.0003366666666666667,
-      "loss": 0.4798,
+      "learning_rate": 0.0003986206896551724,
+      "loss": 0.5633,
       "step": 1238
     },
     {
       "epoch": 22.53,
-      "learning_rate": 0.00033644444444444444,
-      "loss": 0.3773,
+      "learning_rate": 0.0003984827586206897,
+      "loss": 0.4576,
       "step": 1239
     },
     {
       "epoch": 22.55,
-      "learning_rate": 0.0003362222222222222,
-      "loss": 0.4403,
+      "learning_rate": 0.0003983448275862069,
+      "loss": 0.5243,
       "step": 1240
     },
     {
       "epoch": 22.56,
-      "learning_rate": 0.00033600000000000004,
-      "loss": 0.5163,
+      "learning_rate": 0.00039820689655172415,
+      "loss": 0.6213,
       "step": 1241
     },
     {
       "epoch": 22.58,
-      "learning_rate": 0.0003357777777777778,
-      "loss": 0.436,
+      "learning_rate": 0.00039806896551724136,
+      "loss": 0.5226,
       "step": 1242
     },
     {
       "epoch": 22.6,
-      "learning_rate": 0.0003355555555555556,
-      "loss": 0.4033,
+      "learning_rate": 0.00039793103448275863,
+      "loss": 0.5302,
       "step": 1243
     },
     {
       "epoch": 22.62,
-      "learning_rate": 0.0003353333333333333,
-      "loss": 0.476,
+      "learning_rate": 0.00039779310344827584,
+      "loss": 0.566,
       "step": 1244
     },
     {
       "epoch": 22.64,
-      "learning_rate": 0.00033511111111111117,
-      "loss": 0.419,
+      "learning_rate": 0.0003976551724137931,
+      "loss": 0.5202,
       "step": 1245
     },
     {
       "epoch": 22.65,
-      "learning_rate": 0.0003348888888888889,
-      "loss": 0.536,
+      "learning_rate": 0.0003975172413793103,
+      "loss": 0.5584,
       "step": 1246
     },
     {
       "epoch": 22.67,
-      "learning_rate": 0.00033466666666666665,
-      "loss": 0.4942,
+      "learning_rate": 0.0003973793103448276,
+      "loss": 0.5918,
       "step": 1247
     },
     {
       "epoch": 22.69,
-      "learning_rate": 0.00033444444444444445,
-      "loss": 0.4704,
+      "learning_rate": 0.0003972413793103448,
+      "loss": 0.5105,
       "step": 1248
     },
     {
       "epoch": 22.71,
-      "learning_rate": 0.0003342222222222222,
-      "loss": 0.4348,
+      "learning_rate": 0.00039710344827586206,
+      "loss": 0.4694,
       "step": 1249
     },
     {
       "epoch": 22.73,
-      "learning_rate": 0.00033400000000000004,
-      "loss": 0.4464,
+      "learning_rate": 0.00039696551724137933,
+      "loss": 0.4979,
       "step": 1250
     },
     {
       "epoch": 22.75,
-      "learning_rate": 0.0003337777777777778,
-      "loss": 0.366,
+      "learning_rate": 0.0003968275862068966,
+      "loss": 0.4452,
       "step": 1251
     },
     {
       "epoch": 22.76,
-      "learning_rate": 0.0003335555555555556,
-      "loss": 0.4726,
+      "learning_rate": 0.0003966896551724138,
+      "loss": 0.5654,
       "step": 1252
     },
     {
       "epoch": 22.78,
-      "learning_rate": 0.0003333333333333333,
-      "loss": 0.4565,
+      "learning_rate": 0.0003965517241379311,
+      "loss": 0.4919,
       "step": 1253
     },
     {
       "epoch": 22.8,
-      "learning_rate": 0.0003331111111111111,
-      "loss": 0.4151,
+      "learning_rate": 0.0003964137931034483,
+      "loss": 0.4282,
       "step": 1254
     },
     {
       "epoch": 22.82,
-      "learning_rate": 0.0003328888888888889,
-      "loss": 0.4462,
+      "learning_rate": 0.00039627586206896555,
+      "loss": 0.524,
       "step": 1255
     },
     {
       "epoch": 22.84,
-      "learning_rate": 0.00033266666666666666,
-      "loss": 0.4428,
+      "learning_rate": 0.00039613793103448276,
+      "loss": 0.4854,
       "step": 1256
     },
     {
       "epoch": 22.85,
-      "learning_rate": 0.00033244444444444445,
-      "loss": 0.5043,
+      "learning_rate": 0.00039600000000000003,
+      "loss": 0.6051,
       "step": 1257
     },
     {
       "epoch": 22.87,
-      "learning_rate": 0.0003322222222222222,
-      "loss": 0.4251,
+      "learning_rate": 0.00039586206896551724,
+      "loss": 0.489,
       "step": 1258
     },
     {
       "epoch": 22.89,
-      "learning_rate": 0.00033200000000000005,
-      "loss": 0.4716,
+      "learning_rate": 0.0003957241379310345,
+      "loss": 0.5208,
       "step": 1259
     },
     {
       "epoch": 22.91,
-      "learning_rate": 0.0003317777777777778,
-      "loss": 0.4094,
+      "learning_rate": 0.0003955862068965517,
+      "loss": 0.4452,
       "step": 1260
     },
     {
       "epoch": 22.93,
-      "learning_rate": 0.00033155555555555553,
-      "loss": 0.4834,
+      "learning_rate": 0.000395448275862069,
+      "loss": 0.5232,
       "step": 1261
     },
     {
       "epoch": 22.95,
-      "learning_rate": 0.00033133333333333333,
-      "loss": 0.4415,
+      "learning_rate": 0.0003953103448275862,
+      "loss": 0.5064,
       "step": 1262
     },
     {
       "epoch": 22.96,
-      "learning_rate": 0.0003311111111111111,
-      "loss": 0.4463,
+      "learning_rate": 0.00039517241379310346,
+      "loss": 0.4756,
       "step": 1263
     },
     {
       "epoch": 22.98,
-      "learning_rate": 0.0003308888888888889,
-      "loss": 0.4748,
+      "learning_rate": 0.0003950344827586207,
+      "loss": 0.5381,
       "step": 1264
     },
     {
       "epoch": 23.0,
-      "learning_rate": 0.00033066666666666666,
-      "loss": 0.4726,
+      "learning_rate": 0.00039489655172413794,
+      "loss": 0.5377,
       "step": 1265
     },
     {
       "epoch": 23.02,
-      "learning_rate": 0.00033044444444444446,
-      "loss": 0.4308,
+      "learning_rate": 0.00039475862068965515,
+      "loss": 0.4884,
       "step": 1266
     },
     {
       "epoch": 23.04,
-      "learning_rate": 0.0003302222222222222,
-      "loss": 0.4256,
+      "learning_rate": 0.0003946206896551725,
+      "loss": 0.4981,
       "step": 1267
     },
     {
       "epoch": 23.05,
-      "learning_rate": 0.00033,
-      "loss": 0.4628,
+      "learning_rate": 0.0003944827586206897,
+      "loss": 0.4972,
       "step": 1268
     },
     {
       "epoch": 23.07,
-      "learning_rate": 0.0003297777777777778,
-      "loss": 0.4722,
+      "learning_rate": 0.0003943448275862069,
+      "loss": 0.5131,
       "step": 1269
     },
     {
       "epoch": 23.09,
-      "learning_rate": 0.00032955555555555554,
-      "loss": 0.429,
+      "learning_rate": 0.00039420689655172416,
+      "loss": 0.389,
       "step": 1270
     },
     {
       "epoch": 23.11,
-      "learning_rate": 0.00032933333333333333,
-      "loss": 0.4034,
+      "learning_rate": 0.0003940689655172414,
+      "loss": 0.44,
       "step": 1271
     },
     {
       "epoch": 23.13,
-      "learning_rate": 0.00032911111111111113,
-      "loss": 0.3931,
+      "learning_rate": 0.00039393103448275864,
+      "loss": 0.4436,
       "step": 1272
     },
     {
       "epoch": 23.15,
-      "learning_rate": 0.0003288888888888889,
-      "loss": 0.475,
+      "learning_rate": 0.00039379310344827585,
+      "loss": 0.521,
       "step": 1273
     },
     {
       "epoch": 23.16,
-      "learning_rate": 0.00032866666666666667,
-      "loss": 0.4379,
+      "learning_rate": 0.0003936551724137931,
+      "loss": 0.4892,
       "step": 1274
     },
     {
       "epoch": 23.18,
-      "learning_rate": 0.0003284444444444444,
-      "loss": 0.3896,
+      "learning_rate": 0.00039351724137931033,
+      "loss": 0.4416,
       "step": 1275
     },
     {
       "epoch": 23.2,
-      "learning_rate": 0.00032822222222222226,
-      "loss": 0.4728,
+      "learning_rate": 0.0003933793103448276,
+      "loss": 0.5411,
       "step": 1276
     },
     {
       "epoch": 23.22,
-      "learning_rate": 0.000328,
-      "loss": 0.49,
+      "learning_rate": 0.0003932413793103448,
+      "loss": 0.5555,
       "step": 1277
     },
     {
       "epoch": 23.24,
-      "learning_rate": 0.0003277777777777778,
-      "loss": 0.4321,
+      "learning_rate": 0.0003931034482758621,
+      "loss": 0.4817,
       "step": 1278
     },
     {
       "epoch": 23.25,
-      "learning_rate": 0.00032755555555555554,
-      "loss": 0.4994,
+      "learning_rate": 0.0003929655172413793,
+      "loss": 0.5215,
       "step": 1279
     },
     {
       "epoch": 23.27,
-      "learning_rate": 0.00032733333333333334,
-      "loss": 0.4574,
+      "learning_rate": 0.00039282758620689655,
+      "loss": 0.4897,
       "step": 1280
     },
     {
       "epoch": 23.29,
-      "learning_rate": 0.00032711111111111114,
-      "loss": 0.4234,
+      "learning_rate": 0.00039268965517241377,
+      "loss": 0.4506,
       "step": 1281
     },
     {
       "epoch": 23.31,
-      "learning_rate": 0.0003268888888888889,
-      "loss": 0.4383,
+      "learning_rate": 0.00039255172413793103,
+      "loss": 0.4567,
       "step": 1282
     },
     {
       "epoch": 23.33,
-      "learning_rate": 0.0003266666666666667,
-      "loss": 0.3884,
+      "learning_rate": 0.00039241379310344824,
+      "loss": 0.44,
       "step": 1283
     },
     {
       "epoch": 23.35,
-      "learning_rate": 0.0003264444444444444,
-      "loss": 0.53,
+      "learning_rate": 0.00039227586206896556,
+      "loss": 0.4941,
       "step": 1284
     },
     {
       "epoch": 23.36,
-      "learning_rate": 0.00032622222222222227,
-      "loss": 0.4567,
+      "learning_rate": 0.0003921379310344828,
+      "loss": 0.5162,
       "step": 1285
     },
     {
       "epoch": 23.38,
-      "learning_rate": 0.000326,
-      "loss": 0.4157,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 0.4412,
       "step": 1286
     },
     {
       "epoch": 23.4,
-      "learning_rate": 0.0003257777777777778,
-      "loss": 0.4171,
+      "learning_rate": 0.00039186206896551725,
+      "loss": 0.4657,
       "step": 1287
     },
     {
       "epoch": 23.42,
-      "learning_rate": 0.00032555555555555555,
-      "loss": 0.4681,
+      "learning_rate": 0.0003917241379310345,
+      "loss": 0.5021,
       "step": 1288
     },
     {
       "epoch": 23.44,
-      "learning_rate": 0.0003253333333333333,
-      "loss": 0.4652,
+      "learning_rate": 0.00039158620689655173,
+      "loss": 0.4719,
       "step": 1289
     },
     {
       "epoch": 23.45,
-      "learning_rate": 0.00032511111111111114,
-      "loss": 0.3938,
+      "learning_rate": 0.000391448275862069,
+      "loss": 0.454,
       "step": 1290
     },
     {
       "epoch": 23.47,
-      "learning_rate": 0.0003248888888888889,
-      "loss": 0.4786,
+      "learning_rate": 0.0003913103448275862,
+      "loss": 0.5098,
       "step": 1291
     },
     {
       "epoch": 23.49,
-      "learning_rate": 0.0003246666666666667,
-      "loss": 0.5314,
+      "learning_rate": 0.0003911724137931035,
+      "loss": 0.5702,
       "step": 1292
     },
     {
       "epoch": 23.51,
-      "learning_rate": 0.0003244444444444444,
-      "loss": 0.4751,
+      "learning_rate": 0.0003910344827586207,
+      "loss": 0.5027,
       "step": 1293
     },
     {
       "epoch": 23.53,
-      "learning_rate": 0.00032422222222222227,
-      "loss": 0.4025,
+      "learning_rate": 0.00039089655172413795,
+      "loss": 0.4447,
       "step": 1294
     },
     {
       "epoch": 23.55,
-      "learning_rate": 0.000324,
-      "loss": 0.4919,
+      "learning_rate": 0.00039075862068965517,
+      "loss": 0.5047,
       "step": 1295
     },
     {
       "epoch": 23.56,
-      "learning_rate": 0.00032377777777777776,
-      "loss": 0.417,
+      "learning_rate": 0.00039062068965517243,
+      "loss": 0.4972,
       "step": 1296
     },
     {
       "epoch": 23.58,
-      "learning_rate": 0.00032355555555555555,
-      "loss": 0.4389,
+      "learning_rate": 0.00039048275862068964,
+      "loss": 0.5317,
       "step": 1297
     },
     {
       "epoch": 23.6,
-      "learning_rate": 0.0003233333333333333,
-      "loss": 0.4395,
+      "learning_rate": 0.0003903448275862069,
+      "loss": 0.507,
       "step": 1298
     },
     {
       "epoch": 23.62,
-      "learning_rate": 0.00032311111111111115,
-      "loss": 0.458,
+      "learning_rate": 0.0003902068965517241,
+      "loss": 0.4701,
       "step": 1299
     },
     {
       "epoch": 23.64,
-      "learning_rate": 0.0003228888888888889,
-      "loss": 0.4691,
+      "learning_rate": 0.00039006896551724144,
+      "loss": 0.5111,
       "step": 1300
     },
     {
       "epoch": 23.65,
-      "learning_rate": 0.0003226666666666667,
-      "loss": 0.4186,
+      "learning_rate": 0.00038993103448275865,
+      "loss": 0.431,
       "step": 1301
     },
     {
       "epoch": 23.67,
-      "learning_rate": 0.00032244444444444443,
-      "loss": 0.5006,
+      "learning_rate": 0.00038979310344827586,
+      "loss": 0.5222,
       "step": 1302
     },
     {
       "epoch": 23.69,
-      "learning_rate": 0.0003222222222222222,
-      "loss": 0.4164,
+      "learning_rate": 0.00038965517241379313,
+      "loss": 0.4836,
       "step": 1303
     },
     {
       "epoch": 23.71,
-      "learning_rate": 0.000322,
-      "loss": 0.4513,
+      "learning_rate": 0.00038951724137931034,
+      "loss": 0.4728,
       "step": 1304
     },
     {
       "epoch": 23.73,
-      "learning_rate": 0.00032177777777777776,
-      "loss": 0.4569,
+      "learning_rate": 0.0003893793103448276,
+      "loss": 0.467,
       "step": 1305
     },
     {
       "epoch": 23.75,
-      "learning_rate": 0.00032155555555555556,
-      "loss": 0.3846,
+      "learning_rate": 0.0003892413793103448,
+      "loss": 0.4123,
       "step": 1306
     },
     {
       "epoch": 23.76,
-      "learning_rate": 0.00032133333333333336,
-      "loss": 0.4545,
+      "learning_rate": 0.0003891034482758621,
+      "loss": 0.4954,
       "step": 1307
     },
     {
       "epoch": 23.78,
-      "learning_rate": 0.00032111111111111115,
-      "loss": 0.4341,
+      "learning_rate": 0.0003889655172413793,
+      "loss": 0.4599,
       "step": 1308
     },
     {
       "epoch": 23.8,
-      "learning_rate": 0.0003208888888888889,
-      "loss": 0.3943,
+      "learning_rate": 0.00038882758620689656,
+      "loss": 0.4416,
       "step": 1309
     },
     {
       "epoch": 23.82,
-      "learning_rate": 0.00032066666666666664,
-      "loss": 0.4613,
+      "learning_rate": 0.0003886896551724138,
+      "loss": 0.5055,
       "step": 1310
     },
     {
       "epoch": 23.84,
-      "learning_rate": 0.00032044444444444443,
-      "loss": 0.4269,
+      "learning_rate": 0.00038855172413793104,
+      "loss": 0.4733,
       "step": 1311
     },
     {
       "epoch": 23.85,
-      "learning_rate": 0.00032022222222222223,
-      "loss": 0.3864,
+      "learning_rate": 0.00038841379310344825,
+      "loss": 0.455,
       "step": 1312
     },
     {
       "epoch": 23.87,
-      "learning_rate": 0.00032,
-      "loss": 0.3784,
+      "learning_rate": 0.0003882758620689655,
+      "loss": 0.445,
       "step": 1313
     },
     {
       "epoch": 23.89,
-      "learning_rate": 0.00031977777777777777,
-      "loss": 0.4961,
+      "learning_rate": 0.00038813793103448273,
+      "loss": 0.5067,
       "step": 1314
     },
     {
       "epoch": 23.91,
-      "learning_rate": 0.00031955555555555556,
-      "loss": 0.3975,
+      "learning_rate": 0.000388,
+      "loss": 0.4401,
       "step": 1315
     },
     {
       "epoch": 23.93,
-      "learning_rate": 0.00031933333333333336,
-      "loss": 0.4374,
+      "learning_rate": 0.0003878620689655172,
+      "loss": 0.5081,
       "step": 1316
     },
     {
       "epoch": 23.95,
-      "learning_rate": 0.0003191111111111111,
-      "loss": 0.4305,
+      "learning_rate": 0.00038772413793103453,
+      "loss": 0.5105,
       "step": 1317
     },
     {
       "epoch": 23.96,
-      "learning_rate": 0.0003188888888888889,
-      "loss": 0.4046,
+      "learning_rate": 0.00038758620689655174,
+      "loss": 0.4734,
       "step": 1318
     },
     {
       "epoch": 23.98,
-      "learning_rate": 0.00031866666666666664,
-      "loss": 0.4276,
+      "learning_rate": 0.000387448275862069,
+      "loss": 0.4777,
       "step": 1319
     },
     {
       "epoch": 24.0,
-      "learning_rate": 0.00031844444444444444,
-      "loss": 0.4033,
+      "learning_rate": 0.0003873103448275862,
+      "loss": 0.4027,
       "step": 1320
     },
     {
       "epoch": 24.02,
-      "learning_rate": 0.00031822222222222224,
-      "loss": 0.4425,
+      "learning_rate": 0.0003871724137931035,
+      "loss": 0.5112,
       "step": 1321
     },
     {
       "epoch": 24.04,
-      "learning_rate": 0.00031800000000000003,
-      "loss": 0.4525,
+      "learning_rate": 0.0003870344827586207,
+      "loss": 0.5036,
       "step": 1322
     },
     {
       "epoch": 24.05,
-      "learning_rate": 0.0003177777777777778,
-      "loss": 0.4466,
+      "learning_rate": 0.00038689655172413796,
+      "loss": 0.4938,
       "step": 1323
     },
     {
       "epoch": 24.07,
-      "learning_rate": 0.0003175555555555555,
-      "loss": 0.3955,
+      "learning_rate": 0.0003867586206896552,
+      "loss": 0.473,
       "step": 1324
     },
     {
       "epoch": 24.09,
-      "learning_rate": 0.00031733333333333337,
-      "loss": 0.3806,
+      "learning_rate": 0.00038662068965517244,
+      "loss": 0.464,
       "step": 1325
     },
     {
       "epoch": 24.11,
-      "learning_rate": 0.0003171111111111111,
-      "loss": 0.4086,
+      "learning_rate": 0.00038648275862068965,
+      "loss": 0.4326,
       "step": 1326
     },
     {
       "epoch": 24.13,
-      "learning_rate": 0.0003168888888888889,
-      "loss": 0.4271,
+      "learning_rate": 0.0003863448275862069,
+      "loss": 0.4744,
       "step": 1327
     },
     {
       "epoch": 24.15,
-      "learning_rate": 0.00031666666666666665,
-      "loss": 0.4319,
+      "learning_rate": 0.00038620689655172413,
+      "loss": 0.4558,
       "step": 1328
     },
     {
       "epoch": 24.16,
-      "learning_rate": 0.00031644444444444444,
-      "loss": 0.4114,
+      "learning_rate": 0.0003860689655172414,
+      "loss": 0.4654,
       "step": 1329
     },
     {
       "epoch": 24.18,
-      "learning_rate": 0.00031622222222222224,
-      "loss": 0.4229,
+      "learning_rate": 0.0003859310344827586,
+      "loss": 0.5039,
       "step": 1330
     },
     {
       "epoch": 24.2,
-      "learning_rate": 0.000316,
-      "loss": 0.386,
+      "learning_rate": 0.0003857931034482759,
+      "loss": 0.435,
       "step": 1331
     },
     {
       "epoch": 24.22,
-      "learning_rate": 0.0003157777777777778,
-      "loss": 0.4215,
+      "learning_rate": 0.0003856551724137931,
+      "loss": 0.4699,
       "step": 1332
     },
     {
       "epoch": 24.24,
-      "learning_rate": 0.0003155555555555555,
-      "loss": 0.4079,
+      "learning_rate": 0.00038551724137931035,
+      "loss": 0.4696,
       "step": 1333
     },
     {
       "epoch": 24.25,
-      "learning_rate": 0.00031533333333333337,
-      "loss": 0.4387,
+      "learning_rate": 0.0003853793103448276,
+      "loss": 0.5155,
       "step": 1334
     },
     {
       "epoch": 24.27,
-      "learning_rate": 0.0003151111111111111,
-      "loss": 0.4079,
+      "learning_rate": 0.00038524137931034483,
+      "loss": 0.4581,
       "step": 1335
     },
     {
       "epoch": 24.29,
-      "learning_rate": 0.0003148888888888889,
-      "loss": 0.3533,
+      "learning_rate": 0.0003851034482758621,
+      "loss": 0.4251,
       "step": 1336
     },
     {
       "epoch": 24.31,
-      "learning_rate": 0.00031466666666666665,
-      "loss": 0.3653,
+      "learning_rate": 0.0003849655172413793,
+      "loss": 0.4146,
       "step": 1337
     },
     {
       "epoch": 24.33,
-      "learning_rate": 0.0003144444444444445,
-      "loss": 0.4564,
+      "learning_rate": 0.0003848275862068966,
+      "loss": 0.4879,
       "step": 1338
     },
     {
       "epoch": 24.35,
-      "learning_rate": 0.00031422222222222225,
-      "loss": 0.4423,
+      "learning_rate": 0.0003846896551724138,
+      "loss": 0.4558,
       "step": 1339
     },
     {
       "epoch": 24.36,
-      "learning_rate": 0.000314,
-      "loss": 0.3915,
+      "learning_rate": 0.00038455172413793105,
+      "loss": 0.4352,
       "step": 1340
     },
     {
       "epoch": 24.38,
-      "learning_rate": 0.0003137777777777778,
-      "loss": 0.406,
+      "learning_rate": 0.00038441379310344827,
+      "loss": 0.4208,
       "step": 1341
     },
     {
       "epoch": 24.4,
-      "learning_rate": 0.00031355555555555553,
-      "loss": 0.4195,
+      "learning_rate": 0.00038427586206896553,
+      "loss": 0.4732,
       "step": 1342
     },
     {
       "epoch": 24.42,
-      "learning_rate": 0.0003133333333333334,
-      "loss": 0.3856,
+      "learning_rate": 0.00038413793103448274,
+      "loss": 0.4524,
       "step": 1343
     },
     {
       "epoch": 24.44,
-      "learning_rate": 0.0003131111111111111,
-      "loss": 0.4669,
+      "learning_rate": 0.000384,
+      "loss": 0.5031,
       "step": 1344
     },
     {
       "epoch": 24.45,
-      "learning_rate": 0.00031288888888888886,
-      "loss": 0.4172,
+      "learning_rate": 0.0003838620689655172,
+      "loss": 0.4434,
       "step": 1345
     },
     {
       "epoch": 24.47,
-      "learning_rate": 0.00031266666666666666,
-      "loss": 0.4785,
+      "learning_rate": 0.0003837241379310345,
+      "loss": 0.5295,
       "step": 1346
     },
     {
       "epoch": 24.49,
-      "learning_rate": 0.00031244444444444446,
-      "loss": 0.4557,
+      "learning_rate": 0.0003835862068965517,
+      "loss": 0.506,
       "step": 1347
     },
     {
       "epoch": 24.51,
-      "learning_rate": 0.00031222222222222225,
-      "loss": 0.4377,
+      "learning_rate": 0.00038344827586206897,
+      "loss": 0.4741,
       "step": 1348
     },
     {
       "epoch": 24.53,
-      "learning_rate": 0.000312,
-      "loss": 0.4393,
+      "learning_rate": 0.0003833103448275862,
+      "loss": 0.4744,
       "step": 1349
     },
     {
       "epoch": 24.55,
-      "learning_rate": 0.0003117777777777778,
-      "loss": 0.4631,
+      "learning_rate": 0.00038317241379310344,
+      "loss": 0.5077,
       "step": 1350
     },
     {
       "epoch": 24.56,
-      "learning_rate": 0.00031155555555555553,
-      "loss": 0.3922,
+      "learning_rate": 0.0003830344827586207,
+      "loss": 0.4226,
       "step": 1351
     },
     {
       "epoch": 24.58,
-      "learning_rate": 0.0003113333333333334,
-      "loss": 0.4194,
+      "learning_rate": 0.000382896551724138,
+      "loss": 0.4937,
       "step": 1352
     },
     {
       "epoch": 24.6,
-      "learning_rate": 0.0003111111111111111,
-      "loss": 0.3981,
+      "learning_rate": 0.0003827586206896552,
+      "loss": 0.423,
       "step": 1353
     },
     {
       "epoch": 24.62,
-      "learning_rate": 0.00031088888888888887,
-      "loss": 0.3899,
+      "learning_rate": 0.00038262068965517245,
+      "loss": 0.4362,
       "step": 1354
     },
     {
       "epoch": 24.64,
-      "learning_rate": 0.00031066666666666666,
-      "loss": 0.4372,
+      "learning_rate": 0.00038248275862068967,
+      "loss": 0.4844,
       "step": 1355
     },
     {
       "epoch": 24.65,
-      "learning_rate": 0.00031044444444444446,
-      "loss": 0.4343,
+      "learning_rate": 0.00038234482758620693,
+      "loss": 0.4937,
       "step": 1356
     },
     {
       "epoch": 24.67,
-      "learning_rate": 0.00031022222222222226,
-      "loss": 0.4788,
+      "learning_rate": 0.00038220689655172414,
+      "loss": 0.5477,
       "step": 1357
     },
     {
       "epoch": 24.69,
-      "learning_rate": 0.00031,
-      "loss": 0.4016,
+      "learning_rate": 0.0003820689655172414,
+      "loss": 0.4334,
       "step": 1358
     },
     {
       "epoch": 24.71,
-      "learning_rate": 0.0003097777777777778,
-      "loss": 0.4269,
+      "learning_rate": 0.0003819310344827586,
+      "loss": 0.4705,
       "step": 1359
     },
     {
       "epoch": 24.73,
-      "learning_rate": 0.00030955555555555554,
-      "loss": 0.3948,
+      "learning_rate": 0.0003817931034482759,
+      "loss": 0.4255,
       "step": 1360
     },
     {
       "epoch": 24.75,
-      "learning_rate": 0.00030933333333333334,
-      "loss": 0.4908,
+      "learning_rate": 0.0003816551724137931,
+      "loss": 0.5021,
       "step": 1361
     },
     {
       "epoch": 24.76,
-      "learning_rate": 0.00030911111111111113,
-      "loss": 0.3653,
+      "learning_rate": 0.00038151724137931037,
+      "loss": 0.3808,
       "step": 1362
     },
     {
       "epoch": 24.78,
-      "learning_rate": 0.0003088888888888889,
-      "loss": 0.4347,
+      "learning_rate": 0.0003813793103448276,
+      "loss": 0.5279,
       "step": 1363
     },
     {
       "epoch": 24.8,
-      "learning_rate": 0.00030866666666666667,
-      "loss": 0.452,
+      "learning_rate": 0.00038124137931034484,
+      "loss": 0.4489,
       "step": 1364
     },
     {
       "epoch": 24.82,
-      "learning_rate": 0.00030844444444444447,
-      "loss": 0.4513,
+      "learning_rate": 0.00038110344827586206,
+      "loss": 0.4884,
       "step": 1365
     },
     {
       "epoch": 24.84,
-      "learning_rate": 0.00030822222222222226,
-      "loss": 0.3954,
+      "learning_rate": 0.0003809655172413793,
+      "loss": 0.4525,
       "step": 1366
     },
     {
       "epoch": 24.85,
-      "learning_rate": 0.000308,
-      "loss": 0.4366,
+      "learning_rate": 0.00038082758620689653,
+      "loss": 0.4823,
       "step": 1367
     },
     {
       "epoch": 24.87,
-      "learning_rate": 0.00030777777777777775,
-      "loss": 0.444,
+      "learning_rate": 0.0003806896551724138,
+      "loss": 0.5119,
       "step": 1368
     },
     {
       "epoch": 24.89,
-      "learning_rate": 0.0003075555555555556,
-      "loss": 0.4592,
+      "learning_rate": 0.00038055172413793107,
+      "loss": 0.5073,
       "step": 1369
     },
     {
       "epoch": 24.91,
-      "learning_rate": 0.00030733333333333334,
-      "loss": 0.4377,
+      "learning_rate": 0.0003804137931034483,
+      "loss": 0.4787,
       "step": 1370
     },
     {
       "epoch": 24.93,
-      "learning_rate": 0.00030711111111111114,
-      "loss": 0.4321,
+      "learning_rate": 0.00038027586206896554,
+      "loss": 0.4652,
       "step": 1371
     },
     {
       "epoch": 24.95,
-      "learning_rate": 0.0003068888888888889,
-      "loss": 0.4309,
+      "learning_rate": 0.00038013793103448275,
+      "loss": 0.4896,
       "step": 1372
     },
     {
       "epoch": 24.96,
-      "learning_rate": 0.0003066666666666667,
-      "loss": 0.4139,
+      "learning_rate": 0.00038,
+      "loss": 0.4357,
       "step": 1373
     },
     {
       "epoch": 24.98,
-      "learning_rate": 0.00030644444444444447,
-      "loss": 0.4808,
+      "learning_rate": 0.00037986206896551723,
+      "loss": 0.4797,
       "step": 1374
     },
     {
       "epoch": 25.0,
-      "learning_rate": 0.0003062222222222222,
-      "loss": 0.404,
+      "learning_rate": 0.0003797241379310345,
+      "loss": 0.4977,
       "step": 1375
     },
     {
       "epoch": 25.02,
-      "learning_rate": 0.000306,
-      "loss": 0.4206,
+      "learning_rate": 0.0003795862068965517,
+      "loss": 0.4447,
       "step": 1376
     },
     {
       "epoch": 25.04,
-      "learning_rate": 0.00030577777777777775,
-      "loss": 0.3975,
+      "learning_rate": 0.000379448275862069,
+      "loss": 0.4526,
       "step": 1377
     },
     {
       "epoch": 25.05,
-      "learning_rate": 0.0003055555555555556,
-      "loss": 0.4143,
+      "learning_rate": 0.0003793103448275862,
+      "loss": 0.4642,
       "step": 1378
     },
     {
       "epoch": 25.07,
-      "learning_rate": 0.00030533333333333335,
-      "loss": 0.394,
+      "learning_rate": 0.00037917241379310345,
+      "loss": 0.4327,
       "step": 1379
     },
     {
       "epoch": 25.09,
-      "learning_rate": 0.00030511111111111114,
-      "loss": 0.4164,
+      "learning_rate": 0.00037903448275862067,
+      "loss": 0.4491,
       "step": 1380
     },
     {
       "epoch": 25.11,
-      "learning_rate": 0.0003048888888888889,
-      "loss": 0.3617,
+      "learning_rate": 0.00037889655172413793,
+      "loss": 0.3997,
       "step": 1381
     },
     {
       "epoch": 25.13,
-      "learning_rate": 0.00030466666666666663,
-      "loss": 0.3328,
+      "learning_rate": 0.00037875862068965514,
+      "loss": 0.3761,
       "step": 1382
     },
     {
       "epoch": 25.15,
-      "learning_rate": 0.0003044444444444445,
-      "loss": 0.3934,
+      "learning_rate": 0.0003786206896551724,
+      "loss": 0.4194,
       "step": 1383
     },
     {
       "epoch": 25.16,
-      "learning_rate": 0.0003042222222222222,
-      "loss": 0.3916,
+      "learning_rate": 0.0003784827586206896,
+      "loss": 0.4417,
       "step": 1384
     },
     {
       "epoch": 25.18,
-      "learning_rate": 0.000304,
-      "loss": 0.3988,
+      "learning_rate": 0.00037834482758620694,
+      "loss": 0.4435,
       "step": 1385
     },
     {
       "epoch": 25.2,
-      "learning_rate": 0.00030377777777777776,
-      "loss": 0.3941,
+      "learning_rate": 0.00037820689655172415,
+      "loss": 0.4514,
       "step": 1386
     },
     {
       "epoch": 25.22,
-      "learning_rate": 0.0003035555555555556,
-      "loss": 0.3511,
+      "learning_rate": 0.0003780689655172414,
+      "loss": 0.4221,
       "step": 1387
     },
     {
       "epoch": 25.24,
-      "learning_rate": 0.00030333333333333335,
-      "loss": 0.3958,
+      "learning_rate": 0.00037793103448275863,
+      "loss": 0.4331,
       "step": 1388
     },
     {
       "epoch": 25.25,
-      "learning_rate": 0.0003031111111111111,
-      "loss": 0.4841,
+      "learning_rate": 0.0003777931034482759,
+      "loss": 0.5034,
       "step": 1389
     },
     {
       "epoch": 25.27,
-      "learning_rate": 0.0003028888888888889,
-      "loss": 0.3927,
+      "learning_rate": 0.0003776551724137931,
+      "loss": 0.4334,
       "step": 1390
     },
     {
       "epoch": 25.29,
-      "learning_rate": 0.00030266666666666663,
-      "loss": 0.3895,
+      "learning_rate": 0.0003775172413793104,
+      "loss": 0.4364,
       "step": 1391
     },
     {
       "epoch": 25.31,
-      "learning_rate": 0.0003024444444444445,
-      "loss": 0.4362,
+      "learning_rate": 0.0003773793103448276,
+      "loss": 0.4809,
       "step": 1392
     },
     {
       "epoch": 25.33,
-      "learning_rate": 0.0003022222222222222,
-      "loss": 0.4828,
+      "learning_rate": 0.00037724137931034485,
+      "loss": 0.5467,
       "step": 1393
     },
     {
       "epoch": 25.35,
-      "learning_rate": 0.000302,
-      "loss": 0.4866,
+      "learning_rate": 0.00037710344827586207,
+      "loss": 0.4554,
       "step": 1394
     },
     {
       "epoch": 25.36,
-      "learning_rate": 0.00030177777777777776,
-      "loss": 0.3786,
+      "learning_rate": 0.00037696551724137933,
+      "loss": 0.4282,
       "step": 1395
     },
     {
       "epoch": 25.38,
-      "learning_rate": 0.00030155555555555556,
-      "loss": 0.3709,
+      "learning_rate": 0.00037682758620689654,
+      "loss": 0.4265,
       "step": 1396
     },
     {
       "epoch": 25.4,
-      "learning_rate": 0.00030133333333333336,
-      "loss": 0.4484,
+      "learning_rate": 0.0003766896551724138,
+      "loss": 0.5173,
       "step": 1397
     },
     {
       "epoch": 25.42,
-      "learning_rate": 0.0003011111111111111,
-      "loss": 0.3726,
+      "learning_rate": 0.000376551724137931,
+      "loss": 0.4407,
       "step": 1398
     },
     {
       "epoch": 25.44,
-      "learning_rate": 0.0003008888888888889,
-      "loss": 0.3891,
+      "learning_rate": 0.0003764137931034483,
+      "loss": 0.4444,
       "step": 1399
     },
     {
       "epoch": 25.45,
-      "learning_rate": 0.0003006666666666667,
-      "loss": 0.3328,
+      "learning_rate": 0.0003762758620689655,
+      "loss": 0.4301,
       "step": 1400
     },
     {
       "epoch": 25.47,
-      "learning_rate": 0.0003004444444444445,
-      "loss": 0.4332,
+      "learning_rate": 0.0003761379310344827,
+      "loss": 0.4685,
       "step": 1401
     },
     {
       "epoch": 25.49,
-      "learning_rate": 0.00030022222222222223,
-      "loss": 0.3959,
+      "learning_rate": 0.00037600000000000003,
+      "loss": 0.4153,
       "step": 1402
     },
     {
       "epoch": 25.51,
-      "learning_rate": 0.0003,
-      "loss": 0.4018,
+      "learning_rate": 0.00037586206896551724,
+      "loss": 0.4357,
       "step": 1403
     },
     {
       "epoch": 25.53,
-      "learning_rate": 0.00029977777777777777,
-      "loss": 0.3746,
+      "learning_rate": 0.0003757241379310345,
+      "loss": 0.4346,
       "step": 1404
     },
     {
       "epoch": 25.55,
-      "learning_rate": 0.00029955555555555557,
-      "loss": 0.3803,
+      "learning_rate": 0.0003755862068965517,
+      "loss": 0.4138,
       "step": 1405
     },
     {
       "epoch": 25.56,
-      "learning_rate": 0.00029933333333333336,
-      "loss": 0.4981,
+      "learning_rate": 0.000375448275862069,
+      "loss": 0.5451,
       "step": 1406
     },
     {
       "epoch": 25.58,
-      "learning_rate": 0.0002991111111111111,
-      "loss": 0.4486,
+      "learning_rate": 0.0003753103448275862,
+      "loss": 0.4564,
       "step": 1407
     },
     {
       "epoch": 25.6,
-      "learning_rate": 0.0002988888888888889,
-      "loss": 0.4516,
+      "learning_rate": 0.00037517241379310347,
+      "loss": 0.4829,
       "step": 1408
     },
     {
       "epoch": 25.62,
-      "learning_rate": 0.0002986666666666667,
-      "loss": 0.372,
+      "learning_rate": 0.0003750344827586207,
+      "loss": 0.4137,
       "step": 1409
     },
     {
       "epoch": 25.64,
-      "learning_rate": 0.00029844444444444444,
-      "loss": 0.4203,
+      "learning_rate": 0.00037489655172413794,
+      "loss": 0.4685,
       "step": 1410
     },
     {
       "epoch": 25.65,
-      "learning_rate": 0.00029822222222222224,
-      "loss": 0.4362,
+      "learning_rate": 0.00037475862068965516,
+      "loss": 0.4475,
       "step": 1411
     },
     {
       "epoch": 25.67,
-      "learning_rate": 0.000298,
-      "loss": 0.4637,
+      "learning_rate": 0.0003746206896551724,
+      "loss": 0.4752,
       "step": 1412
     },
     {
       "epoch": 25.69,
-      "learning_rate": 0.0002977777777777778,
-      "loss": 0.4399,
+      "learning_rate": 0.00037448275862068963,
+      "loss": 0.4465,
       "step": 1413
     },
     {
       "epoch": 25.71,
-      "learning_rate": 0.00029755555555555557,
-      "loss": 0.4125,
+      "learning_rate": 0.0003743448275862069,
+      "loss": 0.471,
       "step": 1414
     },
     {
       "epoch": 25.73,
-      "learning_rate": 0.00029733333333333337,
-      "loss": 0.4239,
+      "learning_rate": 0.0003742068965517241,
+      "loss": 0.4559,
       "step": 1415
     },
     {
       "epoch": 25.75,
-      "learning_rate": 0.0002971111111111111,
-      "loss": 0.3852,
+      "learning_rate": 0.0003740689655172414,
+      "loss": 0.4091,
       "step": 1416
     },
     {
       "epoch": 25.76,
-      "learning_rate": 0.00029688888888888885,
-      "loss": 0.3952,
+      "learning_rate": 0.0003739310344827586,
+      "loss": 0.452,
       "step": 1417
     },
     {
       "epoch": 25.78,
-      "learning_rate": 0.0002966666666666667,
-      "loss": 0.3946,
+      "learning_rate": 0.0003737931034482759,
+      "loss": 0.4598,
       "step": 1418
     },
     {
       "epoch": 25.8,
-      "learning_rate": 0.00029644444444444445,
-      "loss": 0.4761,
+      "learning_rate": 0.0003736551724137931,
+      "loss": 0.506,
       "step": 1419
     },
     {
       "epoch": 25.82,
-      "learning_rate": 0.00029622222222222224,
-      "loss": 0.4747,
+      "learning_rate": 0.0003735172413793104,
+      "loss": 0.464,
       "step": 1420
     },
     {
       "epoch": 25.84,
-      "learning_rate": 0.000296,
-      "loss": 0.4144,
+      "learning_rate": 0.0003733793103448276,
+      "loss": 0.4168,
       "step": 1421
     },
     {
       "epoch": 25.85,
-      "learning_rate": 0.0002957777777777778,
-      "loss": 0.4353,
+      "learning_rate": 0.00037324137931034487,
+      "loss": 0.4899,
       "step": 1422
     },
     {
       "epoch": 25.87,
-      "learning_rate": 0.0002955555555555556,
-      "loss": 0.462,
+      "learning_rate": 0.0003731034482758621,
+      "loss": 0.4733,
       "step": 1423
     },
     {
       "epoch": 25.89,
-      "learning_rate": 0.0002953333333333333,
-      "loss": 0.382,
+      "learning_rate": 0.00037296551724137934,
+      "loss": 0.4828,
       "step": 1424
     },
     {
       "epoch": 25.91,
-      "learning_rate": 0.0002951111111111111,
-      "loss": 0.4343,
+      "learning_rate": 0.00037282758620689656,
+      "loss": 0.503,
       "step": 1425
     },
     {
       "epoch": 25.93,
-      "learning_rate": 0.00029488888888888886,
-      "loss": 0.4255,
+      "learning_rate": 0.0003726896551724138,
+      "loss": 0.4639,
       "step": 1426
     },
     {
       "epoch": 25.95,
-      "learning_rate": 0.0002946666666666667,
-      "loss": 0.3855,
+      "learning_rate": 0.00037255172413793103,
+      "loss": 0.4196,
       "step": 1427
     },
     {
       "epoch": 25.96,
-      "learning_rate": 0.00029444444444444445,
-      "loss": 0.3635,
+      "learning_rate": 0.0003724137931034483,
+      "loss": 0.4115,
       "step": 1428
     },
     {
       "epoch": 25.98,
-      "learning_rate": 0.00029422222222222225,
-      "loss": 0.4216,
+      "learning_rate": 0.0003722758620689655,
+      "loss": 0.4777,
       "step": 1429
     },
     {
       "epoch": 26.0,
-      "learning_rate": 0.000294,
-      "loss": 0.4017,
+      "learning_rate": 0.0003721379310344828,
+      "loss": 0.4695,
       "step": 1430
     },
     {
       "epoch": 26.02,
-      "learning_rate": 0.0002937777777777778,
-      "loss": 0.3955,
+      "learning_rate": 0.000372,
+      "loss": 0.473,
       "step": 1431
     },
     {
       "epoch": 26.04,
-      "learning_rate": 0.0002935555555555556,
-      "loss": 0.383,
+      "learning_rate": 0.00037186206896551726,
+      "loss": 0.3992,
       "step": 1432
     },
     {
       "epoch": 26.05,
-      "learning_rate": 0.0002933333333333333,
-      "loss": 0.3941,
+      "learning_rate": 0.00037172413793103447,
+      "loss": 0.4121,
       "step": 1433
     },
     {
       "epoch": 26.07,
-      "learning_rate": 0.0002931111111111111,
-      "loss": 0.4175,
+      "learning_rate": 0.00037158620689655173,
+      "loss": 0.4585,
       "step": 1434
     },
     {
       "epoch": 26.09,
-      "learning_rate": 0.00029288888888888886,
-      "loss": 0.4061,
+      "learning_rate": 0.000371448275862069,
+      "loss": 0.4066,
       "step": 1435
     },
     {
       "epoch": 26.11,
-      "learning_rate": 0.0002926666666666667,
-      "loss": 0.3951,
+      "learning_rate": 0.0003713103448275862,
+      "loss": 0.4379,
       "step": 1436
     },
     {
       "epoch": 26.13,
-      "learning_rate": 0.00029244444444444446,
-      "loss": 0.4301,
+      "learning_rate": 0.0003711724137931035,
+      "loss": 0.5174,
       "step": 1437
     },
     {
       "epoch": 26.15,
-      "learning_rate": 0.0002922222222222222,
-      "loss": 0.3264,
+      "learning_rate": 0.0003710344827586207,
+      "loss": 0.3902,
       "step": 1438
     },
     {
       "epoch": 26.16,
-      "learning_rate": 0.000292,
-      "loss": 0.3001,
+      "learning_rate": 0.00037089655172413795,
+      "loss": 0.3264,
       "step": 1439
     },
     {
       "epoch": 26.18,
-      "learning_rate": 0.0002917777777777778,
-      "loss": 0.4598,
+      "learning_rate": 0.00037075862068965517,
+      "loss": 0.4953,
       "step": 1440
     },
     {
       "epoch": 26.2,
-      "learning_rate": 0.0002915555555555556,
-      "loss": 0.3981,
+      "learning_rate": 0.00037062068965517243,
+      "loss": 0.4863,
       "step": 1441
     },
     {
       "epoch": 26.22,
-      "learning_rate": 0.00029133333333333333,
-      "loss": 0.4469,
+      "learning_rate": 0.00037048275862068964,
+      "loss": 0.5503,
       "step": 1442
     },
     {
       "epoch": 26.24,
-      "learning_rate": 0.00029111111111111113,
-      "loss": 0.3482,
+      "learning_rate": 0.0003703448275862069,
+      "loss": 0.3737,
       "step": 1443
     },
     {
       "epoch": 26.25,
-      "learning_rate": 0.00029088888888888887,
-      "loss": 0.4001,
+      "learning_rate": 0.0003702068965517241,
+      "loss": 0.4499,
       "step": 1444
     },
     {
       "epoch": 26.27,
-      "learning_rate": 0.00029066666666666667,
-      "loss": 0.3682,
+      "learning_rate": 0.0003700689655172414,
+      "loss": 0.426,
       "step": 1445
     },
     {
       "epoch": 26.29,
-      "learning_rate": 0.00029044444444444446,
-      "loss": 0.4122,
+      "learning_rate": 0.0003699310344827586,
+      "loss": 0.4888,
       "step": 1446
     },
     {
       "epoch": 26.31,
-      "learning_rate": 0.0002902222222222222,
-      "loss": 0.3566,
+      "learning_rate": 0.00036979310344827587,
+      "loss": 0.4039,
       "step": 1447
     },
     {
       "epoch": 26.33,
-      "learning_rate": 0.00029,
-      "loss": 0.4383,
+      "learning_rate": 0.0003696551724137931,
+      "loss": 0.455,
       "step": 1448
     },
     {
       "epoch": 26.35,
-      "learning_rate": 0.0002897777777777778,
-      "loss": 0.3229,
+      "learning_rate": 0.00036951724137931034,
+      "loss": 0.3908,
       "step": 1449
     },
     {
       "epoch": 26.36,
-      "learning_rate": 0.0002895555555555556,
-      "loss": 0.4256,
+      "learning_rate": 0.00036937931034482756,
+      "loss": 0.478,
       "step": 1450
     },
     {
       "epoch": 26.38,
-      "learning_rate": 0.00028933333333333334,
-      "loss": 0.3653,
+      "learning_rate": 0.0003692413793103448,
+      "loss": 0.4238,
       "step": 1451
     },
     {
       "epoch": 26.4,
-      "learning_rate": 0.0002891111111111111,
-      "loss": 0.3865,
+      "learning_rate": 0.0003691034482758621,
+      "loss": 0.4637,
       "step": 1452
     },
     {
       "epoch": 26.42,
-      "learning_rate": 0.0002888888888888889,
-      "loss": 0.4257,
+      "learning_rate": 0.00036896551724137935,
+      "loss": 0.4608,
       "step": 1453
     },
     {
       "epoch": 26.44,
-      "learning_rate": 0.00028866666666666667,
-      "loss": 0.3824,
+      "learning_rate": 0.00036882758620689657,
+      "loss": 0.4673,
       "step": 1454
     },
     {
       "epoch": 26.45,
-      "learning_rate": 0.00028844444444444447,
-      "loss": 0.431,
+      "learning_rate": 0.00036868965517241383,
+      "loss": 0.4866,
       "step": 1455
     },
     {
       "epoch": 26.47,
-      "learning_rate": 0.0002882222222222222,
-      "loss": 0.3342,
+      "learning_rate": 0.00036855172413793104,
+      "loss": 0.3533,
       "step": 1456
     },
     {
       "epoch": 26.49,
-      "learning_rate": 0.000288,
-      "loss": 0.3675,
+      "learning_rate": 0.0003684137931034483,
+      "loss": 0.4295,
       "step": 1457
     },
     {
       "epoch": 26.51,
-      "learning_rate": 0.0002877777777777778,
-      "loss": 0.3483,
+      "learning_rate": 0.0003682758620689655,
+      "loss": 0.3976,
       "step": 1458
     },
     {
       "epoch": 26.53,
-      "learning_rate": 0.00028755555555555555,
-      "loss": 0.4216,
+      "learning_rate": 0.0003681379310344828,
+      "loss": 0.4341,
       "step": 1459
     },
     {
       "epoch": 26.55,
-      "learning_rate": 0.00028733333333333334,
-      "loss": 0.4391,
+      "learning_rate": 0.000368,
+      "loss": 0.453,
       "step": 1460
     },
     {
       "epoch": 26.56,
-      "learning_rate": 0.0002871111111111111,
-      "loss": 0.436,
+      "learning_rate": 0.00036786206896551727,
+      "loss": 0.5009,
       "step": 1461
     },
     {
       "epoch": 26.58,
-      "learning_rate": 0.00028688888888888894,
-      "loss": 0.4447,
+      "learning_rate": 0.0003677241379310345,
+      "loss": 0.4592,
       "step": 1462
     },
     {
       "epoch": 26.6,
-      "learning_rate": 0.0002866666666666667,
-      "loss": 0.3639,
+      "learning_rate": 0.00036758620689655174,
+      "loss": 0.4004,
       "step": 1463
     },
     {
       "epoch": 26.62,
-      "learning_rate": 0.0002864444444444445,
-      "loss": 0.3611,
+      "learning_rate": 0.00036744827586206896,
+      "loss": 0.4625,
       "step": 1464
     },
     {
       "epoch": 26.64,
-      "learning_rate": 0.0002862222222222222,
-      "loss": 0.3984,
+      "learning_rate": 0.0003673103448275862,
+      "loss": 0.4662,
       "step": 1465
     },
     {
       "epoch": 26.65,
-      "learning_rate": 0.00028599999999999996,
-      "loss": 0.4514,
+      "learning_rate": 0.00036717241379310343,
+      "loss": 0.4587,
       "step": 1466
     },
     {
       "epoch": 26.67,
-      "learning_rate": 0.0002857777777777778,
-      "loss": 0.3941,
+      "learning_rate": 0.0003670344827586207,
+      "loss": 0.4425,
       "step": 1467
     },
     {
       "epoch": 26.69,
-      "learning_rate": 0.00028555555555555555,
-      "loss": 0.4164,
+      "learning_rate": 0.00036689655172413797,
+      "loss": 0.4481,
       "step": 1468
     },
     {
       "epoch": 26.71,
-      "learning_rate": 0.00028533333333333335,
-      "loss": 0.3539,
+      "learning_rate": 0.0003667586206896552,
+      "loss": 0.4289,
       "step": 1469
     },
     {
       "epoch": 26.73,
-      "learning_rate": 0.0002851111111111111,
-      "loss": 0.3573,
+      "learning_rate": 0.00036662068965517244,
+      "loss": 0.4394,
       "step": 1470
     },
     {
       "epoch": 26.75,
-      "learning_rate": 0.00028488888888888894,
-      "loss": 0.3789,
+      "learning_rate": 0.00036648275862068966,
+      "loss": 0.4301,
       "step": 1471
     },
     {
       "epoch": 26.76,
-      "learning_rate": 0.0002846666666666667,
-      "loss": 0.3519,
+      "learning_rate": 0.0003663448275862069,
+      "loss": 0.4337,
       "step": 1472
     },
     {
       "epoch": 26.78,
-      "learning_rate": 0.0002844444444444444,
-      "loss": 0.3583,
+      "learning_rate": 0.00036620689655172413,
+      "loss": 0.4598,
       "step": 1473
     },
     {
       "epoch": 26.8,
-      "learning_rate": 0.0002842222222222222,
-      "loss": 0.3879,
+      "learning_rate": 0.0003660689655172414,
+      "loss": 0.4085,
       "step": 1474
     },
     {
       "epoch": 26.82,
-      "learning_rate": 0.00028399999999999996,
-      "loss": 0.4425,
+      "learning_rate": 0.0003659310344827586,
+      "loss": 0.4963,
       "step": 1475
     },
     {
       "epoch": 26.84,
-      "learning_rate": 0.0002837777777777778,
-      "loss": 0.4043,
+      "learning_rate": 0.0003657931034482759,
+      "loss": 0.4507,
       "step": 1476
     },
     {
       "epoch": 26.85,
-      "learning_rate": 0.00028355555555555556,
-      "loss": 0.3248,
+      "learning_rate": 0.0003656551724137931,
+      "loss": 0.3798,
       "step": 1477
     },
     {
       "epoch": 26.87,
-      "learning_rate": 0.00028333333333333335,
-      "loss": 0.3764,
+      "learning_rate": 0.00036551724137931036,
+      "loss": 0.4418,
       "step": 1478
     },
     {
       "epoch": 26.89,
-      "learning_rate": 0.0002831111111111111,
-      "loss": 0.4336,
+      "learning_rate": 0.00036537931034482757,
+      "loss": 0.4249,
       "step": 1479
     },
     {
       "epoch": 26.91,
-      "learning_rate": 0.0002828888888888889,
-      "loss": 0.4246,
+      "learning_rate": 0.00036524137931034483,
+      "loss": 0.415,
       "step": 1480
     },
     {
       "epoch": 26.93,
-      "learning_rate": 0.0002826666666666667,
-      "loss": 0.4466,
+      "learning_rate": 0.00036510344827586205,
+      "loss": 0.4878,
       "step": 1481
     },
     {
       "epoch": 26.95,
-      "learning_rate": 0.00028244444444444443,
-      "loss": 0.3368,
+      "learning_rate": 0.0003649655172413793,
+      "loss": 0.4313,
       "step": 1482
     },
     {
       "epoch": 26.96,
-      "learning_rate": 0.00028222222222222223,
-      "loss": 0.3934,
+      "learning_rate": 0.0003648275862068965,
+      "loss": 0.4876,
       "step": 1483
     },
     {
       "epoch": 26.98,
-      "learning_rate": 0.00028199999999999997,
-      "loss": 0.3919,
+      "learning_rate": 0.0003646896551724138,
+      "loss": 0.464,
       "step": 1484
     },
     {
       "epoch": 27.0,
-      "learning_rate": 0.0002817777777777778,
-      "loss": 0.2876,
+      "learning_rate": 0.00036455172413793106,
+      "loss": 0.3934,
       "step": 1485
     },
     {
       "epoch": 27.02,
-      "learning_rate": 0.00028155555555555556,
-      "loss": 0.4504,
+      "learning_rate": 0.0003644137931034483,
+      "loss": 0.4902,
       "step": 1486
     },
     {
       "epoch": 27.04,
-      "learning_rate": 0.0002813333333333333,
-      "loss": 0.3474,
+      "learning_rate": 0.00036427586206896553,
+      "loss": 0.4051,
       "step": 1487
     },
     {
       "epoch": 27.05,
-      "learning_rate": 0.0002811111111111111,
-      "loss": 0.3725,
+      "learning_rate": 0.0003641379310344828,
+      "loss": 0.3797,
       "step": 1488
     },
     {
       "epoch": 27.07,
-      "learning_rate": 0.0002808888888888889,
-      "loss": 0.3844,
+      "learning_rate": 0.000364,
+      "loss": 0.4541,
       "step": 1489
     },
     {
       "epoch": 27.09,
-      "learning_rate": 0.0002806666666666667,
-      "loss": 0.3579,
+      "learning_rate": 0.0003638620689655173,
+      "loss": 0.4243,
       "step": 1490
     },
     {
       "epoch": 27.11,
-      "learning_rate": 0.00028044444444444444,
-      "loss": 0.4214,
+      "learning_rate": 0.0003637241379310345,
+      "loss": 0.489,
       "step": 1491
     },
     {
       "epoch": 27.13,
-      "learning_rate": 0.00028022222222222223,
-      "loss": 0.3728,
+      "learning_rate": 0.00036358620689655176,
+      "loss": 0.4024,
       "step": 1492
     },
     {
       "epoch": 27.15,
-      "learning_rate": 0.00028000000000000003,
-      "loss": 0.3645,
+      "learning_rate": 0.00036344827586206897,
+      "loss": 0.426,
       "step": 1493
     },
     {
       "epoch": 27.16,
-      "learning_rate": 0.00027977777777777777,
-      "loss": 0.3662,
+      "learning_rate": 0.00036331034482758623,
+      "loss": 0.4303,
       "step": 1494
     },
     {
       "epoch": 27.18,
-      "learning_rate": 0.00027955555555555557,
-      "loss": 0.4295,
+      "learning_rate": 0.00036317241379310345,
+      "loss": 0.4856,
       "step": 1495
     },
     {
       "epoch": 27.2,
-      "learning_rate": 0.0002793333333333333,
-      "loss": 0.3497,
+      "learning_rate": 0.0003630344827586207,
+      "loss": 0.4062,
       "step": 1496
     },
     {
       "epoch": 27.22,
-      "learning_rate": 0.0002791111111111111,
-      "loss": 0.3716,
+      "learning_rate": 0.0003628965517241379,
+      "loss": 0.432,
       "step": 1497
     },
     {
       "epoch": 27.24,
-      "learning_rate": 0.0002788888888888889,
-      "loss": 0.3103,
+      "learning_rate": 0.0003627586206896552,
+      "loss": 0.3498,
       "step": 1498
     },
     {
       "epoch": 27.25,
-      "learning_rate": 0.0002786666666666667,
-      "loss": 0.43,
+      "learning_rate": 0.0003626206896551724,
+      "loss": 0.4592,
       "step": 1499
     },
     {
       "epoch": 27.27,
-      "learning_rate": 0.00027844444444444444,
-      "loss": 0.4012,
+      "learning_rate": 0.00036248275862068967,
+      "loss": 0.4343,
       "step": 1500
     },
     {
       "epoch": 27.27,
-      "eval_cer": 0.09349465509228025,
-      "eval_loss": 0.37162724137306213,
-      "eval_runtime": 13.8851,
-      "eval_samples_per_second": 118.617,
-      "eval_steps_per_second": 1.873,
-      "eval_wer": 0.3781023388826473,
+      "eval_cer": 0.1011709320979325,
+      "eval_loss": 0.3818909823894501,
+      "eval_runtime": 14.0857,
+      "eval_samples_per_second": 116.927,
+      "eval_steps_per_second": 1.846,
+      "eval_wer": 0.4059850883464406,
       "step": 1500
     },
     {
       "epoch": 27.29,
-      "learning_rate": 0.0002782222222222222,
-      "loss": 0.3785,
+      "learning_rate": 0.0003623448275862069,
+      "loss": 0.3948,
       "step": 1501
     },
     {
       "epoch": 27.31,
-      "learning_rate": 0.00027800000000000004,
-      "loss": 0.4128,
+      "learning_rate": 0.00036220689655172415,
+      "loss": 0.4366,
       "step": 1502
     },
     {
       "epoch": 27.33,
-      "learning_rate": 0.0002777777777777778,
-      "loss": 0.3695,
+      "learning_rate": 0.0003620689655172414,
+      "loss": 0.4099,
       "step": 1503
     },
     {
       "epoch": 27.35,
-      "learning_rate": 0.0002775555555555556,
-      "loss": 0.3762,
+      "learning_rate": 0.0003619310344827586,
+      "loss": 0.4204,
       "step": 1504
     },
     {
       "epoch": 27.36,
-      "learning_rate": 0.0002773333333333333,
-      "loss": 0.3923,
+      "learning_rate": 0.0003617931034482759,
+      "loss": 0.4469,
       "step": 1505
     },
     {
       "epoch": 27.38,
-      "learning_rate": 0.0002771111111111111,
-      "loss": 0.37,
+      "learning_rate": 0.0003616551724137931,
+      "loss": 0.4186,
       "step": 1506
     },
     {
       "epoch": 27.4,
-      "learning_rate": 0.0002768888888888889,
-      "loss": 0.4316,
+      "learning_rate": 0.00036151724137931037,
+      "loss": 0.4865,
       "step": 1507
     },
     {
       "epoch": 27.42,
-      "learning_rate": 0.00027666666666666665,
-      "loss": 0.4314,
+      "learning_rate": 0.0003613793103448276,
+      "loss": 0.486,
       "step": 1508
     },
     {
       "epoch": 27.44,
-      "learning_rate": 0.00027644444444444445,
-      "loss": 0.4222,
+      "learning_rate": 0.00036124137931034484,
+      "loss": 0.4347,
       "step": 1509
     },
     {
       "epoch": 27.45,
-      "learning_rate": 0.0002762222222222222,
-      "loss": 0.388,
+      "learning_rate": 0.00036110344827586206,
+      "loss": 0.4759,
       "step": 1510
     },
     {
       "epoch": 27.47,
-      "learning_rate": 0.00027600000000000004,
-      "loss": 0.4004,
+      "learning_rate": 0.0003609655172413793,
+      "loss": 0.4186,
       "step": 1511
     },
     {
       "epoch": 27.49,
-      "learning_rate": 0.0002757777777777778,
-      "loss": 0.4431,
+      "learning_rate": 0.00036082758620689653,
+      "loss": 0.4856,
       "step": 1512
     },
     {
       "epoch": 27.51,
-      "learning_rate": 0.0002755555555555556,
-      "loss": 0.3421,
+      "learning_rate": 0.0003606896551724138,
+      "loss": 0.3937,
       "step": 1513
     },
     {
       "epoch": 27.53,
-      "learning_rate": 0.0002753333333333333,
-      "loss": 0.4423,
+      "learning_rate": 0.000360551724137931,
+      "loss": 0.4658,
       "step": 1514
     },
     {
       "epoch": 27.55,
-      "learning_rate": 0.00027511111111111106,
-      "loss": 0.3721,
+      "learning_rate": 0.0003604137931034483,
+      "loss": 0.4066,
       "step": 1515
     },
     {
       "epoch": 27.56,
-      "learning_rate": 0.0002748888888888889,
-      "loss": 0.3829,
+      "learning_rate": 0.0003602758620689655,
+      "loss": 0.4441,
       "step": 1516
     },
     {
       "epoch": 27.58,
-      "learning_rate": 0.00027466666666666666,
-      "loss": 0.421,
+      "learning_rate": 0.00036013793103448276,
+      "loss": 0.4578,
       "step": 1517
     },
     {
       "epoch": 27.6,
-      "learning_rate": 0.00027444444444444445,
-      "loss": 0.3233,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 0.3771,
       "step": 1518
     },
     {
       "epoch": 27.62,
-      "learning_rate": 0.0002742222222222222,
-      "loss": 0.3462,
+      "learning_rate": 0.0003598620689655173,
+      "loss": 0.4175,
       "step": 1519
     },
     {
       "epoch": 27.64,
-      "learning_rate": 0.00027400000000000005,
-      "loss": 0.4082,
+      "learning_rate": 0.0003597241379310345,
+      "loss": 0.4995,
       "step": 1520
     },
     {
       "epoch": 27.65,
-      "learning_rate": 0.0002737777777777778,
-      "loss": 0.3764,
+      "learning_rate": 0.00035958620689655177,
+      "loss": 0.4092,
       "step": 1521
     },
     {
       "epoch": 27.67,
-      "learning_rate": 0.00027355555555555553,
-      "loss": 0.4119,
+      "learning_rate": 0.000359448275862069,
+      "loss": 0.4649,
       "step": 1522
     },
     {
       "epoch": 27.69,
-      "learning_rate": 0.00027333333333333333,
-      "loss": 0.3415,
+      "learning_rate": 0.00035931034482758624,
+      "loss": 0.3914,
       "step": 1523
     },
     {
       "epoch": 27.71,
-      "learning_rate": 0.0002731111111111111,
-      "loss": 0.3544,
+      "learning_rate": 0.00035917241379310346,
+      "loss": 0.3888,
       "step": 1524
     },
     {
       "epoch": 27.73,
-      "learning_rate": 0.0002728888888888889,
-      "loss": 0.3673,
+      "learning_rate": 0.0003590344827586207,
+      "loss": 0.4395,
       "step": 1525
     },
     {
       "epoch": 27.75,
-      "learning_rate": 0.00027266666666666666,
-      "loss": 0.414,
+      "learning_rate": 0.00035889655172413793,
+      "loss": 0.4837,
       "step": 1526
     },
     {
       "epoch": 27.76,
-      "learning_rate": 0.00027244444444444446,
-      "loss": 0.4025,
+      "learning_rate": 0.0003587586206896552,
+      "loss": 0.4308,
       "step": 1527
     },
     {
       "epoch": 27.78,
-      "learning_rate": 0.0002722222222222222,
-      "loss": 0.3294,
+      "learning_rate": 0.0003586206896551724,
+      "loss": 0.3716,
       "step": 1528
     },
     {
       "epoch": 27.8,
-      "learning_rate": 0.00027200000000000005,
-      "loss": 0.4318,
+      "learning_rate": 0.0003584827586206897,
+      "loss": 0.5344,
       "step": 1529
     },
     {
       "epoch": 27.82,
-      "learning_rate": 0.0002717777777777778,
-      "loss": 0.4454,
+      "learning_rate": 0.0003583448275862069,
+      "loss": 0.4593,
       "step": 1530
     },
     {
       "epoch": 27.84,
-      "learning_rate": 0.00027155555555555554,
-      "loss": 0.3414,
+      "learning_rate": 0.00035820689655172416,
+      "loss": 0.4505,
       "step": 1531
     },
     {
       "epoch": 27.85,
-      "learning_rate": 0.00027133333333333333,
-      "loss": 0.438,
+      "learning_rate": 0.00035806896551724137,
+      "loss": 0.4895,
       "step": 1532
     },
     {
       "epoch": 27.87,
-      "learning_rate": 0.00027111111111111113,
-      "loss": 0.3477,
+      "learning_rate": 0.00035793103448275863,
+      "loss": 0.4203,
       "step": 1533
     },
     {
       "epoch": 27.89,
-      "learning_rate": 0.0002708888888888889,
-      "loss": 0.4351,
+      "learning_rate": 0.00035779310344827585,
+      "loss": 0.52,
       "step": 1534
     },
     {
       "epoch": 27.91,
-      "learning_rate": 0.00027066666666666667,
-      "loss": 0.3463,
+      "learning_rate": 0.00035765517241379306,
+      "loss": 0.4233,
       "step": 1535
     },
     {
       "epoch": 27.93,
-      "learning_rate": 0.0002704444444444444,
-      "loss": 0.4222,
+      "learning_rate": 0.0003575172413793104,
+      "loss": 0.4544,
       "step": 1536
     },
     {
       "epoch": 27.95,
-      "learning_rate": 0.0002702222222222222,
-      "loss": 0.3914,
+      "learning_rate": 0.0003573793103448276,
+      "loss": 0.4479,
       "step": 1537
     },
     {
       "epoch": 27.96,
-      "learning_rate": 0.00027,
-      "loss": 0.3499,
+      "learning_rate": 0.00035724137931034486,
+      "loss": 0.4455,
       "step": 1538
     },
     {
       "epoch": 27.98,
-      "learning_rate": 0.0002697777777777778,
-      "loss": 0.3583,
+      "learning_rate": 0.00035710344827586207,
+      "loss": 0.4616,
       "step": 1539
     },
     {
       "epoch": 28.0,
-      "learning_rate": 0.00026955555555555554,
-      "loss": 0.2918,
+      "learning_rate": 0.00035696551724137933,
+      "loss": 0.4045,
       "step": 1540
     },
     {
       "epoch": 28.02,
-      "learning_rate": 0.00026933333333333334,
-      "loss": 0.3882,
+      "learning_rate": 0.00035682758620689655,
+      "loss": 0.4187,
       "step": 1541
     },
     {
       "epoch": 28.04,
-      "learning_rate": 0.00026911111111111114,
-      "loss": 0.4008,
+      "learning_rate": 0.0003566896551724138,
+      "loss": 0.4405,
       "step": 1542
     },
     {
       "epoch": 28.05,
-      "learning_rate": 0.00026888888888888893,
-      "loss": 0.3907,
+      "learning_rate": 0.000356551724137931,
+      "loss": 0.4545,
       "step": 1543
     },
     {
       "epoch": 28.07,
-      "learning_rate": 0.0002686666666666667,
-      "loss": 0.3367,
+      "learning_rate": 0.0003564137931034483,
+      "loss": 0.4008,
       "step": 1544
     },
     {
       "epoch": 28.09,
-      "learning_rate": 0.0002684444444444444,
-      "loss": 0.3292,
+      "learning_rate": 0.0003562758620689655,
+      "loss": 0.4283,
       "step": 1545
     },
     {
       "epoch": 28.11,
-      "learning_rate": 0.0002682222222222222,
-      "loss": 0.3574,
+      "learning_rate": 0.00035613793103448277,
+      "loss": 0.4222,
       "step": 1546
     },
     {
       "epoch": 28.13,
-      "learning_rate": 0.000268,
-      "loss": 0.3836,
+      "learning_rate": 0.000356,
+      "loss": 0.4188,
       "step": 1547
     },
     {
       "epoch": 28.15,
-      "learning_rate": 0.0002677777777777778,
-      "loss": 0.3612,
+      "learning_rate": 0.00035586206896551725,
+      "loss": 0.4273,
       "step": 1548
     },
     {
       "epoch": 28.16,
-      "learning_rate": 0.00026755555555555555,
-      "loss": 0.3995,
+      "learning_rate": 0.00035572413793103446,
+      "loss": 0.4745,
       "step": 1549
     },
     {
       "epoch": 28.18,
-      "learning_rate": 0.00026733333333333334,
-      "loss": 0.3558,
+      "learning_rate": 0.0003555862068965517,
+      "loss": 0.462,
       "step": 1550
     },
     {
       "epoch": 28.2,
-      "learning_rate": 0.00026711111111111114,
-      "loss": 0.3731,
+      "learning_rate": 0.00035544827586206894,
+      "loss": 0.4237,
       "step": 1551
     },
     {
       "epoch": 28.22,
-      "learning_rate": 0.0002668888888888889,
-      "loss": 0.3966,
+      "learning_rate": 0.00035531034482758626,
+      "loss": 0.4449,
       "step": 1552
     },
     {
       "epoch": 28.24,
-      "learning_rate": 0.0002666666666666667,
-      "loss": 0.4063,
+      "learning_rate": 0.00035517241379310347,
+      "loss": 0.4982,
       "step": 1553
     },
     {
       "epoch": 28.25,
-      "learning_rate": 0.0002664444444444444,
-      "loss": 0.4044,
+      "learning_rate": 0.00035503448275862073,
+      "loss": 0.4467,
       "step": 1554
     },
     {
       "epoch": 28.27,
-      "learning_rate": 0.00026622222222222227,
-      "loss": 0.3613,
+      "learning_rate": 0.00035489655172413795,
+      "loss": 0.4062,
       "step": 1555
     },
     {
       "epoch": 28.29,
-      "learning_rate": 0.000266,
-      "loss": 0.3572,
+      "learning_rate": 0.0003547586206896552,
+      "loss": 0.4366,
       "step": 1556
     },
     {
       "epoch": 28.31,
-      "learning_rate": 0.0002657777777777778,
-      "loss": 0.4295,
+      "learning_rate": 0.0003546206896551724,
+      "loss": 0.4942,
       "step": 1557
     },
     {
       "epoch": 28.33,
-      "learning_rate": 0.00026555555555555555,
-      "loss": 0.4183,
+      "learning_rate": 0.0003544827586206897,
+      "loss": 0.4628,
       "step": 1558
     },
     {
       "epoch": 28.35,
-      "learning_rate": 0.0002653333333333333,
-      "loss": 0.385,
+      "learning_rate": 0.0003543448275862069,
+      "loss": 0.4381,
       "step": 1559
     },
     {
       "epoch": 28.36,
-      "learning_rate": 0.00026511111111111115,
-      "loss": 0.4036,
+      "learning_rate": 0.00035420689655172417,
+      "loss": 0.4572,
       "step": 1560
     },
     {
       "epoch": 28.38,
-      "learning_rate": 0.0002648888888888889,
-      "loss": 0.3205,
+      "learning_rate": 0.0003540689655172414,
+      "loss": 0.3861,
       "step": 1561
     },
     {
       "epoch": 28.4,
-      "learning_rate": 0.0002646666666666667,
-      "loss": 0.3932,
+      "learning_rate": 0.00035393103448275865,
+      "loss": 0.408,
       "step": 1562
     },
     {
       "epoch": 28.42,
-      "learning_rate": 0.00026444444444444443,
-      "loss": 0.3194,
+      "learning_rate": 0.00035379310344827586,
+      "loss": 0.3481,
       "step": 1563
     },
     {
       "epoch": 28.44,
-      "learning_rate": 0.0002642222222222223,
-      "loss": 0.3884,
+      "learning_rate": 0.0003536551724137931,
+      "loss": 0.4294,
       "step": 1564
     },
     {
       "epoch": 28.45,
-      "learning_rate": 0.000264,
-      "loss": 0.3718,
+      "learning_rate": 0.00035351724137931034,
+      "loss": 0.4143,
       "step": 1565
     },
     {
       "epoch": 28.47,
-      "learning_rate": 0.00026377777777777776,
-      "loss": 0.394,
+      "learning_rate": 0.0003533793103448276,
+      "loss": 0.4562,
       "step": 1566
     },
     {
       "epoch": 28.49,
-      "learning_rate": 0.00026355555555555556,
-      "loss": 0.346,
+      "learning_rate": 0.0003532413793103448,
+      "loss": 0.4227,
       "step": 1567
     },
     {
       "epoch": 28.51,
-      "learning_rate": 0.0002633333333333333,
-      "loss": 0.3939,
+      "learning_rate": 0.000353103448275862,
+      "loss": 0.4627,
       "step": 1568
     },
     {
       "epoch": 28.53,
-      "learning_rate": 0.00026311111111111115,
-      "loss": 0.3424,
+      "learning_rate": 0.00035296551724137935,
+      "loss": 0.4332,
       "step": 1569
     },
     {
       "epoch": 28.55,
-      "learning_rate": 0.0002628888888888889,
-      "loss": 0.3354,
+      "learning_rate": 0.00035282758620689656,
+      "loss": 0.3763,
       "step": 1570
     },
     {
       "epoch": 28.56,
-      "learning_rate": 0.0002626666666666667,
-      "loss": 0.372,
+      "learning_rate": 0.0003526896551724138,
+      "loss": 0.4199,
       "step": 1571
     },
     {
       "epoch": 28.58,
-      "learning_rate": 0.00026244444444444443,
-      "loss": 0.329,
+      "learning_rate": 0.00035255172413793104,
+      "loss": 0.4024,
       "step": 1572
     },
     {
       "epoch": 28.6,
-      "learning_rate": 0.00026222222222222223,
-      "loss": 0.4093,
+      "learning_rate": 0.0003524137931034483,
+      "loss": 0.4698,
       "step": 1573
     },
     {
       "epoch": 28.62,
-      "learning_rate": 0.000262,
-      "loss": 0.3668,
+      "learning_rate": 0.0003522758620689655,
+      "loss": 0.4229,
       "step": 1574
     },
     {
       "epoch": 28.64,
-      "learning_rate": 0.00026177777777777777,
-      "loss": 0.4422,
+      "learning_rate": 0.0003521379310344828,
+      "loss": 0.483,
       "step": 1575
     },
     {
       "epoch": 28.65,
-      "learning_rate": 0.00026155555555555556,
-      "loss": 0.3845,
+      "learning_rate": 0.000352,
+      "loss": 0.4532,
       "step": 1576
     },
     {
       "epoch": 28.67,
-      "learning_rate": 0.0002613333333333333,
-      "loss": 0.3696,
+      "learning_rate": 0.00035186206896551726,
+      "loss": 0.4405,
       "step": 1577
     },
     {
       "epoch": 28.69,
-      "learning_rate": 0.00026111111111111116,
-      "loss": 0.3126,
+      "learning_rate": 0.00035172413793103447,
+      "loss": 0.36,
       "step": 1578
     },
     {
       "epoch": 28.71,
-      "learning_rate": 0.0002608888888888889,
-      "loss": 0.3295,
+      "learning_rate": 0.00035158620689655173,
+      "loss": 0.3772,
       "step": 1579
     },
     {
       "epoch": 28.73,
-      "learning_rate": 0.00026066666666666664,
-      "loss": 0.3388,
+      "learning_rate": 0.00035144827586206895,
+      "loss": 0.4053,
       "step": 1580
     },
     {
       "epoch": 28.75,
-      "learning_rate": 0.00026044444444444444,
-      "loss": 0.3499,
+      "learning_rate": 0.0003513103448275862,
+      "loss": 0.3634,
       "step": 1581
     },
     {
       "epoch": 28.76,
-      "learning_rate": 0.00026022222222222223,
-      "loss": 0.3889,
+      "learning_rate": 0.0003511724137931034,
+      "loss": 0.437,
       "step": 1582
     },
     {
       "epoch": 28.78,
-      "learning_rate": 0.00026000000000000003,
-      "loss": 0.3649,
+      "learning_rate": 0.0003510344827586207,
+      "loss": 0.384,
       "step": 1583
     },
     {
       "epoch": 28.8,
-      "learning_rate": 0.0002597777777777778,
-      "loss": 0.3883,
+      "learning_rate": 0.0003508965517241379,
+      "loss": 0.431,
       "step": 1584
     },
     {
       "epoch": 28.82,
-      "learning_rate": 0.00025955555555555557,
-      "loss": 0.4425,
+      "learning_rate": 0.00035075862068965517,
+      "loss": 0.4416,
       "step": 1585
     },
     {
       "epoch": 28.84,
-      "learning_rate": 0.00025933333333333337,
-      "loss": 0.3545,
+      "learning_rate": 0.00035062068965517243,
+      "loss": 0.4037,
       "step": 1586
     },
     {
       "epoch": 28.85,
-      "learning_rate": 0.0002591111111111111,
-      "loss": 0.355,
+      "learning_rate": 0.0003504827586206897,
+      "loss": 0.3903,
       "step": 1587
     },
     {
       "epoch": 28.87,
-      "learning_rate": 0.0002588888888888889,
-      "loss": 0.3615,
+      "learning_rate": 0.0003503448275862069,
+      "loss": 0.4303,
       "step": 1588
     },
     {
       "epoch": 28.89,
-      "learning_rate": 0.00025866666666666665,
-      "loss": 0.3521,
+      "learning_rate": 0.0003502068965517242,
+      "loss": 0.4129,
       "step": 1589
     },
     {
       "epoch": 28.91,
-      "learning_rate": 0.00025844444444444444,
-      "loss": 0.3767,
+      "learning_rate": 0.0003500689655172414,
+      "loss": 0.4732,
       "step": 1590
     },
     {
       "epoch": 28.93,
-      "learning_rate": 0.00025822222222222224,
-      "loss": 0.4132,
+      "learning_rate": 0.00034993103448275866,
+      "loss": 0.4782,
       "step": 1591
     },
     {
       "epoch": 28.95,
-      "learning_rate": 0.00025800000000000004,
-      "loss": 0.4013,
+      "learning_rate": 0.00034979310344827587,
+      "loss": 0.4379,
       "step": 1592
     },
     {
       "epoch": 28.96,
-      "learning_rate": 0.0002577777777777778,
-      "loss": 0.3759,
+      "learning_rate": 0.00034965517241379313,
+      "loss": 0.4639,
       "step": 1593
     },
     {
       "epoch": 28.98,
-      "learning_rate": 0.0002575555555555555,
-      "loss": 0.3583,
+      "learning_rate": 0.00034951724137931035,
+      "loss": 0.4052,
       "step": 1594
     },
     {
       "epoch": 29.0,
-      "learning_rate": 0.00025733333333333337,
-      "loss": 0.3787,
+      "learning_rate": 0.0003493793103448276,
+      "loss": 0.4448,
       "step": 1595
     },
     {
       "epoch": 29.02,
-      "learning_rate": 0.0002571111111111111,
-      "loss": 0.3744,
+      "learning_rate": 0.0003492413793103448,
+      "loss": 0.4383,
       "step": 1596
     },
     {
       "epoch": 29.04,
-      "learning_rate": 0.0002568888888888889,
-      "loss": 0.361,
+      "learning_rate": 0.0003491034482758621,
+      "loss": 0.4158,
       "step": 1597
     },
     {
       "epoch": 29.05,
-      "learning_rate": 0.00025666666666666665,
-      "loss": 0.3503,
+      "learning_rate": 0.0003489655172413793,
+      "loss": 0.401,
       "step": 1598
     },
     {
       "epoch": 29.07,
-      "learning_rate": 0.00025644444444444445,
-      "loss": 0.4261,
+      "learning_rate": 0.00034882758620689657,
+      "loss": 0.49,
       "step": 1599
     },
     {
       "epoch": 29.09,
-      "learning_rate": 0.00025622222222222225,
-      "loss": 0.351,
+      "learning_rate": 0.0003486896551724138,
+      "loss": 0.4027,
       "step": 1600
     },
     {
       "epoch": 29.11,
-      "learning_rate": 0.000256,
-      "loss": 0.3691,
+      "learning_rate": 0.000348551724137931,
+      "loss": 0.4325,
       "step": 1601
     },
     {
       "epoch": 29.13,
-      "learning_rate": 0.0002557777777777778,
-      "loss": 0.372,
+      "learning_rate": 0.00034841379310344826,
+      "loss": 0.3991,
       "step": 1602
     },
     {
       "epoch": 29.15,
-      "learning_rate": 0.00025555555555555553,
-      "loss": 0.3781,
+      "learning_rate": 0.0003482758620689655,
+      "loss": 0.4509,
       "step": 1603
     },
     {
       "epoch": 29.16,
-      "learning_rate": 0.0002553333333333334,
-      "loss": 0.3368,
+      "learning_rate": 0.0003481379310344828,
+      "loss": 0.4093,
       "step": 1604
     },
     {
       "epoch": 29.18,
-      "learning_rate": 0.0002551111111111111,
-      "loss": 0.364,
+      "learning_rate": 0.000348,
+      "loss": 0.3979,
       "step": 1605
     },
     {
       "epoch": 29.2,
-      "learning_rate": 0.0002548888888888889,
-      "loss": 0.3829,
+      "learning_rate": 0.00034786206896551727,
+      "loss": 0.4297,
       "step": 1606
     },
     {
       "epoch": 29.22,
-      "learning_rate": 0.00025466666666666666,
-      "loss": 0.3557,
+      "learning_rate": 0.0003477241379310345,
+      "loss": 0.4236,
       "step": 1607
     },
     {
       "epoch": 29.24,
-      "learning_rate": 0.0002544444444444444,
-      "loss": 0.3604,
+      "learning_rate": 0.00034758620689655175,
+      "loss": 0.4492,
       "step": 1608
     },
     {
       "epoch": 29.25,
-      "learning_rate": 0.00025422222222222225,
-      "loss": 0.3565,
+      "learning_rate": 0.00034744827586206896,
+      "loss": 0.4086,
       "step": 1609
     },
     {
       "epoch": 29.27,
-      "learning_rate": 0.000254,
-      "loss": 0.3402,
+      "learning_rate": 0.0003473103448275862,
+      "loss": 0.3809,
       "step": 1610
     },
     {
       "epoch": 29.29,
-      "learning_rate": 0.0002537777777777778,
-      "loss": 0.3017,
+      "learning_rate": 0.00034717241379310344,
+      "loss": 0.3748,
       "step": 1611
     },
     {
       "epoch": 29.31,
-      "learning_rate": 0.00025355555555555553,
-      "loss": 0.3402,
+      "learning_rate": 0.0003470344827586207,
+      "loss": 0.3793,
       "step": 1612
     },
     {
       "epoch": 29.33,
-      "learning_rate": 0.0002533333333333334,
-      "loss": 0.346,
+      "learning_rate": 0.0003468965517241379,
+      "loss": 0.4506,
       "step": 1613
     },
     {
       "epoch": 29.35,
-      "learning_rate": 0.0002531111111111111,
-      "loss": 0.3673,
+      "learning_rate": 0.0003467586206896552,
+      "loss": 0.4544,
       "step": 1614
     },
     {
       "epoch": 29.36,
-      "learning_rate": 0.00025288888888888887,
-      "loss": 0.3646,
+      "learning_rate": 0.0003466206896551724,
+      "loss": 0.4356,
       "step": 1615
     },
     {
       "epoch": 29.38,
-      "learning_rate": 0.00025266666666666666,
-      "loss": 0.4031,
+      "learning_rate": 0.00034648275862068966,
+      "loss": 0.4409,
       "step": 1616
     },
     {
       "epoch": 29.4,
-      "learning_rate": 0.00025244444444444446,
-      "loss": 0.355,
+      "learning_rate": 0.00034634482758620687,
+      "loss": 0.4203,
       "step": 1617
     },
     {
       "epoch": 29.42,
-      "learning_rate": 0.00025222222222222226,
-      "loss": 0.3214,
+      "learning_rate": 0.00034620689655172414,
+      "loss": 0.4151,
       "step": 1618
     },
     {
       "epoch": 29.44,
-      "learning_rate": 0.000252,
-      "loss": 0.3617,
+      "learning_rate": 0.00034606896551724135,
+      "loss": 0.4501,
       "step": 1619
     },
     {
       "epoch": 29.45,
-      "learning_rate": 0.0002517777777777778,
-      "loss": 0.3024,
+      "learning_rate": 0.00034593103448275867,
+      "loss": 0.3621,
       "step": 1620
     },
     {
       "epoch": 29.47,
-      "learning_rate": 0.00025155555555555554,
-      "loss": 0.3429,
+      "learning_rate": 0.0003457931034482759,
+      "loss": 0.444,
       "step": 1621
     },
     {
       "epoch": 29.49,
-      "learning_rate": 0.00025133333333333333,
-      "loss": 0.3326,
+      "learning_rate": 0.00034565517241379315,
+      "loss": 0.3833,
       "step": 1622
     },
     {
       "epoch": 29.51,
-      "learning_rate": 0.00025111111111111113,
-      "loss": 0.3495,
+      "learning_rate": 0.00034551724137931036,
+      "loss": 0.4444,
       "step": 1623
     },
     {
       "epoch": 29.53,
-      "learning_rate": 0.0002508888888888889,
-      "loss": 0.3734,
+      "learning_rate": 0.0003453793103448276,
+      "loss": 0.418,
       "step": 1624
     },
     {
       "epoch": 29.55,
-      "learning_rate": 0.00025066666666666667,
-      "loss": 0.399,
+      "learning_rate": 0.00034524137931034484,
+      "loss": 0.4895,
       "step": 1625
     },
     {
       "epoch": 29.56,
-      "learning_rate": 0.00025044444444444447,
-      "loss": 0.3186,
+      "learning_rate": 0.0003451034482758621,
+      "loss": 0.3954,
       "step": 1626
     },
     {
       "epoch": 29.58,
-      "learning_rate": 0.00025022222222222226,
-      "loss": 0.3744,
+      "learning_rate": 0.0003449655172413793,
+      "loss": 0.4279,
       "step": 1627
     },
     {
       "epoch": 29.6,
-      "learning_rate": 0.00025,
-      "loss": 0.3904,
+      "learning_rate": 0.0003448275862068966,
+      "loss": 0.4148,
       "step": 1628
     },
     {
       "epoch": 29.62,
-      "learning_rate": 0.00024977777777777775,
-      "loss": 0.2997,
+      "learning_rate": 0.0003446896551724138,
+      "loss": 0.3863,
       "step": 1629
     },
     {
       "epoch": 29.64,
-      "learning_rate": 0.00024955555555555554,
-      "loss": 0.3339,
+      "learning_rate": 0.00034455172413793106,
+      "loss": 0.4135,
       "step": 1630
     },
     {
       "epoch": 29.65,
-      "learning_rate": 0.00024933333333333334,
-      "loss": 0.3487,
+      "learning_rate": 0.00034441379310344827,
+      "loss": 0.4474,
       "step": 1631
     },
     {
       "epoch": 29.67,
-      "learning_rate": 0.00024911111111111114,
-      "loss": 0.3556,
+      "learning_rate": 0.00034427586206896554,
+      "loss": 0.4192,
       "step": 1632
     },
     {
       "epoch": 29.69,
-      "learning_rate": 0.0002488888888888889,
-      "loss": 0.3321,
+      "learning_rate": 0.00034413793103448275,
+      "loss": 0.3851,
       "step": 1633
     },
     {
       "epoch": 29.71,
-      "learning_rate": 0.0002486666666666667,
-      "loss": 0.3867,
+      "learning_rate": 0.00034399999999999996,
+      "loss": 0.4409,
       "step": 1634
     },
     {
       "epoch": 29.73,
-      "learning_rate": 0.00024844444444444447,
-      "loss": 0.2982,
+      "learning_rate": 0.0003438620689655172,
+      "loss": 0.3677,
       "step": 1635
     },
     {
       "epoch": 29.75,
-      "learning_rate": 0.0002482222222222222,
-      "loss": 0.4175,
+      "learning_rate": 0.00034372413793103444,
+      "loss": 0.5136,
       "step": 1636
     },
     {
       "epoch": 29.76,
-      "learning_rate": 0.000248,
-      "loss": 0.3688,
+      "learning_rate": 0.00034358620689655176,
+      "loss": 0.4849,
       "step": 1637
     },
     {
       "epoch": 29.78,
-      "learning_rate": 0.0002477777777777778,
-      "loss": 0.3804,
+      "learning_rate": 0.00034344827586206897,
+      "loss": 0.4536,
       "step": 1638
     },
     {
       "epoch": 29.8,
-      "learning_rate": 0.00024755555555555555,
-      "loss": 0.4076,
+      "learning_rate": 0.00034331034482758624,
+      "loss": 0.4637,
       "step": 1639
     },
     {
       "epoch": 29.82,
-      "learning_rate": 0.00024733333333333335,
-      "loss": 0.3785,
+      "learning_rate": 0.00034317241379310345,
+      "loss": 0.4664,
       "step": 1640
     },
     {
       "epoch": 29.84,
-      "learning_rate": 0.00024711111111111114,
-      "loss": 0.3409,
+      "learning_rate": 0.0003430344827586207,
+      "loss": 0.4633,
       "step": 1641
     },
     {
       "epoch": 29.85,
-      "learning_rate": 0.0002468888888888889,
-      "loss": 0.3866,
+      "learning_rate": 0.0003428965517241379,
+      "loss": 0.4363,
       "step": 1642
     },
     {
       "epoch": 29.87,
-      "learning_rate": 0.0002466666666666667,
-      "loss": 0.3254,
+      "learning_rate": 0.0003427586206896552,
+      "loss": 0.4114,
       "step": 1643
     },
     {
       "epoch": 29.89,
-      "learning_rate": 0.0002464444444444444,
-      "loss": 0.3355,
+      "learning_rate": 0.0003426206896551724,
+      "loss": 0.3893,
       "step": 1644
     },
     {
       "epoch": 29.91,
-      "learning_rate": 0.0002462222222222222,
-      "loss": 0.3362,
+      "learning_rate": 0.00034248275862068967,
+      "loss": 0.4097,
       "step": 1645
     },
     {
       "epoch": 29.93,
-      "learning_rate": 0.000246,
-      "loss": 0.3271,
+      "learning_rate": 0.0003423448275862069,
+      "loss": 0.3983,
       "step": 1646
     },
     {
       "epoch": 29.95,
-      "learning_rate": 0.0002457777777777778,
-      "loss": 0.3545,
+      "learning_rate": 0.00034220689655172415,
+      "loss": 0.4497,
       "step": 1647
     },
     {
       "epoch": 29.96,
-      "learning_rate": 0.00024555555555555556,
-      "loss": 0.3913,
+      "learning_rate": 0.00034206896551724136,
+      "loss": 0.4614,
       "step": 1648
     },
     {
       "epoch": 29.98,
-      "learning_rate": 0.00024533333333333335,
-      "loss": 0.3862,
+      "learning_rate": 0.0003419310344827586,
+      "loss": 0.464,
       "step": 1649
     },
     {
       "epoch": 30.0,
-      "learning_rate": 0.0002451111111111111,
-      "loss": 0.2773,
+      "learning_rate": 0.00034179310344827584,
+      "loss": 0.3383,
       "step": 1650
     },
     {
       "epoch": 30.02,
-      "learning_rate": 0.0002448888888888889,
-      "loss": 0.3448,
+      "learning_rate": 0.0003416551724137931,
+      "loss": 0.3942,
       "step": 1651
     },
     {
       "epoch": 30.04,
-      "learning_rate": 0.0002446666666666667,
-      "loss": 0.3371,
+      "learning_rate": 0.0003415172413793103,
+      "loss": 0.4242,
       "step": 1652
     },
     {
       "epoch": 30.05,
-      "learning_rate": 0.00024444444444444443,
-      "loss": 0.3382,
+      "learning_rate": 0.00034137931034482763,
+      "loss": 0.4026,
       "step": 1653
     },
     {
       "epoch": 30.07,
-      "learning_rate": 0.0002442222222222222,
-      "loss": 0.431,
+      "learning_rate": 0.00034124137931034485,
+      "loss": 0.5147,
       "step": 1654
     },
     {
       "epoch": 30.09,
-      "learning_rate": 0.000244,
-      "loss": 0.3955,
+      "learning_rate": 0.0003411034482758621,
+      "loss": 0.4638,
       "step": 1655
     },
     {
       "epoch": 30.11,
-      "learning_rate": 0.0002437777777777778,
-      "loss": 0.3325,
+      "learning_rate": 0.0003409655172413793,
+      "loss": 0.4236,
       "step": 1656
     },
     {
       "epoch": 30.13,
-      "learning_rate": 0.00024355555555555556,
-      "loss": 0.3018,
+      "learning_rate": 0.0003408275862068966,
+      "loss": 0.4012,
       "step": 1657
     },
     {
       "epoch": 30.15,
-      "learning_rate": 0.00024333333333333336,
-      "loss": 0.3264,
+      "learning_rate": 0.0003406896551724138,
+      "loss": 0.4043,
       "step": 1658
     },
     {
       "epoch": 30.16,
-      "learning_rate": 0.00024311111111111113,
-      "loss": 0.3625,
+      "learning_rate": 0.00034055172413793107,
+      "loss": 0.4381,
       "step": 1659
     },
     {
       "epoch": 30.18,
-      "learning_rate": 0.00024288888888888887,
-      "loss": 0.3966,
+      "learning_rate": 0.0003404137931034483,
+      "loss": 0.4777,
       "step": 1660
     },
     {
       "epoch": 30.2,
-      "learning_rate": 0.00024266666666666667,
-      "loss": 0.3572,
+      "learning_rate": 0.00034027586206896555,
+      "loss": 0.4059,
       "step": 1661
     },
     {
       "epoch": 30.22,
-      "learning_rate": 0.00024244444444444443,
-      "loss": 0.3944,
+      "learning_rate": 0.00034013793103448276,
+      "loss": 0.4942,
       "step": 1662
     },
     {
       "epoch": 30.24,
-      "learning_rate": 0.00024222222222222223,
-      "loss": 0.3829,
+      "learning_rate": 0.00034,
+      "loss": 0.4978,
       "step": 1663
     },
     {
       "epoch": 30.25,
-      "learning_rate": 0.000242,
-      "loss": 0.3476,
+      "learning_rate": 0.00033986206896551724,
+      "loss": 0.4263,
       "step": 1664
     },
     {
       "epoch": 30.27,
-      "learning_rate": 0.0002417777777777778,
-      "loss": 0.3775,
+      "learning_rate": 0.0003397241379310345,
+      "loss": 0.4565,
       "step": 1665
     },
     {
       "epoch": 30.29,
-      "learning_rate": 0.00024155555555555557,
-      "loss": 0.3597,
+      "learning_rate": 0.0003395862068965517,
+      "loss": 0.4469,
       "step": 1666
     },
     {
       "epoch": 30.31,
-      "learning_rate": 0.00024133333333333334,
-      "loss": 0.32,
+      "learning_rate": 0.0003394482758620689,
+      "loss": 0.3929,
       "step": 1667
     },
     {
       "epoch": 30.33,
-      "learning_rate": 0.0002411111111111111,
-      "loss": 0.4363,
+      "learning_rate": 0.0003393103448275862,
+      "loss": 0.4846,
       "step": 1668
     },
     {
       "epoch": 30.35,
-      "learning_rate": 0.0002408888888888889,
-      "loss": 0.3887,
+      "learning_rate": 0.0003391724137931034,
+      "loss": 0.4386,
       "step": 1669
     },
     {
       "epoch": 30.36,
-      "learning_rate": 0.00024066666666666667,
-      "loss": 0.3429,
+      "learning_rate": 0.0003390344827586207,
+      "loss": 0.3799,
       "step": 1670
     },
     {
       "epoch": 30.38,
-      "learning_rate": 0.00024044444444444444,
-      "loss": 0.3165,
+      "learning_rate": 0.00033889655172413794,
+      "loss": 0.4054,
       "step": 1671
     },
     {
       "epoch": 30.4,
-      "learning_rate": 0.00024022222222222224,
-      "loss": 0.3512,
+      "learning_rate": 0.0003387586206896552,
+      "loss": 0.4269,
       "step": 1672
     },
     {
       "epoch": 30.42,
-      "learning_rate": 0.00024,
-      "loss": 0.3098,
+      "learning_rate": 0.0003386206896551724,
+      "loss": 0.3842,
       "step": 1673
     },
     {
       "epoch": 30.44,
-      "learning_rate": 0.0002397777777777778,
-      "loss": 0.3155,
+      "learning_rate": 0.0003384827586206897,
+      "loss": 0.3859,
       "step": 1674
     },
     {
       "epoch": 30.45,
-      "learning_rate": 0.00023955555555555554,
-      "loss": 0.3022,
+      "learning_rate": 0.0003383448275862069,
+      "loss": 0.3171,
       "step": 1675
     },
     {
       "epoch": 30.47,
-      "learning_rate": 0.00023933333333333334,
-      "loss": 0.3063,
+      "learning_rate": 0.00033820689655172416,
+      "loss": 0.3778,
       "step": 1676
     },
     {
       "epoch": 30.49,
-      "learning_rate": 0.0002391111111111111,
-      "loss": 0.2967,
+      "learning_rate": 0.00033806896551724137,
+      "loss": 0.3845,
       "step": 1677
     },
     {
       "epoch": 30.51,
-      "learning_rate": 0.0002388888888888889,
-      "loss": 0.2859,
+      "learning_rate": 0.00033793103448275864,
+      "loss": 0.3751,
       "step": 1678
     },
     {
       "epoch": 30.53,
-      "learning_rate": 0.00023866666666666668,
-      "loss": 0.306,
+      "learning_rate": 0.00033779310344827585,
+      "loss": 0.3844,
       "step": 1679
     },
     {
       "epoch": 30.55,
-      "learning_rate": 0.00023844444444444445,
-      "loss": 0.3002,
+      "learning_rate": 0.0003376551724137931,
+      "loss": 0.343,
       "step": 1680
     },
     {
       "epoch": 30.56,
-      "learning_rate": 0.00023822222222222224,
-      "loss": 0.3321,
+      "learning_rate": 0.0003375172413793103,
+      "loss": 0.4095,
       "step": 1681
     },
     {
       "epoch": 30.58,
-      "learning_rate": 0.00023799999999999998,
-      "loss": 0.368,
+      "learning_rate": 0.0003373793103448276,
+      "loss": 0.4411,
       "step": 1682
     },
     {
       "epoch": 30.6,
-      "learning_rate": 0.00023777777777777778,
-      "loss": 0.323,
+      "learning_rate": 0.0003372413793103448,
+      "loss": 0.417,
       "step": 1683
     },
     {
       "epoch": 30.62,
-      "learning_rate": 0.00023755555555555555,
-      "loss": 0.3287,
+      "learning_rate": 0.00033710344827586207,
+      "loss": 0.4165,
       "step": 1684
     },
     {
       "epoch": 30.64,
-      "learning_rate": 0.00023733333333333335,
-      "loss": 0.3329,
+      "learning_rate": 0.0003369655172413793,
+      "loss": 0.4126,
       "step": 1685
     },
     {
       "epoch": 30.65,
-      "learning_rate": 0.00023711111111111112,
-      "loss": 0.3213,
+      "learning_rate": 0.00033682758620689655,
+      "loss": 0.4056,
       "step": 1686
     },
     {
       "epoch": 30.67,
-      "learning_rate": 0.0002368888888888889,
-      "loss": 0.2946,
+      "learning_rate": 0.0003366896551724138,
+      "loss": 0.3672,
       "step": 1687
     },
     {
       "epoch": 30.69,
-      "learning_rate": 0.00023666666666666668,
-      "loss": 0.3376,
+      "learning_rate": 0.0003365517241379311,
+      "loss": 0.4317,
       "step": 1688
     },
     {
       "epoch": 30.71,
-      "learning_rate": 0.00023644444444444445,
-      "loss": 0.3268,
+      "learning_rate": 0.0003364137931034483,
+      "loss": 0.4026,
       "step": 1689
     },
     {
       "epoch": 30.73,
-      "learning_rate": 0.00023622222222222222,
-      "loss": 0.3215,
+      "learning_rate": 0.00033627586206896556,
+      "loss": 0.3838,
       "step": 1690
     },
     {
       "epoch": 30.75,
-      "learning_rate": 0.000236,
-      "loss": 0.2826,
+      "learning_rate": 0.00033613793103448277,
+      "loss": 0.3849,
       "step": 1691
     },
     {
       "epoch": 30.76,
-      "learning_rate": 0.0002357777777777778,
-      "loss": 0.3596,
+      "learning_rate": 0.00033600000000000004,
+      "loss": 0.4362,
       "step": 1692
     },
     {
       "epoch": 30.78,
-      "learning_rate": 0.00023555555555555556,
-      "loss": 0.4355,
+      "learning_rate": 0.00033586206896551725,
+      "loss": 0.4962,
       "step": 1693
     },
     {
       "epoch": 30.8,
-      "learning_rate": 0.00023533333333333335,
-      "loss": 0.3316,
+      "learning_rate": 0.0003357241379310345,
+      "loss": 0.427,
       "step": 1694
     },
     {
       "epoch": 30.82,
-      "learning_rate": 0.00023511111111111112,
-      "loss": 0.395,
+      "learning_rate": 0.0003355862068965517,
+      "loss": 0.4689,
       "step": 1695
     },
     {
       "epoch": 30.84,
-      "learning_rate": 0.0002348888888888889,
-      "loss": 0.4225,
+      "learning_rate": 0.000335448275862069,
+      "loss": 0.448,
       "step": 1696
     },
     {
       "epoch": 30.85,
-      "learning_rate": 0.00023466666666666666,
-      "loss": 0.3336,
+      "learning_rate": 0.0003353103448275862,
+      "loss": 0.4157,
       "step": 1697
     },
     {
       "epoch": 30.87,
-      "learning_rate": 0.00023444444444444446,
-      "loss": 0.3539,
+      "learning_rate": 0.00033517241379310347,
+      "loss": 0.4387,
       "step": 1698
     },
     {
       "epoch": 30.89,
-      "learning_rate": 0.00023422222222222223,
-      "loss": 0.3845,
+      "learning_rate": 0.0003350344827586207,
+      "loss": 0.4381,
       "step": 1699
     },
     {
       "epoch": 30.91,
-      "learning_rate": 0.00023400000000000002,
-      "loss": 0.3525,
+      "learning_rate": 0.00033489655172413795,
+      "loss": 0.399,
       "step": 1700
     },
     {
       "epoch": 30.93,
-      "learning_rate": 0.0002337777777777778,
-      "loss": 0.3453,
+      "learning_rate": 0.00033475862068965516,
+      "loss": 0.4133,
       "step": 1701
     },
     {
       "epoch": 30.95,
-      "learning_rate": 0.00023355555555555556,
-      "loss": 0.3489,
+      "learning_rate": 0.00033462068965517237,
+      "loss": 0.4148,
       "step": 1702
     },
     {
       "epoch": 30.96,
-      "learning_rate": 0.00023333333333333333,
-      "loss": 0.3717,
+      "learning_rate": 0.0003344827586206897,
+      "loss": 0.4978,
       "step": 1703
     },
     {
       "epoch": 30.98,
-      "learning_rate": 0.0002331111111111111,
-      "loss": 0.3928,
+      "learning_rate": 0.0003343448275862069,
+      "loss": 0.4795,
       "step": 1704
     },
     {
       "epoch": 31.0,
-      "learning_rate": 0.0002328888888888889,
-      "loss": 0.4324,
+      "learning_rate": 0.00033420689655172417,
+      "loss": 0.5169,
       "step": 1705
     },
     {
       "epoch": 31.02,
-      "learning_rate": 0.00023266666666666667,
-      "loss": 0.3122,
+      "learning_rate": 0.0003340689655172414,
+      "loss": 0.4097,
       "step": 1706
     },
     {
       "epoch": 31.04,
-      "learning_rate": 0.00023244444444444446,
-      "loss": 0.3375,
+      "learning_rate": 0.00033393103448275865,
+      "loss": 0.4039,
       "step": 1707
     },
     {
       "epoch": 31.05,
-      "learning_rate": 0.00023222222222222223,
-      "loss": 0.3475,
+      "learning_rate": 0.00033379310344827586,
+      "loss": 0.4057,
       "step": 1708
     },
     {
       "epoch": 31.07,
-      "learning_rate": 0.00023200000000000003,
-      "loss": 0.3369,
+      "learning_rate": 0.0003336551724137931,
+      "loss": 0.4006,
       "step": 1709
     },
     {
       "epoch": 31.09,
-      "learning_rate": 0.00023177777777777777,
-      "loss": 0.3289,
+      "learning_rate": 0.00033351724137931034,
+      "loss": 0.4297,
       "step": 1710
     },
     {
       "epoch": 31.11,
-      "learning_rate": 0.00023155555555555554,
-      "loss": 0.3352,
+      "learning_rate": 0.0003333793103448276,
+      "loss": 0.42,
       "step": 1711
     },
     {
       "epoch": 31.13,
-      "learning_rate": 0.00023133333333333334,
-      "loss": 0.3333,
+      "learning_rate": 0.0003332413793103448,
+      "loss": 0.395,
       "step": 1712
     },
     {
       "epoch": 31.15,
-      "learning_rate": 0.0002311111111111111,
-      "loss": 0.3249,
+      "learning_rate": 0.0003331034482758621,
+      "loss": 0.432,
       "step": 1713
     },
     {
       "epoch": 31.16,
-      "learning_rate": 0.0002308888888888889,
-      "loss": 0.3219,
+      "learning_rate": 0.0003329655172413793,
+      "loss": 0.3953,
       "step": 1714
     },
     {
       "epoch": 31.18,
-      "learning_rate": 0.00023066666666666667,
-      "loss": 0.3334,
+      "learning_rate": 0.00033282758620689656,
+      "loss": 0.4201,
       "step": 1715
     },
     {
       "epoch": 31.2,
-      "learning_rate": 0.00023044444444444447,
-      "loss": 0.4274,
+      "learning_rate": 0.00033268965517241377,
+      "loss": 0.4609,
       "step": 1716
     },
     {
       "epoch": 31.22,
-      "learning_rate": 0.0002302222222222222,
-      "loss": 0.4649,
+      "learning_rate": 0.00033255172413793104,
+      "loss": 0.496,
       "step": 1717
     },
     {
       "epoch": 31.24,
-      "learning_rate": 0.00023,
-      "loss": 0.3172,
+      "learning_rate": 0.00033241379310344825,
+      "loss": 0.4155,
       "step": 1718
     },
     {
       "epoch": 31.25,
-      "learning_rate": 0.00022977777777777778,
-      "loss": 0.3593,
+      "learning_rate": 0.0003322758620689655,
+      "loss": 0.4203,
       "step": 1719
     },
     {
       "epoch": 31.27,
-      "learning_rate": 0.00022955555555555557,
-      "loss": 0.3585,
+      "learning_rate": 0.0003321379310344828,
+      "loss": 0.4426,
       "step": 1720
     },
     {
       "epoch": 31.29,
-      "learning_rate": 0.00022933333333333334,
-      "loss": 0.3292,
+      "learning_rate": 0.00033200000000000005,
+      "loss": 0.4008,
       "step": 1721
     },
     {
       "epoch": 31.31,
-      "learning_rate": 0.0002291111111111111,
-      "loss": 0.2924,
+      "learning_rate": 0.00033186206896551726,
+      "loss": 0.3853,
       "step": 1722
     },
     {
       "epoch": 31.33,
-      "learning_rate": 0.0002288888888888889,
-      "loss": 0.3355,
+      "learning_rate": 0.0003317241379310345,
+      "loss": 0.4454,
       "step": 1723
     },
     {
       "epoch": 31.35,
-      "learning_rate": 0.00022866666666666665,
-      "loss": 0.3552,
+      "learning_rate": 0.00033158620689655174,
+      "loss": 0.4538,
       "step": 1724
     },
     {
       "epoch": 31.36,
-      "learning_rate": 0.00022844444444444445,
-      "loss": 0.3849,
+      "learning_rate": 0.000331448275862069,
+      "loss": 0.4559,
       "step": 1725
     },
     {
       "epoch": 31.38,
-      "learning_rate": 0.00022822222222222222,
-      "loss": 0.3583,
+      "learning_rate": 0.0003313103448275862,
+      "loss": 0.433,
       "step": 1726
     },
     {
       "epoch": 31.4,
-      "learning_rate": 0.000228,
-      "loss": 0.2557,
+      "learning_rate": 0.0003311724137931035,
+      "loss": 0.3258,
       "step": 1727
     },
     {
       "epoch": 31.42,
-      "learning_rate": 0.00022777777777777778,
-      "loss": 0.3542,
+      "learning_rate": 0.0003310344827586207,
+      "loss": 0.4145,
       "step": 1728
     },
     {
       "epoch": 31.44,
-      "learning_rate": 0.00022755555555555558,
-      "loss": 0.3493,
+      "learning_rate": 0.00033089655172413796,
+      "loss": 0.4406,
       "step": 1729
     },
     {
       "epoch": 31.45,
-      "learning_rate": 0.00022733333333333335,
-      "loss": 0.3371,
+      "learning_rate": 0.00033075862068965517,
+      "loss": 0.3776,
       "step": 1730
     },
     {
       "epoch": 31.47,
-      "learning_rate": 0.00022711111111111112,
-      "loss": 0.3298,
+      "learning_rate": 0.00033062068965517244,
+      "loss": 0.3854,
       "step": 1731
     },
     {
       "epoch": 31.49,
-      "learning_rate": 0.0002268888888888889,
-      "loss": 0.2939,
+      "learning_rate": 0.00033048275862068965,
+      "loss": 0.3835,
       "step": 1732
     },
     {
       "epoch": 31.51,
-      "learning_rate": 0.00022666666666666666,
-      "loss": 0.3609,
+      "learning_rate": 0.0003303448275862069,
+      "loss": 0.4437,
       "step": 1733
     },
     {
       "epoch": 31.53,
-      "learning_rate": 0.00022644444444444445,
-      "loss": 0.3185,
+      "learning_rate": 0.0003302068965517241,
+      "loss": 0.409,
       "step": 1734
     },
     {
       "epoch": 31.55,
-      "learning_rate": 0.00022622222222222222,
-      "loss": 0.3206,
+      "learning_rate": 0.00033006896551724134,
+      "loss": 0.3878,
       "step": 1735
     },
     {
       "epoch": 31.56,
-      "learning_rate": 0.00022600000000000002,
-      "loss": 0.3253,
+      "learning_rate": 0.0003299310344827586,
+      "loss": 0.4062,
       "step": 1736
     },
     {
       "epoch": 31.58,
-      "learning_rate": 0.0002257777777777778,
-      "loss": 0.3231,
+      "learning_rate": 0.00032979310344827587,
+      "loss": 0.4078,
       "step": 1737
     },
     {
       "epoch": 31.6,
-      "learning_rate": 0.00022555555555555556,
-      "loss": 0.3498,
+      "learning_rate": 0.00032965517241379314,
+      "loss": 0.4138,
       "step": 1738
     },
     {
       "epoch": 31.62,
-      "learning_rate": 0.00022533333333333333,
-      "loss": 0.3452,
+      "learning_rate": 0.00032951724137931035,
+      "loss": 0.3658,
       "step": 1739
     },
     {
       "epoch": 31.64,
-      "learning_rate": 0.00022511111111111112,
-      "loss": 0.3427,
+      "learning_rate": 0.0003293793103448276,
+      "loss": 0.4261,
       "step": 1740
     },
     {
       "epoch": 31.65,
-      "learning_rate": 0.0002248888888888889,
-      "loss": 0.3704,
+      "learning_rate": 0.0003292413793103448,
+      "loss": 0.4683,
       "step": 1741
     },
     {
       "epoch": 31.67,
-      "learning_rate": 0.00022466666666666666,
-      "loss": 0.3837,
+      "learning_rate": 0.0003291034482758621,
+      "loss": 0.4707,
       "step": 1742
     },
     {
       "epoch": 31.69,
-      "learning_rate": 0.00022444444444444446,
-      "loss": 0.3264,
+      "learning_rate": 0.0003289655172413793,
+      "loss": 0.3859,
       "step": 1743
     },
     {
       "epoch": 31.71,
-      "learning_rate": 0.00022422222222222223,
-      "loss": 0.3743,
+      "learning_rate": 0.00032882758620689657,
+      "loss": 0.4389,
       "step": 1744
     },
     {
       "epoch": 31.73,
-      "learning_rate": 0.000224,
-      "loss": 0.3266,
+      "learning_rate": 0.0003286896551724138,
+      "loss": 0.3877,
       "step": 1745
     },
     {
       "epoch": 31.75,
-      "learning_rate": 0.00022377777777777777,
-      "loss": 0.3585,
+      "learning_rate": 0.00032855172413793105,
+      "loss": 0.4386,
       "step": 1746
     },
     {
       "epoch": 31.76,
-      "learning_rate": 0.00022355555555555556,
-      "loss": 0.3705,
+      "learning_rate": 0.00032841379310344826,
+      "loss": 0.4503,
       "step": 1747
     },
     {
       "epoch": 31.78,
-      "learning_rate": 0.00022333333333333333,
-      "loss": 0.265,
+      "learning_rate": 0.0003282758620689655,
+      "loss": 0.3555,
       "step": 1748
     },
     {
       "epoch": 31.8,
-      "learning_rate": 0.00022311111111111113,
-      "loss": 0.3308,
+      "learning_rate": 0.00032813793103448274,
+      "loss": 0.3968,
       "step": 1749
     },
     {
       "epoch": 31.82,
-      "learning_rate": 0.0002228888888888889,
-      "loss": 0.4123,
+      "learning_rate": 0.000328,
+      "loss": 0.4628,
       "step": 1750
     },
     {
       "epoch": 31.84,
-      "learning_rate": 0.0002226666666666667,
-      "loss": 0.3706,
+      "learning_rate": 0.0003278620689655172,
+      "loss": 0.4369,
       "step": 1751
     },
     {
       "epoch": 31.85,
-      "learning_rate": 0.00022244444444444444,
-      "loss": 0.3217,
+      "learning_rate": 0.0003277241379310345,
+      "loss": 0.4081,
       "step": 1752
     },
     {
       "epoch": 31.87,
-      "learning_rate": 0.0002222222222222222,
-      "loss": 0.3314,
+      "learning_rate": 0.0003275862068965517,
+      "loss": 0.4116,
       "step": 1753
     },
     {
       "epoch": 31.89,
-      "learning_rate": 0.000222,
-      "loss": 0.3302,
+      "learning_rate": 0.000327448275862069,
+      "loss": 0.4167,
       "step": 1754
     },
     {
       "epoch": 31.91,
-      "learning_rate": 0.00022177777777777777,
-      "loss": 0.3917,
+      "learning_rate": 0.0003273103448275862,
+      "loss": 0.4576,
       "step": 1755
     },
     {
       "epoch": 31.93,
-      "learning_rate": 0.00022155555555555557,
-      "loss": 0.3321,
+      "learning_rate": 0.0003271724137931035,
+      "loss": 0.4075,
       "step": 1756
     },
     {
       "epoch": 31.95,
-      "learning_rate": 0.00022133333333333334,
-      "loss": 0.3019,
+      "learning_rate": 0.0003270344827586207,
+      "loss": 0.369,
       "step": 1757
     },
     {
       "epoch": 31.96,
-      "learning_rate": 0.00022111111111111113,
-      "loss": 0.3991,
+      "learning_rate": 0.00032689655172413797,
+      "loss": 0.4357,
       "step": 1758
     },
     {
       "epoch": 31.98,
-      "learning_rate": 0.00022088888888888888,
-      "loss": 0.4029,
+      "learning_rate": 0.0003267586206896552,
+      "loss": 0.4728,
       "step": 1759
     },
     {
       "epoch": 32.0,
-      "learning_rate": 0.00022066666666666667,
-      "loss": 0.2918,
+      "learning_rate": 0.00032662068965517245,
+      "loss": 0.4433,
       "step": 1760
     },
     {
       "epoch": 32.02,
-      "learning_rate": 0.00022044444444444444,
-      "loss": 0.2992,
+      "learning_rate": 0.00032648275862068966,
+      "loss": 0.3647,
       "step": 1761
     },
     {
       "epoch": 32.04,
-      "learning_rate": 0.00022022222222222224,
-      "loss": 0.3248,
+      "learning_rate": 0.0003263448275862069,
+      "loss": 0.3804,
       "step": 1762
     },
     {
       "epoch": 32.05,
-      "learning_rate": 0.00022,
-      "loss": 0.3635,
+      "learning_rate": 0.00032620689655172414,
+      "loss": 0.457,
       "step": 1763
     },
     {
       "epoch": 32.07,
-      "learning_rate": 0.00021977777777777778,
-      "loss": 0.3431,
+      "learning_rate": 0.0003260689655172414,
+      "loss": 0.4122,
       "step": 1764
     },
     {
       "epoch": 32.09,
-      "learning_rate": 0.00021955555555555557,
-      "loss": 0.3192,
+      "learning_rate": 0.0003259310344827586,
+      "loss": 0.3804,
       "step": 1765
     },
     {
       "epoch": 32.11,
-      "learning_rate": 0.00021933333333333332,
-      "loss": 0.3576,
+      "learning_rate": 0.0003257931034482759,
+      "loss": 0.368,
       "step": 1766
     },
     {
       "epoch": 32.13,
-      "learning_rate": 0.0002191111111111111,
-      "loss": 0.3612,
+      "learning_rate": 0.0003256551724137931,
+      "loss": 0.4296,
       "step": 1767
     },
     {
       "epoch": 32.15,
-      "learning_rate": 0.00021888888888888888,
-      "loss": 0.3051,
+      "learning_rate": 0.0003255172413793103,
+      "loss": 0.4011,
       "step": 1768
     },
     {
       "epoch": 32.16,
-      "learning_rate": 0.00021866666666666668,
-      "loss": 0.3078,
+      "learning_rate": 0.00032537931034482757,
+      "loss": 0.3962,
       "step": 1769
     },
     {
       "epoch": 32.18,
-      "learning_rate": 0.00021844444444444445,
-      "loss": 0.3385,
+      "learning_rate": 0.0003252413793103448,
+      "loss": 0.3955,
       "step": 1770
     },
     {
       "epoch": 32.2,
-      "learning_rate": 0.00021822222222222224,
-      "loss": 0.2794,
+      "learning_rate": 0.0003251034482758621,
+      "loss": 0.3418,
       "step": 1771
     },
     {
       "epoch": 32.22,
-      "learning_rate": 0.000218,
-      "loss": 0.4185,
+      "learning_rate": 0.0003249655172413793,
+      "loss": 0.4343,
       "step": 1772
     },
     {
       "epoch": 32.24,
-      "learning_rate": 0.00021777777777777776,
-      "loss": 0.3067,
+      "learning_rate": 0.0003248275862068966,
+      "loss": 0.3665,
       "step": 1773
     },
     {
       "epoch": 32.25,
-      "learning_rate": 0.00021755555555555555,
-      "loss": 0.3158,
+      "learning_rate": 0.0003246896551724138,
+      "loss": 0.4208,
       "step": 1774
     },
     {
       "epoch": 32.27,
-      "learning_rate": 0.00021733333333333332,
-      "loss": 0.3178,
+      "learning_rate": 0.00032455172413793106,
+      "loss": 0.3784,
       "step": 1775
     },
     {
       "epoch": 32.29,
-      "learning_rate": 0.00021711111111111112,
-      "loss": 0.2993,
+      "learning_rate": 0.00032441379310344827,
+      "loss": 0.3646,
       "step": 1776
     },
     {
       "epoch": 32.31,
-      "learning_rate": 0.0002168888888888889,
-      "loss": 0.3292,
+      "learning_rate": 0.00032427586206896554,
+      "loss": 0.3618,
       "step": 1777
     },
     {
       "epoch": 32.33,
-      "learning_rate": 0.00021666666666666668,
-      "loss": 0.3671,
+      "learning_rate": 0.00032413793103448275,
+      "loss": 0.4538,
       "step": 1778
     },
     {
       "epoch": 32.35,
-      "learning_rate": 0.00021644444444444445,
-      "loss": 0.3259,
+      "learning_rate": 0.000324,
+      "loss": 0.4311,
       "step": 1779
     },
     {
       "epoch": 32.36,
-      "learning_rate": 0.00021622222222222225,
-      "loss": 0.3164,
+      "learning_rate": 0.0003238620689655172,
+      "loss": 0.3701,
       "step": 1780
     },
     {
       "epoch": 32.38,
-      "learning_rate": 0.000216,
-      "loss": 0.3595,
+      "learning_rate": 0.0003237241379310345,
+      "loss": 0.4368,
       "step": 1781
     },
     {
       "epoch": 32.4,
-      "learning_rate": 0.0002157777777777778,
-      "loss": 0.3768,
+      "learning_rate": 0.0003235862068965517,
+      "loss": 0.4501,
       "step": 1782
     },
     {
       "epoch": 32.42,
-      "learning_rate": 0.00021555555555555556,
-      "loss": 0.2283,
+      "learning_rate": 0.00032344827586206897,
+      "loss": 0.3314,
       "step": 1783
     },
     {
       "epoch": 32.44,
-      "learning_rate": 0.00021533333333333333,
-      "loss": 0.347,
+      "learning_rate": 0.0003233103448275862,
+      "loss": 0.4155,
       "step": 1784
     },
     {
       "epoch": 32.45,
-      "learning_rate": 0.00021511111111111112,
-      "loss": 0.2915,
+      "learning_rate": 0.00032317241379310345,
+      "loss": 0.3689,
       "step": 1785
     },
     {
       "epoch": 32.47,
-      "learning_rate": 0.0002148888888888889,
-      "loss": 0.3827,
+      "learning_rate": 0.00032303448275862066,
+      "loss": 0.4453,
       "step": 1786
     },
     {
       "epoch": 32.49,
-      "learning_rate": 0.0002146666666666667,
-      "loss": 0.3183,
+      "learning_rate": 0.000322896551724138,
+      "loss": 0.4037,
       "step": 1787
     },
     {
       "epoch": 32.51,
-      "learning_rate": 0.00021444444444444443,
-      "loss": 0.3257,
+      "learning_rate": 0.0003227586206896552,
+      "loss": 0.4296,
       "step": 1788
     },
     {
       "epoch": 32.53,
-      "learning_rate": 0.00021422222222222223,
-      "loss": 0.297,
+      "learning_rate": 0.00032262068965517246,
+      "loss": 0.3399,
       "step": 1789
     },
     {
       "epoch": 32.55,
-      "learning_rate": 0.000214,
-      "loss": 0.333,
+      "learning_rate": 0.00032248275862068967,
+      "loss": 0.4098,
       "step": 1790
     },
     {
       "epoch": 32.56,
-      "learning_rate": 0.0002137777777777778,
-      "loss": 0.3133,
+      "learning_rate": 0.00032234482758620694,
+      "loss": 0.4286,
       "step": 1791
     },
     {
       "epoch": 32.58,
-      "learning_rate": 0.00021355555555555556,
-      "loss": 0.3306,
+      "learning_rate": 0.00032220689655172415,
+      "loss": 0.4141,
       "step": 1792
     },
     {
       "epoch": 32.6,
-      "learning_rate": 0.00021333333333333336,
-      "loss": 0.3031,
+      "learning_rate": 0.0003220689655172414,
+      "loss": 0.3643,
       "step": 1793
     },
     {
       "epoch": 32.62,
-      "learning_rate": 0.00021311111111111113,
-      "loss": 0.3198,
+      "learning_rate": 0.0003219310344827586,
+      "loss": 0.384,
       "step": 1794
     },
     {
       "epoch": 32.64,
-      "learning_rate": 0.00021288888888888887,
-      "loss": 0.389,
+      "learning_rate": 0.0003217931034482759,
+      "loss": 0.4834,
       "step": 1795
     },
     {
       "epoch": 32.65,
-      "learning_rate": 0.00021266666666666667,
-      "loss": 0.3669,
+      "learning_rate": 0.0003216551724137931,
+      "loss": 0.397,
       "step": 1796
     },
     {
       "epoch": 32.67,
-      "learning_rate": 0.00021244444444444444,
-      "loss": 0.308,
+      "learning_rate": 0.00032151724137931037,
+      "loss": 0.3608,
       "step": 1797
     },
     {
       "epoch": 32.69,
-      "learning_rate": 0.00021222222222222223,
-      "loss": 0.3538,
+      "learning_rate": 0.0003213793103448276,
+      "loss": 0.3686,
       "step": 1798
     },
     {
       "epoch": 32.71,
-      "learning_rate": 0.000212,
-      "loss": 0.3632,
+      "learning_rate": 0.00032124137931034485,
+      "loss": 0.3938,
       "step": 1799
     },
     {
       "epoch": 32.73,
-      "learning_rate": 0.0002117777777777778,
-      "loss": 0.3319,
+      "learning_rate": 0.00032110344827586206,
+      "loss": 0.3903,
       "step": 1800
     },
     {
       "epoch": 32.75,
-      "learning_rate": 0.00021155555555555557,
-      "loss": 0.2799,
+      "learning_rate": 0.00032096551724137927,
+      "loss": 0.3275,
       "step": 1801
     },
     {
       "epoch": 32.76,
-      "learning_rate": 0.00021133333333333334,
-      "loss": 0.3198,
+      "learning_rate": 0.00032082758620689654,
+      "loss": 0.3728,
       "step": 1802
     },
     {
       "epoch": 32.78,
-      "learning_rate": 0.0002111111111111111,
-      "loss": 0.2993,
+      "learning_rate": 0.00032068965517241375,
+      "loss": 0.3702,
       "step": 1803
     },
     {
       "epoch": 32.8,
-      "learning_rate": 0.00021088888888888888,
-      "loss": 0.3546,
+      "learning_rate": 0.00032055172413793107,
+      "loss": 0.3986,
       "step": 1804
     },
     {
       "epoch": 32.82,
-      "learning_rate": 0.00021066666666666667,
-      "loss": 0.3119,
+      "learning_rate": 0.0003204137931034483,
+      "loss": 0.3947,
       "step": 1805
     },
     {
       "epoch": 32.84,
-      "learning_rate": 0.00021044444444444444,
-      "loss": 0.3305,
+      "learning_rate": 0.00032027586206896555,
+      "loss": 0.3901,
       "step": 1806
     },
     {
       "epoch": 32.85,
-      "learning_rate": 0.00021022222222222224,
-      "loss": 0.3359,
+      "learning_rate": 0.00032013793103448276,
+      "loss": 0.4055,
       "step": 1807
     },
     {
       "epoch": 32.87,
-      "learning_rate": 0.00021,
-      "loss": 0.2945,
+      "learning_rate": 0.00032,
+      "loss": 0.3766,
       "step": 1808
     },
     {
       "epoch": 32.89,
-      "learning_rate": 0.00020977777777777778,
-      "loss": 0.3615,
+      "learning_rate": 0.00031986206896551724,
+      "loss": 0.4634,
       "step": 1809
     },
     {
       "epoch": 32.91,
-      "learning_rate": 0.00020955555555555555,
-      "loss": 0.31,
+      "learning_rate": 0.0003197241379310345,
+      "loss": 0.391,
       "step": 1810
     },
     {
       "epoch": 32.93,
-      "learning_rate": 0.00020933333333333334,
-      "loss": 0.3526,
+      "learning_rate": 0.0003195862068965517,
+      "loss": 0.404,
       "step": 1811
     },
     {
       "epoch": 32.95,
-      "learning_rate": 0.0002091111111111111,
-      "loss": 0.3457,
+      "learning_rate": 0.000319448275862069,
+      "loss": 0.4229,
       "step": 1812
     },
     {
       "epoch": 32.96,
-      "learning_rate": 0.0002088888888888889,
-      "loss": 0.3283,
+      "learning_rate": 0.0003193103448275862,
+      "loss": 0.3738,
       "step": 1813
     },
     {
       "epoch": 32.98,
-      "learning_rate": 0.00020866666666666668,
-      "loss": 0.3283,
+      "learning_rate": 0.00031917241379310346,
+      "loss": 0.3714,
       "step": 1814
     },
     {
       "epoch": 33.0,
-      "learning_rate": 0.00020844444444444445,
-      "loss": 0.3181,
+      "learning_rate": 0.00031903448275862067,
+      "loss": 0.36,
       "step": 1815
     },
     {
       "epoch": 33.02,
-      "learning_rate": 0.00020822222222222222,
-      "loss": 0.2742,
+      "learning_rate": 0.00031889655172413794,
+      "loss": 0.3419,
       "step": 1816
     },
     {
       "epoch": 33.04,
-      "learning_rate": 0.000208,
-      "loss": 0.3585,
+      "learning_rate": 0.00031875862068965515,
+      "loss": 0.3821,
       "step": 1817
     },
     {
       "epoch": 33.05,
-      "learning_rate": 0.00020777777777777778,
-      "loss": 0.2959,
+      "learning_rate": 0.0003186206896551724,
+      "loss": 0.3233,
       "step": 1818
     },
     {
       "epoch": 33.07,
-      "learning_rate": 0.00020755555555555555,
-      "loss": 0.2806,
+      "learning_rate": 0.00031848275862068963,
+      "loss": 0.3319,
       "step": 1819
     },
     {
       "epoch": 33.09,
-      "learning_rate": 0.00020733333333333335,
-      "loss": 0.306,
+      "learning_rate": 0.0003183448275862069,
+      "loss": 0.4224,
       "step": 1820
     },
     {
       "epoch": 33.11,
-      "learning_rate": 0.00020711111111111112,
-      "loss": 0.2902,
+      "learning_rate": 0.00031820689655172416,
+      "loss": 0.3254,
       "step": 1821
     },
     {
       "epoch": 33.13,
-      "learning_rate": 0.00020688888888888892,
-      "loss": 0.3269,
+      "learning_rate": 0.0003180689655172414,
+      "loss": 0.3794,
       "step": 1822
     },
     {
       "epoch": 33.15,
-      "learning_rate": 0.00020666666666666666,
-      "loss": 0.3196,
+      "learning_rate": 0.00031793103448275864,
+      "loss": 0.3716,
       "step": 1823
     },
     {
       "epoch": 33.16,
-      "learning_rate": 0.00020644444444444445,
-      "loss": 0.2588,
+      "learning_rate": 0.0003177931034482759,
+      "loss": 0.3689,
       "step": 1824
     },
     {
       "epoch": 33.18,
-      "learning_rate": 0.00020622222222222222,
-      "loss": 0.3586,
+      "learning_rate": 0.0003176551724137931,
+      "loss": 0.439,
       "step": 1825
     },
     {
       "epoch": 33.2,
-      "learning_rate": 0.000206,
-      "loss": 0.4114,
+      "learning_rate": 0.0003175172413793104,
+      "loss": 0.4805,
       "step": 1826
     },
     {
       "epoch": 33.22,
-      "learning_rate": 0.0002057777777777778,
-      "loss": 0.3192,
+      "learning_rate": 0.0003173793103448276,
+      "loss": 0.3556,
       "step": 1827
     },
     {
       "epoch": 33.24,
-      "learning_rate": 0.00020555555555555556,
-      "loss": 0.28,
+      "learning_rate": 0.00031724137931034486,
+      "loss": 0.3483,
       "step": 1828
     },
     {
       "epoch": 33.25,
-      "learning_rate": 0.00020533333333333336,
-      "loss": 0.3207,
+      "learning_rate": 0.00031710344827586207,
+      "loss": 0.407,
       "step": 1829
     },
     {
       "epoch": 33.27,
-      "learning_rate": 0.0002051111111111111,
-      "loss": 0.304,
+      "learning_rate": 0.00031696551724137934,
+      "loss": 0.3757,
       "step": 1830
     },
     {
       "epoch": 33.29,
-      "learning_rate": 0.0002048888888888889,
-      "loss": 0.3087,
+      "learning_rate": 0.00031682758620689655,
+      "loss": 0.3618,
       "step": 1831
     },
     {
       "epoch": 33.31,
-      "learning_rate": 0.00020466666666666666,
-      "loss": 0.2988,
+      "learning_rate": 0.0003166896551724138,
+      "loss": 0.3635,
       "step": 1832
     },
     {
       "epoch": 33.33,
-      "learning_rate": 0.00020444444444444446,
-      "loss": 0.3234,
+      "learning_rate": 0.00031655172413793103,
+      "loss": 0.3864,
       "step": 1833
     },
     {
       "epoch": 33.35,
-      "learning_rate": 0.00020422222222222223,
-      "loss": 0.3277,
+      "learning_rate": 0.00031641379310344824,
+      "loss": 0.3669,
       "step": 1834
     },
     {
       "epoch": 33.36,
-      "learning_rate": 0.000204,
-      "loss": 0.2755,
+      "learning_rate": 0.0003162758620689655,
+      "loss": 0.3808,
       "step": 1835
     },
     {
       "epoch": 33.38,
-      "learning_rate": 0.0002037777777777778,
-      "loss": 0.3492,
+      "learning_rate": 0.0003161379310344827,
+      "loss": 0.4177,
       "step": 1836
     },
     {
       "epoch": 33.4,
-      "learning_rate": 0.00020355555555555554,
-      "loss": 0.344,
+      "learning_rate": 0.000316,
+      "loss": 0.4016,
       "step": 1837
     },
     {
       "epoch": 33.42,
-      "learning_rate": 0.00020333333333333333,
-      "loss": 0.2581,
+      "learning_rate": 0.00031586206896551725,
+      "loss": 0.3497,
       "step": 1838
     },
     {
       "epoch": 33.44,
-      "learning_rate": 0.0002031111111111111,
-      "loss": 0.3147,
+      "learning_rate": 0.0003157241379310345,
+      "loss": 0.3803,
       "step": 1839
     },
     {
       "epoch": 33.45,
-      "learning_rate": 0.0002028888888888889,
-      "loss": 0.3091,
+      "learning_rate": 0.00031558620689655173,
+      "loss": 0.3863,
       "step": 1840
     },
     {
       "epoch": 33.47,
-      "learning_rate": 0.00020266666666666667,
-      "loss": 0.2619,
+      "learning_rate": 0.000315448275862069,
+      "loss": 0.3575,
       "step": 1841
     },
     {
       "epoch": 33.49,
-      "learning_rate": 0.00020244444444444447,
-      "loss": 0.308,
+      "learning_rate": 0.0003153103448275862,
+      "loss": 0.3361,
       "step": 1842
     },
     {
       "epoch": 33.51,
-      "learning_rate": 0.00020222222222222223,
-      "loss": 0.2803,
+      "learning_rate": 0.00031517241379310347,
+      "loss": 0.3731,
       "step": 1843
     },
     {
       "epoch": 33.53,
-      "learning_rate": 0.000202,
-      "loss": 0.2961,
+      "learning_rate": 0.0003150344827586207,
+      "loss": 0.3751,
       "step": 1844
     },
     {
       "epoch": 33.55,
-      "learning_rate": 0.00020177777777777777,
-      "loss": 0.3655,
+      "learning_rate": 0.00031489655172413795,
+      "loss": 0.4403,
       "step": 1845
     },
     {
       "epoch": 33.56,
-      "learning_rate": 0.00020155555555555554,
-      "loss": 0.3313,
+      "learning_rate": 0.00031475862068965516,
+      "loss": 0.3765,
       "step": 1846
     },
     {
       "epoch": 33.58,
-      "learning_rate": 0.00020133333333333334,
-      "loss": 0.2793,
+      "learning_rate": 0.0003146206896551724,
+      "loss": 0.3413,
       "step": 1847
     },
     {
       "epoch": 33.6,
-      "learning_rate": 0.0002011111111111111,
-      "loss": 0.3217,
+      "learning_rate": 0.00031448275862068964,
+      "loss": 0.4025,
       "step": 1848
     },
     {
       "epoch": 33.62,
-      "learning_rate": 0.0002008888888888889,
-      "loss": 0.3445,
+      "learning_rate": 0.0003143448275862069,
+      "loss": 0.4023,
       "step": 1849
     },
     {
       "epoch": 33.64,
-      "learning_rate": 0.00020066666666666667,
-      "loss": 0.3107,
+      "learning_rate": 0.0003142068965517241,
+      "loss": 0.3538,
       "step": 1850
     },
     {
       "epoch": 33.65,
-      "learning_rate": 0.00020044444444444444,
-      "loss": 0.3027,
+      "learning_rate": 0.0003140689655172414,
+      "loss": 0.3741,
       "step": 1851
     },
     {
       "epoch": 33.67,
-      "learning_rate": 0.0002002222222222222,
-      "loss": 0.3228,
+      "learning_rate": 0.0003139310344827586,
+      "loss": 0.3546,
       "step": 1852
     },
     {
       "epoch": 33.69,
-      "learning_rate": 0.0002,
-      "loss": 0.3289,
+      "learning_rate": 0.00031379310344827586,
+      "loss": 0.4277,
       "step": 1853
     },
     {
       "epoch": 33.71,
-      "learning_rate": 0.00019977777777777778,
-      "loss": 0.3325,
+      "learning_rate": 0.00031365517241379307,
+      "loss": 0.411,
       "step": 1854
     },
     {
       "epoch": 33.73,
-      "learning_rate": 0.00019955555555555558,
-      "loss": 0.2841,
+      "learning_rate": 0.0003135172413793104,
+      "loss": 0.3844,
       "step": 1855
     },
     {
       "epoch": 33.75,
-      "learning_rate": 0.00019933333333333334,
-      "loss": 0.3044,
+      "learning_rate": 0.0003133793103448276,
+      "loss": 0.4085,
       "step": 1856
     },
     {
       "epoch": 33.76,
-      "learning_rate": 0.00019911111111111111,
-      "loss": 0.307,
+      "learning_rate": 0.00031324137931034487,
+      "loss": 0.3674,
       "step": 1857
     },
     {
       "epoch": 33.78,
-      "learning_rate": 0.00019888888888888888,
-      "loss": 0.335,
+      "learning_rate": 0.0003131034482758621,
+      "loss": 0.3828,
       "step": 1858
     },
     {
       "epoch": 33.8,
-      "learning_rate": 0.00019866666666666665,
-      "loss": 0.3366,
+      "learning_rate": 0.00031296551724137935,
+      "loss": 0.4145,
       "step": 1859
     },
     {
       "epoch": 33.82,
-      "learning_rate": 0.00019844444444444445,
-      "loss": 0.3245,
+      "learning_rate": 0.00031282758620689656,
+      "loss": 0.4276,
       "step": 1860
     },
     {
       "epoch": 33.84,
-      "learning_rate": 0.00019822222222222222,
-      "loss": 0.2956,
+      "learning_rate": 0.0003126896551724138,
+      "loss": 0.3417,
       "step": 1861
     },
     {
       "epoch": 33.85,
-      "learning_rate": 0.00019800000000000002,
-      "loss": 0.3205,
+      "learning_rate": 0.00031255172413793104,
+      "loss": 0.3691,
       "step": 1862
     },
     {
       "epoch": 33.87,
-      "learning_rate": 0.00019777777777777778,
-      "loss": 0.3487,
+      "learning_rate": 0.0003124137931034483,
+      "loss": 0.4537,
       "step": 1863
     },
     {
       "epoch": 33.89,
-      "learning_rate": 0.00019755555555555558,
-      "loss": 0.2752,
+      "learning_rate": 0.0003122758620689655,
+      "loss": 0.3814,
       "step": 1864
     },
     {
       "epoch": 33.91,
-      "learning_rate": 0.00019733333333333335,
-      "loss": 0.3464,
+      "learning_rate": 0.0003121379310344828,
+      "loss": 0.4206,
       "step": 1865
     },
     {
       "epoch": 33.93,
-      "learning_rate": 0.0001971111111111111,
-      "loss": 0.304,
+      "learning_rate": 0.000312,
+      "loss": 0.4057,
       "step": 1866
     },
     {
       "epoch": 33.95,
-      "learning_rate": 0.0001968888888888889,
-      "loss": 0.2909,
+      "learning_rate": 0.0003118620689655172,
+      "loss": 0.3629,
       "step": 1867
     },
     {
       "epoch": 33.96,
-      "learning_rate": 0.00019666666666666666,
-      "loss": 0.344,
+      "learning_rate": 0.00031172413793103447,
+      "loss": 0.4351,
       "step": 1868
     },
     {
       "epoch": 33.98,
-      "learning_rate": 0.00019644444444444445,
-      "loss": 0.3194,
+      "learning_rate": 0.0003115862068965517,
+      "loss": 0.3778,
       "step": 1869
     },
     {
       "epoch": 34.0,
-      "learning_rate": 0.00019622222222222222,
-      "loss": 0.2586,
+      "learning_rate": 0.00031144827586206895,
+      "loss": 0.3671,
       "step": 1870
     },
     {
       "epoch": 34.02,
-      "learning_rate": 0.00019600000000000002,
-      "loss": 0.3335,
+      "learning_rate": 0.0003113103448275862,
+      "loss": 0.3863,
       "step": 1871
     },
     {
       "epoch": 34.04,
-      "learning_rate": 0.0001957777777777778,
-      "loss": 0.2778,
+      "learning_rate": 0.0003111724137931035,
+      "loss": 0.3692,
       "step": 1872
     },
     {
       "epoch": 34.05,
-      "learning_rate": 0.00019555555555555556,
-      "loss": 0.3276,
+      "learning_rate": 0.0003110344827586207,
+      "loss": 0.364,
       "step": 1873
     },
     {
       "epoch": 34.07,
-      "learning_rate": 0.00019533333333333333,
-      "loss": 0.2947,
+      "learning_rate": 0.00031089655172413796,
+      "loss": 0.3798,
       "step": 1874
     },
     {
       "epoch": 34.09,
-      "learning_rate": 0.00019511111111111113,
-      "loss": 0.3275,
+      "learning_rate": 0.00031075862068965517,
+      "loss": 0.4335,
       "step": 1875
     },
     {
       "epoch": 34.11,
-      "learning_rate": 0.0001948888888888889,
-      "loss": 0.2852,
+      "learning_rate": 0.00031062068965517244,
+      "loss": 0.386,
       "step": 1876
     },
     {
       "epoch": 34.13,
-      "learning_rate": 0.00019466666666666666,
-      "loss": 0.3441,
+      "learning_rate": 0.00031048275862068965,
+      "loss": 0.4077,
       "step": 1877
     },
     {
       "epoch": 34.15,
-      "learning_rate": 0.00019444444444444446,
-      "loss": 0.2914,
+      "learning_rate": 0.0003103448275862069,
+      "loss": 0.37,
       "step": 1878
     },
     {
       "epoch": 34.16,
-      "learning_rate": 0.00019422222222222223,
-      "loss": 0.27,
+      "learning_rate": 0.00031020689655172413,
+      "loss": 0.3122,
       "step": 1879
     },
     {
       "epoch": 34.18,
-      "learning_rate": 0.000194,
-      "loss": 0.3307,
+      "learning_rate": 0.0003100689655172414,
+      "loss": 0.4072,
       "step": 1880
     },
     {
       "epoch": 34.2,
-      "learning_rate": 0.00019377777777777777,
-      "loss": 0.2712,
+      "learning_rate": 0.0003099310344827586,
+      "loss": 0.3848,
       "step": 1881
     },
     {
       "epoch": 34.22,
-      "learning_rate": 0.00019355555555555557,
-      "loss": 0.3265,
+      "learning_rate": 0.00030979310344827587,
+      "loss": 0.4148,
       "step": 1882
     },
     {
       "epoch": 34.24,
-      "learning_rate": 0.00019333333333333333,
-      "loss": 0.3818,
+      "learning_rate": 0.0003096551724137931,
+      "loss": 0.4403,
       "step": 1883
     },
     {
       "epoch": 34.25,
-      "learning_rate": 0.00019311111111111113,
-      "loss": 0.318,
+      "learning_rate": 0.00030951724137931035,
+      "loss": 0.3974,
       "step": 1884
     },
     {
       "epoch": 34.27,
-      "learning_rate": 0.0001928888888888889,
-      "loss": 0.3326,
+      "learning_rate": 0.00030937931034482756,
+      "loss": 0.367,
       "step": 1885
     },
     {
       "epoch": 34.29,
-      "learning_rate": 0.0001926666666666667,
-      "loss": 0.322,
+      "learning_rate": 0.00030924137931034483,
+      "loss": 0.4215,
       "step": 1886
     },
     {
       "epoch": 34.31,
-      "learning_rate": 0.00019244444444444444,
-      "loss": 0.3308,
+      "learning_rate": 0.00030910344827586204,
+      "loss": 0.3932,
       "step": 1887
     },
     {
       "epoch": 34.33,
-      "learning_rate": 0.0001922222222222222,
-      "loss": 0.3067,
+      "learning_rate": 0.00030896551724137936,
+      "loss": 0.3947,
       "step": 1888
     },
     {
       "epoch": 34.35,
-      "learning_rate": 0.000192,
-      "loss": 0.3126,
+      "learning_rate": 0.00030882758620689657,
+      "loss": 0.3765,
       "step": 1889
     },
     {
       "epoch": 34.36,
-      "learning_rate": 0.00019177777777777777,
-      "loss": 0.3063,
+      "learning_rate": 0.00030868965517241384,
+      "loss": 0.3872,
       "step": 1890
     },
     {
       "epoch": 34.38,
-      "learning_rate": 0.00019155555555555557,
-      "loss": 0.3107,
+      "learning_rate": 0.00030855172413793105,
+      "loss": 0.3594,
       "step": 1891
     },
     {
       "epoch": 34.4,
-      "learning_rate": 0.00019133333333333334,
-      "loss": 0.3033,
+      "learning_rate": 0.0003084137931034483,
+      "loss": 0.3762,
       "step": 1892
     },
     {
       "epoch": 34.42,
-      "learning_rate": 0.00019111111111111114,
-      "loss": 0.3116,
+      "learning_rate": 0.00030827586206896553,
+      "loss": 0.3553,
       "step": 1893
     },
     {
       "epoch": 34.44,
-      "learning_rate": 0.00019088888888888888,
-      "loss": 0.3223,
+      "learning_rate": 0.0003081379310344828,
+      "loss": 0.3647,
       "step": 1894
     },
     {
       "epoch": 34.45,
-      "learning_rate": 0.00019066666666666668,
-      "loss": 0.3063,
+      "learning_rate": 0.000308,
+      "loss": 0.3909,
       "step": 1895
     },
     {
       "epoch": 34.47,
-      "learning_rate": 0.00019044444444444444,
-      "loss": 0.3513,
+      "learning_rate": 0.00030786206896551727,
+      "loss": 0.4143,
       "step": 1896
     },
     {
       "epoch": 34.49,
-      "learning_rate": 0.00019022222222222221,
-      "loss": 0.2488,
+      "learning_rate": 0.0003077241379310345,
+      "loss": 0.3204,
       "step": 1897
     },
     {
       "epoch": 34.51,
-      "learning_rate": 0.00019,
-      "loss": 0.3295,
+      "learning_rate": 0.00030758620689655175,
+      "loss": 0.3898,
       "step": 1898
     },
     {
       "epoch": 34.53,
-      "learning_rate": 0.00018977777777777778,
-      "loss": 0.2906,
+      "learning_rate": 0.00030744827586206896,
+      "loss": 0.3636,
       "step": 1899
     },
     {
       "epoch": 34.55,
-      "learning_rate": 0.00018955555555555558,
-      "loss": 0.3478,
+      "learning_rate": 0.0003073103448275862,
+      "loss": 0.4242,
       "step": 1900
     },
     {
       "epoch": 34.56,
-      "learning_rate": 0.00018933333333333332,
-      "loss": 0.2915,
+      "learning_rate": 0.00030717241379310344,
+      "loss": 0.3623,
       "step": 1901
     },
     {
       "epoch": 34.58,
-      "learning_rate": 0.00018911111111111112,
-      "loss": 0.2974,
+      "learning_rate": 0.00030703448275862065,
+      "loss": 0.3913,
       "step": 1902
     },
     {
       "epoch": 34.6,
-      "learning_rate": 0.00018888888888888888,
-      "loss": 0.3183,
+      "learning_rate": 0.0003068965517241379,
+      "loss": 0.4359,
       "step": 1903
     },
     {
       "epoch": 34.62,
-      "learning_rate": 0.00018866666666666668,
-      "loss": 0.2773,
+      "learning_rate": 0.00030675862068965513,
+      "loss": 0.3477,
       "step": 1904
     },
     {
       "epoch": 34.64,
-      "learning_rate": 0.00018844444444444445,
-      "loss": 0.3298,
+      "learning_rate": 0.00030662068965517245,
+      "loss": 0.3625,
       "step": 1905
     },
     {
       "epoch": 34.65,
-      "learning_rate": 0.00018822222222222225,
-      "loss": 0.2823,
+      "learning_rate": 0.00030648275862068966,
+      "loss": 0.3808,
       "step": 1906
     },
     {
       "epoch": 34.67,
-      "learning_rate": 0.00018800000000000002,
-      "loss": 0.3126,
+      "learning_rate": 0.00030634482758620693,
+      "loss": 0.4427,
       "step": 1907
     },
     {
       "epoch": 34.69,
-      "learning_rate": 0.00018777777777777776,
-      "loss": 0.3083,
+      "learning_rate": 0.00030620689655172414,
+      "loss": 0.423,
       "step": 1908
     },
     {
       "epoch": 34.71,
-      "learning_rate": 0.00018755555555555555,
-      "loss": 0.2832,
+      "learning_rate": 0.0003060689655172414,
+      "loss": 0.3465,
       "step": 1909
     },
     {
       "epoch": 34.73,
-      "learning_rate": 0.00018733333333333332,
-      "loss": 0.2741,
+      "learning_rate": 0.0003059310344827586,
+      "loss": 0.386,
       "step": 1910
     },
     {
       "epoch": 34.75,
-      "learning_rate": 0.00018711111111111112,
-      "loss": 0.2341,
+      "learning_rate": 0.0003057931034482759,
+      "loss": 0.3152,
       "step": 1911
     },
     {
       "epoch": 34.76,
-      "learning_rate": 0.0001868888888888889,
-      "loss": 0.3156,
+      "learning_rate": 0.0003056551724137931,
+      "loss": 0.3745,
       "step": 1912
     },
     {
       "epoch": 34.78,
-      "learning_rate": 0.0001866666666666667,
-      "loss": 0.2754,
+      "learning_rate": 0.00030551724137931036,
+      "loss": 0.3429,
       "step": 1913
     },
     {
       "epoch": 34.8,
-      "learning_rate": 0.00018644444444444446,
-      "loss": 0.2587,
+      "learning_rate": 0.0003053793103448276,
+      "loss": 0.3343,
       "step": 1914
     },
     {
       "epoch": 34.82,
-      "learning_rate": 0.00018622222222222223,
-      "loss": 0.3316,
+      "learning_rate": 0.00030524137931034484,
+      "loss": 0.3751,
       "step": 1915
     },
     {
       "epoch": 34.84,
-      "learning_rate": 0.000186,
-      "loss": 0.3452,
+      "learning_rate": 0.00030510344827586205,
+      "loss": 0.4222,
       "step": 1916
     },
     {
       "epoch": 34.85,
-      "learning_rate": 0.00018577777777777776,
-      "loss": 0.2642,
+      "learning_rate": 0.0003049655172413793,
+      "loss": 0.3239,
       "step": 1917
     },
     {
       "epoch": 34.87,
-      "learning_rate": 0.00018555555555555556,
-      "loss": 0.2625,
+      "learning_rate": 0.00030482758620689653,
+      "loss": 0.3734,
       "step": 1918
     },
     {
       "epoch": 34.89,
-      "learning_rate": 0.00018533333333333333,
-      "loss": 0.385,
+      "learning_rate": 0.0003046896551724138,
+      "loss": 0.4056,
       "step": 1919
     },
     {
       "epoch": 34.91,
-      "learning_rate": 0.00018511111111111113,
-      "loss": 0.3632,
+      "learning_rate": 0.000304551724137931,
+      "loss": 0.4087,
       "step": 1920
     },
     {
       "epoch": 34.93,
-      "learning_rate": 0.0001848888888888889,
-      "loss": 0.3038,
+      "learning_rate": 0.0003044137931034483,
+      "loss": 0.3682,
       "step": 1921
     },
     {
       "epoch": 34.95,
-      "learning_rate": 0.00018466666666666666,
-      "loss": 0.3105,
+      "learning_rate": 0.00030427586206896554,
+      "loss": 0.3712,
       "step": 1922
     },
     {
       "epoch": 34.96,
-      "learning_rate": 0.00018444444444444443,
-      "loss": 0.4211,
+      "learning_rate": 0.0003041379310344828,
+      "loss": 0.4756,
       "step": 1923
     },
     {
       "epoch": 34.98,
-      "learning_rate": 0.00018422222222222223,
-      "loss": 0.3461,
+      "learning_rate": 0.000304,
+      "loss": 0.4229,
       "step": 1924
     },
     {
       "epoch": 35.0,
-      "learning_rate": 0.000184,
-      "loss": 0.3228,
+      "learning_rate": 0.0003038620689655173,
+      "loss": 0.3838,
       "step": 1925
     },
     {
       "epoch": 35.02,
-      "learning_rate": 0.0001837777777777778,
-      "loss": 0.379,
+      "learning_rate": 0.0003037241379310345,
+      "loss": 0.4298,
       "step": 1926
     },
     {
       "epoch": 35.04,
-      "learning_rate": 0.00018355555555555557,
-      "loss": 0.3943,
+      "learning_rate": 0.00030358620689655176,
+      "loss": 0.4614,
       "step": 1927
     },
     {
       "epoch": 35.05,
-      "learning_rate": 0.00018333333333333334,
-      "loss": 0.3104,
+      "learning_rate": 0.00030344827586206897,
+      "loss": 0.3754,
       "step": 1928
     },
     {
       "epoch": 35.07,
-      "learning_rate": 0.0001831111111111111,
-      "loss": 0.2848,
+      "learning_rate": 0.00030331034482758624,
+      "loss": 0.3721,
       "step": 1929
     },
     {
       "epoch": 35.09,
-      "learning_rate": 0.00018288888888888887,
-      "loss": 0.2986,
+      "learning_rate": 0.00030317241379310345,
+      "loss": 0.3811,
       "step": 1930
     },
     {
       "epoch": 35.11,
-      "learning_rate": 0.00018266666666666667,
-      "loss": 0.329,
+      "learning_rate": 0.0003030344827586207,
+      "loss": 0.3714,
       "step": 1931
     },
     {
       "epoch": 35.13,
-      "learning_rate": 0.00018244444444444444,
-      "loss": 0.3251,
+      "learning_rate": 0.00030289655172413793,
+      "loss": 0.3642,
       "step": 1932
     },
     {
       "epoch": 35.15,
-      "learning_rate": 0.00018222222222222224,
-      "loss": 0.2973,
+      "learning_rate": 0.0003027586206896552,
+      "loss": 0.4174,
       "step": 1933
     },
     {
       "epoch": 35.16,
-      "learning_rate": 0.000182,
-      "loss": 0.2504,
+      "learning_rate": 0.0003026206896551724,
+      "loss": 0.3061,
       "step": 1934
     },
     {
       "epoch": 35.18,
-      "learning_rate": 0.0001817777777777778,
-      "loss": 0.3243,
+      "learning_rate": 0.0003024827586206896,
+      "loss": 0.391,
       "step": 1935
     },
     {
       "epoch": 35.2,
-      "learning_rate": 0.00018155555555555554,
-      "loss": 0.3168,
+      "learning_rate": 0.0003023448275862069,
+      "loss": 0.3741,
       "step": 1936
     },
     {
       "epoch": 35.22,
-      "learning_rate": 0.00018133333333333334,
-      "loss": 0.2987,
+      "learning_rate": 0.0003022068965517241,
+      "loss": 0.3439,
       "step": 1937
     },
     {
       "epoch": 35.24,
-      "learning_rate": 0.0001811111111111111,
-      "loss": 0.2918,
+      "learning_rate": 0.0003020689655172414,
+      "loss": 0.3427,
       "step": 1938
     },
     {
       "epoch": 35.25,
-      "learning_rate": 0.00018088888888888888,
-      "loss": 0.3004,
+      "learning_rate": 0.00030193103448275863,
+      "loss": 0.3814,
       "step": 1939
     },
     {
       "epoch": 35.27,
-      "learning_rate": 0.00018066666666666668,
-      "loss": 0.2826,
+      "learning_rate": 0.0003017931034482759,
+      "loss": 0.3569,
       "step": 1940
     },
     {
       "epoch": 35.29,
-      "learning_rate": 0.00018044444444444445,
-      "loss": 0.2818,
+      "learning_rate": 0.0003016551724137931,
+      "loss": 0.3729,
       "step": 1941
     },
     {
       "epoch": 35.31,
-      "learning_rate": 0.00018022222222222224,
-      "loss": 0.343,
+      "learning_rate": 0.00030151724137931037,
+      "loss": 0.4129,
       "step": 1942
     },
     {
       "epoch": 35.33,
-      "learning_rate": 0.00017999999999999998,
-      "loss": 0.3253,
+      "learning_rate": 0.0003013793103448276,
+      "loss": 0.3825,
       "step": 1943
     },
     {
       "epoch": 35.35,
-      "learning_rate": 0.00017977777777777778,
-      "loss": 0.3571,
+      "learning_rate": 0.00030124137931034485,
+      "loss": 0.3672,
       "step": 1944
     },
     {
       "epoch": 35.36,
-      "learning_rate": 0.00017955555555555555,
-      "loss": 0.3573,
+      "learning_rate": 0.00030110344827586206,
+      "loss": 0.4551,
       "step": 1945
     },
     {
       "epoch": 35.38,
-      "learning_rate": 0.00017933333333333335,
-      "loss": 0.3144,
+      "learning_rate": 0.00030096551724137933,
+      "loss": 0.3975,
       "step": 1946
     },
     {
       "epoch": 35.4,
-      "learning_rate": 0.00017911111111111112,
-      "loss": 0.3171,
+      "learning_rate": 0.00030082758620689654,
+      "loss": 0.3729,
       "step": 1947
     },
     {
       "epoch": 35.42,
-      "learning_rate": 0.00017888888888888889,
-      "loss": 0.281,
+      "learning_rate": 0.0003006896551724138,
+      "loss": 0.3125,
       "step": 1948
     },
     {
       "epoch": 35.44,
-      "learning_rate": 0.00017866666666666668,
-      "loss": 0.3255,
+      "learning_rate": 0.000300551724137931,
+      "loss": 0.4072,
       "step": 1949
     },
     {
       "epoch": 35.45,
-      "learning_rate": 0.00017844444444444442,
-      "loss": 0.2558,
+      "learning_rate": 0.0003004137931034483,
+      "loss": 0.3229,
       "step": 1950
     },
     {
       "epoch": 35.47,
-      "learning_rate": 0.00017822222222222222,
-      "loss": 0.3453,
+      "learning_rate": 0.0003002758620689655,
+      "loss": 0.3886,
       "step": 1951
     },
     {
       "epoch": 35.49,
-      "learning_rate": 0.000178,
-      "loss": 0.3034,
+      "learning_rate": 0.00030013793103448276,
+      "loss": 0.3409,
       "step": 1952
     },
     {
       "epoch": 35.51,
-      "learning_rate": 0.00017777777777777779,
-      "loss": 0.2953,
+      "learning_rate": 0.0003,
+      "loss": 0.3771,
       "step": 1953
     },
     {
       "epoch": 35.53,
-      "learning_rate": 0.00017755555555555556,
-      "loss": 0.3471,
+      "learning_rate": 0.00029986206896551724,
+      "loss": 0.3987,
       "step": 1954
     },
     {
       "epoch": 35.55,
-      "learning_rate": 0.00017733333333333335,
-      "loss": 0.3659,
+      "learning_rate": 0.0002997241379310345,
+      "loss": 0.4318,
       "step": 1955
     },
     {
       "epoch": 35.56,
-      "learning_rate": 0.00017711111111111112,
-      "loss": 0.3213,
+      "learning_rate": 0.00029958620689655177,
+      "loss": 0.3653,
       "step": 1956
     },
     {
       "epoch": 35.58,
-      "learning_rate": 0.00017688888888888892,
-      "loss": 0.3296,
+      "learning_rate": 0.000299448275862069,
+      "loss": 0.3866,
       "step": 1957
     },
     {
       "epoch": 35.6,
-      "learning_rate": 0.00017666666666666666,
-      "loss": 0.2986,
+      "learning_rate": 0.00029931034482758625,
+      "loss": 0.4206,
       "step": 1958
     },
     {
       "epoch": 35.62,
-      "learning_rate": 0.00017644444444444443,
-      "loss": 0.2926,
+      "learning_rate": 0.00029917241379310346,
+      "loss": 0.3487,
       "step": 1959
     },
     {
       "epoch": 35.64,
-      "learning_rate": 0.00017622222222222223,
-      "loss": 0.3078,
+      "learning_rate": 0.00029903448275862073,
+      "loss": 0.3913,
       "step": 1960
     },
     {
       "epoch": 35.65,
-      "learning_rate": 0.000176,
-      "loss": 0.2699,
+      "learning_rate": 0.00029889655172413794,
+      "loss": 0.2839,
       "step": 1961
     },
     {
       "epoch": 35.67,
-      "learning_rate": 0.0001757777777777778,
-      "loss": 0.2868,
+      "learning_rate": 0.0002987586206896552,
+      "loss": 0.3824,
       "step": 1962
     },
     {
       "epoch": 35.69,
-      "learning_rate": 0.00017555555555555556,
-      "loss": 0.2777,
+      "learning_rate": 0.0002986206896551724,
+      "loss": 0.3815,
       "step": 1963
     },
     {
       "epoch": 35.71,
-      "learning_rate": 0.00017533333333333336,
-      "loss": 0.2946,
+      "learning_rate": 0.0002984827586206897,
+      "loss": 0.3367,
       "step": 1964
     },
     {
       "epoch": 35.73,
-      "learning_rate": 0.0001751111111111111,
-      "loss": 0.2559,
+      "learning_rate": 0.0002983448275862069,
+      "loss": 0.3216,
       "step": 1965
     },
     {
       "epoch": 35.75,
-      "learning_rate": 0.0001748888888888889,
-      "loss": 0.3419,
+      "learning_rate": 0.00029820689655172416,
+      "loss": 0.411,
       "step": 1966
     },
     {
       "epoch": 35.76,
-      "learning_rate": 0.00017466666666666667,
-      "loss": 0.2553,
+      "learning_rate": 0.0002980689655172414,
+      "loss": 0.3177,
       "step": 1967
     },
     {
       "epoch": 35.78,
-      "learning_rate": 0.00017444444444444446,
-      "loss": 0.3101,
+      "learning_rate": 0.0002979310344827586,
+      "loss": 0.4221,
       "step": 1968
     },
     {
       "epoch": 35.8,
-      "learning_rate": 0.00017422222222222223,
-      "loss": 0.3095,
+      "learning_rate": 0.00029779310344827585,
+      "loss": 0.4036,
       "step": 1969
     },
     {
       "epoch": 35.82,
-      "learning_rate": 0.000174,
-      "loss": 0.3692,
+      "learning_rate": 0.00029765517241379306,
+      "loss": 0.456,
       "step": 1970
     },
     {
       "epoch": 35.84,
-      "learning_rate": 0.0001737777777777778,
-      "loss": 0.308,
+      "learning_rate": 0.00029751724137931033,
+      "loss": 0.3433,
       "step": 1971
     },
     {
       "epoch": 35.85,
-      "learning_rate": 0.00017355555555555554,
-      "loss": 0.3189,
+      "learning_rate": 0.0002973793103448276,
+      "loss": 0.3862,
       "step": 1972
     },
     {
       "epoch": 35.87,
-      "learning_rate": 0.00017333333333333334,
-      "loss": 0.3454,
+      "learning_rate": 0.00029724137931034486,
+      "loss": 0.3956,
       "step": 1973
     },
     {
       "epoch": 35.89,
-      "learning_rate": 0.0001731111111111111,
-      "loss": 0.317,
+      "learning_rate": 0.0002971034482758621,
+      "loss": 0.3095,
       "step": 1974
     },
     {
       "epoch": 35.91,
-      "learning_rate": 0.0001728888888888889,
-      "loss": 0.294,
+      "learning_rate": 0.00029696551724137934,
+      "loss": 0.374,
       "step": 1975
     },
     {
       "epoch": 35.93,
-      "learning_rate": 0.00017266666666666667,
-      "loss": 0.2691,
+      "learning_rate": 0.00029682758620689655,
+      "loss": 0.3288,
       "step": 1976
     },
     {
       "epoch": 35.95,
-      "learning_rate": 0.00017244444444444447,
-      "loss": 0.2546,
+      "learning_rate": 0.0002966896551724138,
+      "loss": 0.3139,
       "step": 1977
     },
     {
       "epoch": 35.96,
-      "learning_rate": 0.00017222222222222224,
-      "loss": 0.3309,
+      "learning_rate": 0.00029655172413793103,
+      "loss": 0.4346,
       "step": 1978
     },
     {
       "epoch": 35.98,
-      "learning_rate": 0.00017199999999999998,
-      "loss": 0.3322,
+      "learning_rate": 0.0002964137931034483,
+      "loss": 0.3867,
       "step": 1979
     },
     {
       "epoch": 36.0,
-      "learning_rate": 0.00017177777777777778,
-      "loss": 0.3791,
+      "learning_rate": 0.0002962758620689655,
+      "loss": 0.4654,
       "step": 1980
     },
     {
       "epoch": 36.02,
-      "learning_rate": 0.00017155555555555555,
-      "loss": 0.3601,
+      "learning_rate": 0.0002961379310344828,
+      "loss": 0.383,
       "step": 1981
     },
     {
       "epoch": 36.04,
-      "learning_rate": 0.00017133333333333334,
-      "loss": 0.2915,
+      "learning_rate": 0.000296,
+      "loss": 0.3233,
       "step": 1982
     },
     {
       "epoch": 36.05,
-      "learning_rate": 0.0001711111111111111,
-      "loss": 0.2756,
+      "learning_rate": 0.00029586206896551725,
+      "loss": 0.3008,
       "step": 1983
     },
     {
       "epoch": 36.07,
-      "learning_rate": 0.0001708888888888889,
-      "loss": 0.3229,
+      "learning_rate": 0.00029572413793103446,
+      "loss": 0.3792,
       "step": 1984
     },
     {
       "epoch": 36.09,
-      "learning_rate": 0.00017066666666666668,
-      "loss": 0.3318,
+      "learning_rate": 0.00029558620689655173,
+      "loss": 0.363,
       "step": 1985
     },
     {
       "epoch": 36.11,
-      "learning_rate": 0.00017044444444444445,
-      "loss": 0.2656,
+      "learning_rate": 0.00029544827586206894,
+      "loss": 0.3524,
       "step": 1986
     },
     {
       "epoch": 36.13,
-      "learning_rate": 0.00017022222222222222,
-      "loss": 0.2565,
+      "learning_rate": 0.0002953103448275862,
+      "loss": 0.3656,
       "step": 1987
     },
     {
       "epoch": 36.15,
-      "learning_rate": 0.00017,
-      "loss": 0.2393,
+      "learning_rate": 0.0002951724137931034,
+      "loss": 0.3382,
       "step": 1988
     },
     {
       "epoch": 36.16,
-      "learning_rate": 0.00016977777777777778,
-      "loss": 0.2692,
+      "learning_rate": 0.00029503448275862074,
+      "loss": 0.3296,
       "step": 1989
     },
     {
       "epoch": 36.18,
-      "learning_rate": 0.00016955555555555555,
-      "loss": 0.316,
+      "learning_rate": 0.00029489655172413795,
+      "loss": 0.3903,
       "step": 1990
     },
     {
       "epoch": 36.2,
-      "learning_rate": 0.00016933333333333335,
-      "loss": 0.2534,
+      "learning_rate": 0.0002947586206896552,
+      "loss": 0.3187,
       "step": 1991
     },
     {
       "epoch": 36.22,
-      "learning_rate": 0.00016911111111111112,
-      "loss": 0.2667,
+      "learning_rate": 0.00029462068965517243,
+      "loss": 0.3183,
       "step": 1992
     },
     {
       "epoch": 36.24,
-      "learning_rate": 0.00016888888888888889,
-      "loss": 0.3362,
+      "learning_rate": 0.0002944827586206897,
+      "loss": 0.3985,
       "step": 1993
     },
     {
       "epoch": 36.25,
-      "learning_rate": 0.00016866666666666666,
-      "loss": 0.3027,
+      "learning_rate": 0.0002943448275862069,
+      "loss": 0.3737,
       "step": 1994
     },
     {
       "epoch": 36.27,
-      "learning_rate": 0.00016844444444444445,
-      "loss": 0.2823,
+      "learning_rate": 0.00029420689655172417,
+      "loss": 0.3337,
       "step": 1995
     },
     {
       "epoch": 36.29,
-      "learning_rate": 0.00016822222222222222,
-      "loss": 0.274,
+      "learning_rate": 0.0002940689655172414,
+      "loss": 0.3326,
       "step": 1996
     },
     {
       "epoch": 36.31,
-      "learning_rate": 0.00016800000000000002,
-      "loss": 0.3058,
+      "learning_rate": 0.00029393103448275865,
+      "loss": 0.3791,
       "step": 1997
     },
     {
       "epoch": 36.33,
-      "learning_rate": 0.0001677777777777778,
-      "loss": 0.2983,
+      "learning_rate": 0.00029379310344827586,
+      "loss": 0.3719,
       "step": 1998
     },
     {
       "epoch": 36.35,
-      "learning_rate": 0.00016755555555555558,
-      "loss": 0.2746,
+      "learning_rate": 0.00029365517241379313,
+      "loss": 0.3289,
       "step": 1999
     },
     {
       "epoch": 36.36,
-      "learning_rate": 0.00016733333333333333,
-      "loss": 0.3851,
+      "learning_rate": 0.00029351724137931034,
+      "loss": 0.4401,
       "step": 2000
     },
     {
       "epoch": 36.36,
-      "eval_cer": 0.08358446153615497,
-      "eval_loss": 0.35650908946990967,
-      "eval_runtime": 14.0363,
-      "eval_samples_per_second": 117.338,
-      "eval_steps_per_second": 1.852,
-      "eval_wer": 0.3458278010417731,
+      "eval_cer": 0.10012144110106598,
+      "eval_loss": 0.39910611510276794,
+      "eval_runtime": 14.0065,
+      "eval_samples_per_second": 117.588,
+      "eval_steps_per_second": 1.856,
+      "eval_wer": 0.39536308855070984,
       "step": 2000
     },
     {
       "epoch": 36.38,
-      "learning_rate": 0.0001671111111111111,
-      "loss": 0.3049,
+      "learning_rate": 0.00029337931034482755,
+      "loss": 0.3447,
       "step": 2001
     },
     {
       "epoch": 36.4,
-      "learning_rate": 0.0001668888888888889,
-      "loss": 0.2733,
+      "learning_rate": 0.0002932413793103448,
+      "loss": 0.3504,
       "step": 2002
     },
     {
       "epoch": 36.42,
-      "learning_rate": 0.00016666666666666666,
-      "loss": 0.3274,
+      "learning_rate": 0.00029310344827586203,
+      "loss": 0.4064,
       "step": 2003
     },
     {
       "epoch": 36.44,
-      "learning_rate": 0.00016644444444444446,
-      "loss": 0.268,
+      "learning_rate": 0.0002929655172413793,
+      "loss": 0.3588,
       "step": 2004
     },
     {
       "epoch": 36.45,
-      "learning_rate": 0.00016622222222222223,
-      "loss": 0.312,
+      "learning_rate": 0.0002928275862068965,
+      "loss": 0.3865,
       "step": 2005
     },
     {
       "epoch": 36.47,
-      "learning_rate": 0.00016600000000000002,
-      "loss": 0.2879,
+      "learning_rate": 0.00029268965517241383,
+      "loss": 0.3398,
       "step": 2006
     },
     {
       "epoch": 36.49,
-      "learning_rate": 0.00016577777777777777,
-      "loss": 0.2612,
+      "learning_rate": 0.00029255172413793104,
+      "loss": 0.3371,
       "step": 2007
     },
     {
       "epoch": 36.51,
-      "learning_rate": 0.00016555555555555556,
-      "loss": 0.2807,
+      "learning_rate": 0.0002924137931034483,
+      "loss": 0.3904,
       "step": 2008
     },
     {
       "epoch": 36.53,
-      "learning_rate": 0.00016533333333333333,
-      "loss": 0.3203,
+      "learning_rate": 0.0002922758620689655,
+      "loss": 0.3907,
       "step": 2009
     },
     {
       "epoch": 36.55,
-      "learning_rate": 0.0001651111111111111,
-      "loss": 0.2722,
+      "learning_rate": 0.0002921379310344828,
+      "loss": 0.345,
       "step": 2010
     },
     {
       "epoch": 36.56,
-      "learning_rate": 0.0001648888888888889,
-      "loss": 0.2902,
+      "learning_rate": 0.000292,
+      "loss": 0.3892,
       "step": 2011
     },
     {
       "epoch": 36.58,
-      "learning_rate": 0.00016466666666666667,
-      "loss": 0.2712,
+      "learning_rate": 0.00029186206896551726,
+      "loss": 0.3538,
       "step": 2012
     },
     {
       "epoch": 36.6,
-      "learning_rate": 0.00016444444444444446,
-      "loss": 0.3249,
+      "learning_rate": 0.0002917241379310345,
+      "loss": 0.3895,
       "step": 2013
     },
     {
       "epoch": 36.62,
-      "learning_rate": 0.0001642222222222222,
-      "loss": 0.2574,
+      "learning_rate": 0.00029158620689655174,
+      "loss": 0.3533,
       "step": 2014
     },
     {
       "epoch": 36.64,
-      "learning_rate": 0.000164,
-      "loss": 0.2747,
+      "learning_rate": 0.00029144827586206895,
+      "loss": 0.3486,
       "step": 2015
     },
     {
       "epoch": 36.65,
-      "learning_rate": 0.00016377777777777777,
-      "loss": 0.2688,
+      "learning_rate": 0.0002913103448275862,
+      "loss": 0.3764,
       "step": 2016
     },
     {
       "epoch": 36.67,
-      "learning_rate": 0.00016355555555555557,
-      "loss": 0.3551,
+      "learning_rate": 0.00029117241379310343,
+      "loss": 0.4126,
       "step": 2017
     },
     {
       "epoch": 36.69,
-      "learning_rate": 0.00016333333333333334,
-      "loss": 0.2912,
+      "learning_rate": 0.0002910344827586207,
+      "loss": 0.3427,
       "step": 2018
     },
     {
       "epoch": 36.71,
-      "learning_rate": 0.00016311111111111113,
-      "loss": 0.3227,
+      "learning_rate": 0.0002908965517241379,
+      "loss": 0.4022,
       "step": 2019
     },
     {
       "epoch": 36.73,
-      "learning_rate": 0.0001628888888888889,
-      "loss": 0.3083,
+      "learning_rate": 0.0002907586206896552,
+      "loss": 0.3953,
       "step": 2020
     },
     {
       "epoch": 36.75,
-      "learning_rate": 0.00016266666666666665,
-      "loss": 0.2538,
+      "learning_rate": 0.0002906206896551724,
+      "loss": 0.3484,
       "step": 2021
     },
     {
       "epoch": 36.76,
-      "learning_rate": 0.00016244444444444444,
-      "loss": 0.2912,
+      "learning_rate": 0.0002904827586206897,
+      "loss": 0.3851,
       "step": 2022
     },
     {
       "epoch": 36.78,
-      "learning_rate": 0.0001622222222222222,
-      "loss": 0.256,
+      "learning_rate": 0.0002903448275862069,
+      "loss": 0.3485,
       "step": 2023
     },
     {
       "epoch": 36.8,
-      "learning_rate": 0.000162,
-      "loss": 0.3249,
+      "learning_rate": 0.0002902068965517242,
+      "loss": 0.4022,
       "step": 2024
     },
     {
       "epoch": 36.82,
-      "learning_rate": 0.00016177777777777778,
-      "loss": 0.3132,
+      "learning_rate": 0.0002900689655172414,
+      "loss": 0.373,
       "step": 2025
     },
     {
       "epoch": 36.84,
-      "learning_rate": 0.00016155555555555557,
-      "loss": 0.3008,
+      "learning_rate": 0.00028993103448275866,
+      "loss": 0.4068,
       "step": 2026
     },
     {
       "epoch": 36.85,
-      "learning_rate": 0.00016133333333333334,
-      "loss": 0.3138,
+      "learning_rate": 0.0002897931034482759,
+      "loss": 0.3589,
       "step": 2027
     },
     {
       "epoch": 36.87,
-      "learning_rate": 0.0001611111111111111,
-      "loss": 0.3344,
+      "learning_rate": 0.00028965517241379314,
+      "loss": 0.4088,
       "step": 2028
     },
     {
       "epoch": 36.89,
-      "learning_rate": 0.00016088888888888888,
-      "loss": 0.3028,
+      "learning_rate": 0.00028951724137931035,
+      "loss": 0.3646,
       "step": 2029
     },
     {
       "epoch": 36.91,
-      "learning_rate": 0.00016066666666666668,
-      "loss": 0.2705,
+      "learning_rate": 0.0002893793103448276,
+      "loss": 0.3365,
       "step": 2030
     },
     {
       "epoch": 36.93,
-      "learning_rate": 0.00016044444444444445,
-      "loss": 0.3009,
+      "learning_rate": 0.00028924137931034483,
+      "loss": 0.3618,
       "step": 2031
     },
     {
       "epoch": 36.95,
-      "learning_rate": 0.00016022222222222222,
-      "loss": 0.3166,
+      "learning_rate": 0.0002891034482758621,
+      "loss": 0.3926,
       "step": 2032
     },
     {
       "epoch": 36.96,
-      "learning_rate": 0.00016,
-      "loss": 0.2536,
+      "learning_rate": 0.0002889655172413793,
+      "loss": 0.3197,
       "step": 2033
     },
     {
       "epoch": 36.98,
-      "learning_rate": 0.00015977777777777778,
-      "loss": 0.2672,
+      "learning_rate": 0.0002888275862068965,
+      "loss": 0.3417,
       "step": 2034
     },
     {
       "epoch": 37.0,
-      "learning_rate": 0.00015955555555555555,
-      "loss": 0.2417,
+      "learning_rate": 0.0002886896551724138,
+      "loss": 0.3238,
       "step": 2035
     },
     {
       "epoch": 37.02,
-      "learning_rate": 0.00015933333333333332,
-      "loss": 0.2891,
+      "learning_rate": 0.000288551724137931,
+      "loss": 0.3539,
       "step": 2036
     },
     {
       "epoch": 37.04,
-      "learning_rate": 0.00015911111111111112,
-      "loss": 0.2952,
+      "learning_rate": 0.00028841379310344826,
+      "loss": 0.37,
       "step": 2037
     },
     {
       "epoch": 37.05,
-      "learning_rate": 0.0001588888888888889,
-      "loss": 0.3214,
+      "learning_rate": 0.0002882758620689655,
+      "loss": 0.3959,
       "step": 2038
     },
     {
       "epoch": 37.07,
-      "learning_rate": 0.00015866666666666668,
-      "loss": 0.2818,
+      "learning_rate": 0.0002881379310344828,
+      "loss": 0.3868,
       "step": 2039
     },
     {
       "epoch": 37.09,
-      "learning_rate": 0.00015844444444444445,
-      "loss": 0.3223,
+      "learning_rate": 0.000288,
+      "loss": 0.4149,
       "step": 2040
     },
     {
       "epoch": 37.11,
-      "learning_rate": 0.00015822222222222222,
-      "loss": 0.2689,
+      "learning_rate": 0.0002878620689655173,
+      "loss": 0.3349,
       "step": 2041
     },
     {
       "epoch": 37.13,
-      "learning_rate": 0.000158,
-      "loss": 0.3587,
+      "learning_rate": 0.0002877241379310345,
+      "loss": 0.4305,
       "step": 2042
     },
     {
       "epoch": 37.15,
-      "learning_rate": 0.00015777777777777776,
-      "loss": 0.2535,
+      "learning_rate": 0.00028758620689655175,
+      "loss": 0.3458,
       "step": 2043
     },
     {
       "epoch": 37.16,
-      "learning_rate": 0.00015755555555555556,
-      "loss": 0.2479,
+      "learning_rate": 0.00028744827586206896,
+      "loss": 0.3329,
       "step": 2044
     },
     {
       "epoch": 37.18,
-      "learning_rate": 0.00015733333333333333,
-      "loss": 0.296,
+      "learning_rate": 0.00028731034482758623,
+      "loss": 0.3509,
       "step": 2045
     },
     {
       "epoch": 37.2,
-      "learning_rate": 0.00015711111111111112,
-      "loss": 0.2759,
+      "learning_rate": 0.00028717241379310344,
+      "loss": 0.3679,
       "step": 2046
     },
     {
       "epoch": 37.22,
-      "learning_rate": 0.0001568888888888889,
-      "loss": 0.3435,
+      "learning_rate": 0.0002870344827586207,
+      "loss": 0.4038,
       "step": 2047
     },
     {
       "epoch": 37.24,
-      "learning_rate": 0.0001566666666666667,
-      "loss": 0.2664,
+      "learning_rate": 0.0002868965517241379,
+      "loss": 0.3503,
       "step": 2048
     },
     {
       "epoch": 37.25,
-      "learning_rate": 0.00015644444444444443,
-      "loss": 0.302,
+      "learning_rate": 0.0002867586206896552,
+      "loss": 0.3665,
       "step": 2049
     },
     {
       "epoch": 37.27,
-      "learning_rate": 0.00015622222222222223,
-      "loss": 0.2867,
+      "learning_rate": 0.0002866206896551724,
+      "loss": 0.3929,
       "step": 2050
     },
     {
       "epoch": 37.29,
-      "learning_rate": 0.000156,
-      "loss": 0.2907,
+      "learning_rate": 0.00028648275862068966,
+      "loss": 0.3319,
       "step": 2051
     },
     {
       "epoch": 37.31,
-      "learning_rate": 0.00015577777777777777,
-      "loss": 0.2916,
+      "learning_rate": 0.0002863448275862069,
+      "loss": 0.3733,
       "step": 2052
     },
     {
       "epoch": 37.33,
-      "learning_rate": 0.00015555555555555556,
-      "loss": 0.2798,
+      "learning_rate": 0.00028620689655172414,
+      "loss": 0.3519,
       "step": 2053
     },
     {
       "epoch": 37.35,
-      "learning_rate": 0.00015533333333333333,
-      "loss": 0.2813,
+      "learning_rate": 0.00028606896551724135,
+      "loss": 0.3777,
       "step": 2054
     },
     {
       "epoch": 37.36,
-      "learning_rate": 0.00015511111111111113,
-      "loss": 0.2722,
+      "learning_rate": 0.0002859310344827586,
+      "loss": 0.3618,
       "step": 2055
     },
     {
       "epoch": 37.38,
-      "learning_rate": 0.0001548888888888889,
-      "loss": 0.291,
+      "learning_rate": 0.0002857931034482759,
+      "loss": 0.4012,
       "step": 2056
     },
     {
       "epoch": 37.4,
-      "learning_rate": 0.00015466666666666667,
-      "loss": 0.24,
+      "learning_rate": 0.00028565517241379315,
+      "loss": 0.2959,
       "step": 2057
     },
     {
       "epoch": 37.42,
-      "learning_rate": 0.00015444444444444444,
-      "loss": 0.3005,
+      "learning_rate": 0.00028551724137931036,
+      "loss": 0.3896,
       "step": 2058
     },
     {
       "epoch": 37.44,
-      "learning_rate": 0.00015422222222222223,
-      "loss": 0.2808,
+      "learning_rate": 0.00028537931034482763,
+      "loss": 0.3916,
       "step": 2059
     },
     {
       "epoch": 37.45,
-      "learning_rate": 0.000154,
-      "loss": 0.2648,
+      "learning_rate": 0.00028524137931034484,
+      "loss": 0.3124,
       "step": 2060
     },
     {
       "epoch": 37.47,
-      "learning_rate": 0.0001537777777777778,
-      "loss": 0.293,
+      "learning_rate": 0.0002851034482758621,
+      "loss": 0.3561,
       "step": 2061
     },
     {
       "epoch": 37.49,
-      "learning_rate": 0.00015355555555555557,
-      "loss": 0.269,
+      "learning_rate": 0.0002849655172413793,
+      "loss": 0.341,
       "step": 2062
     },
     {
       "epoch": 37.51,
-      "learning_rate": 0.00015333333333333334,
-      "loss": 0.2585,
+      "learning_rate": 0.0002848275862068966,
+      "loss": 0.3081,
       "step": 2063
     },
     {
       "epoch": 37.53,
-      "learning_rate": 0.0001531111111111111,
-      "loss": 0.2171,
+      "learning_rate": 0.0002846896551724138,
+      "loss": 0.3104,
       "step": 2064
     },
     {
       "epoch": 37.55,
-      "learning_rate": 0.00015288888888888888,
-      "loss": 0.2534,
+      "learning_rate": 0.00028455172413793106,
+      "loss": 0.3042,
       "step": 2065
     },
     {
       "epoch": 37.56,
-      "learning_rate": 0.00015266666666666667,
-      "loss": 0.247,
+      "learning_rate": 0.0002844137931034483,
+      "loss": 0.3305,
       "step": 2066
     },
     {
       "epoch": 37.58,
-      "learning_rate": 0.00015244444444444444,
-      "loss": 0.2807,
+      "learning_rate": 0.0002842758620689655,
+      "loss": 0.3726,
       "step": 2067
     },
     {
       "epoch": 37.6,
-      "learning_rate": 0.00015222222222222224,
-      "loss": 0.2659,
+      "learning_rate": 0.00028413793103448275,
+      "loss": 0.3379,
       "step": 2068
     },
     {
       "epoch": 37.62,
-      "learning_rate": 0.000152,
-      "loss": 0.2323,
+      "learning_rate": 0.00028399999999999996,
+      "loss": 0.3324,
       "step": 2069
     },
     {
       "epoch": 37.64,
-      "learning_rate": 0.0001517777777777778,
-      "loss": 0.2832,
+      "learning_rate": 0.00028386206896551723,
+      "loss": 0.3655,
       "step": 2070
     },
     {
       "epoch": 37.65,
-      "learning_rate": 0.00015155555555555555,
-      "loss": 0.2549,
+      "learning_rate": 0.00028372413793103444,
+      "loss": 0.3214,
       "step": 2071
     },
     {
       "epoch": 37.67,
-      "learning_rate": 0.00015133333333333332,
-      "loss": 0.3242,
+      "learning_rate": 0.0002835862068965517,
+      "loss": 0.4342,
       "step": 2072
     },
     {
       "epoch": 37.69,
-      "learning_rate": 0.0001511111111111111,
-      "loss": 0.2614,
+      "learning_rate": 0.000283448275862069,
+      "loss": 0.3329,
       "step": 2073
     },
     {
       "epoch": 37.71,
-      "learning_rate": 0.00015088888888888888,
-      "loss": 0.2886,
+      "learning_rate": 0.00028331034482758624,
+      "loss": 0.325,
       "step": 2074
     },
     {
       "epoch": 37.73,
-      "learning_rate": 0.00015066666666666668,
-      "loss": 0.3216,
+      "learning_rate": 0.00028317241379310345,
+      "loss": 0.3681,
       "step": 2075
     },
     {
       "epoch": 37.75,
-      "learning_rate": 0.00015044444444444445,
-      "loss": 0.3099,
+      "learning_rate": 0.0002830344827586207,
+      "loss": 0.3415,
       "step": 2076
     },
     {
       "epoch": 37.76,
-      "learning_rate": 0.00015022222222222224,
-      "loss": 0.2918,
+      "learning_rate": 0.00028289655172413793,
+      "loss": 0.3533,
       "step": 2077
     },
     {
       "epoch": 37.78,
-      "learning_rate": 0.00015,
-      "loss": 0.2914,
+      "learning_rate": 0.0002827586206896552,
+      "loss": 0.3953,
       "step": 2078
     },
     {
       "epoch": 37.8,
-      "learning_rate": 0.00014977777777777778,
-      "loss": 0.2541,
+      "learning_rate": 0.0002826206896551724,
+      "loss": 0.3424,
       "step": 2079
     },
     {
       "epoch": 37.82,
-      "learning_rate": 0.00014955555555555555,
-      "loss": 0.2823,
+      "learning_rate": 0.0002824827586206897,
+      "loss": 0.3328,
       "step": 2080
     },
     {
       "epoch": 37.84,
-      "learning_rate": 0.00014933333333333335,
-      "loss": 0.2604,
+      "learning_rate": 0.0002823448275862069,
+      "loss": 0.3353,
       "step": 2081
     },
     {
       "epoch": 37.85,
-      "learning_rate": 0.00014911111111111112,
-      "loss": 0.2972,
+      "learning_rate": 0.00028220689655172415,
+      "loss": 0.3678,
       "step": 2082
     },
     {
       "epoch": 37.87,
-      "learning_rate": 0.0001488888888888889,
-      "loss": 0.3291,
+      "learning_rate": 0.00028206896551724136,
+      "loss": 0.3995,
       "step": 2083
     },
     {
       "epoch": 37.89,
-      "learning_rate": 0.00014866666666666668,
-      "loss": 0.2916,
+      "learning_rate": 0.00028193103448275863,
+      "loss": 0.3535,
       "step": 2084
     },
     {
       "epoch": 37.91,
-      "learning_rate": 0.00014844444444444443,
-      "loss": 0.3394,
+      "learning_rate": 0.00028179310344827584,
+      "loss": 0.3783,
       "step": 2085
     },
     {
       "epoch": 37.93,
-      "learning_rate": 0.00014822222222222222,
-      "loss": 0.2976,
+      "learning_rate": 0.0002816551724137931,
+      "loss": 0.417,
       "step": 2086
     },
     {
       "epoch": 37.95,
-      "learning_rate": 0.000148,
-      "loss": 0.2936,
+      "learning_rate": 0.0002815172413793103,
+      "loss": 0.3476,
       "step": 2087
     },
     {
       "epoch": 37.96,
-      "learning_rate": 0.0001477777777777778,
-      "loss": 0.3071,
+      "learning_rate": 0.0002813793103448276,
+      "loss": 0.378,
       "step": 2088
     },
     {
       "epoch": 37.98,
-      "learning_rate": 0.00014755555555555556,
-      "loss": 0.2845,
+      "learning_rate": 0.0002812413793103448,
+      "loss": 0.3391,
       "step": 2089
     },
     {
       "epoch": 38.0,
-      "learning_rate": 0.00014733333333333335,
-      "loss": 0.2253,
+      "learning_rate": 0.0002811034482758621,
+      "loss": 0.2534,
       "step": 2090
     },
     {
       "epoch": 38.02,
-      "learning_rate": 0.00014711111111111112,
-      "loss": 0.3242,
+      "learning_rate": 0.00028096551724137933,
+      "loss": 0.3683,
       "step": 2091
     },
     {
       "epoch": 38.04,
-      "learning_rate": 0.0001468888888888889,
-      "loss": 0.2656,
+      "learning_rate": 0.0002808275862068966,
+      "loss": 0.3156,
       "step": 2092
     },
     {
       "epoch": 38.05,
-      "learning_rate": 0.00014666666666666666,
-      "loss": 0.2549,
+      "learning_rate": 0.0002806896551724138,
+      "loss": 0.32,
       "step": 2093
     },
     {
       "epoch": 38.07,
-      "learning_rate": 0.00014644444444444443,
-      "loss": 0.2841,
+      "learning_rate": 0.0002805517241379311,
+      "loss": 0.3878,
       "step": 2094
     },
     {
       "epoch": 38.09,
-      "learning_rate": 0.00014622222222222223,
-      "loss": 0.2325,
+      "learning_rate": 0.0002804137931034483,
+      "loss": 0.3286,
       "step": 2095
     },
     {
       "epoch": 38.11,
-      "learning_rate": 0.000146,
-      "loss": 0.2666,
+      "learning_rate": 0.00028027586206896555,
+      "loss": 0.3366,
       "step": 2096
     },
     {
       "epoch": 38.13,
-      "learning_rate": 0.0001457777777777778,
-      "loss": 0.3092,
+      "learning_rate": 0.00028013793103448276,
+      "loss": 0.3139,
       "step": 2097
     },
     {
       "epoch": 38.15,
-      "learning_rate": 0.00014555555555555556,
-      "loss": 0.2951,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 0.3821,
       "step": 2098
     },
     {
       "epoch": 38.16,
-      "learning_rate": 0.00014533333333333333,
-      "loss": 0.2524,
+      "learning_rate": 0.00027986206896551724,
+      "loss": 0.3068,
       "step": 2099
     },
     {
       "epoch": 38.18,
-      "learning_rate": 0.0001451111111111111,
-      "loss": 0.3432,
+      "learning_rate": 0.00027972413793103445,
+      "loss": 0.3986,
       "step": 2100
     },
     {
       "epoch": 38.2,
-      "learning_rate": 0.0001448888888888889,
-      "loss": 0.2871,
+      "learning_rate": 0.0002795862068965517,
+      "loss": 0.3974,
       "step": 2101
     },
     {
       "epoch": 38.22,
-      "learning_rate": 0.00014466666666666667,
-      "loss": 0.2756,
+      "learning_rate": 0.00027944827586206893,
+      "loss": 0.3421,
       "step": 2102
     },
     {
       "epoch": 38.24,
-      "learning_rate": 0.00014444444444444444,
-      "loss": 0.2405,
+      "learning_rate": 0.0002793103448275862,
+      "loss": 0.3143,
       "step": 2103
     },
     {
       "epoch": 38.25,
-      "learning_rate": 0.00014422222222222223,
-      "loss": 0.3029,
+      "learning_rate": 0.0002791724137931034,
+      "loss": 0.3653,
       "step": 2104
     },
     {
       "epoch": 38.27,
-      "learning_rate": 0.000144,
-      "loss": 0.25,
+      "learning_rate": 0.0002790344827586207,
+      "loss": 0.2832,
       "step": 2105
     },
     {
       "epoch": 38.29,
-      "learning_rate": 0.00014377777777777777,
-      "loss": 0.2641,
+      "learning_rate": 0.0002790344827586207,
+      "loss": 0.319,
       "step": 2106
     },
     {
       "epoch": 38.31,
-      "learning_rate": 0.00014355555555555554,
-      "loss": 0.2771,
+      "learning_rate": 0.00027889655172413794,
+      "loss": 0.3589,
       "step": 2107
     },
     {
       "epoch": 38.33,
-      "learning_rate": 0.00014333333333333334,
-      "loss": 0.2919,
+      "learning_rate": 0.0002787586206896552,
+      "loss": 0.3523,
       "step": 2108
     },
     {
       "epoch": 38.35,
-      "learning_rate": 0.0001431111111111111,
-      "loss": 0.3052,
+      "learning_rate": 0.0002786206896551724,
+      "loss": 0.3674,
       "step": 2109
     },
     {
       "epoch": 38.36,
-      "learning_rate": 0.0001428888888888889,
-      "loss": 0.3006,
+      "learning_rate": 0.0002784827586206897,
+      "loss": 0.3506,
       "step": 2110
     },
     {
       "epoch": 38.38,
-      "learning_rate": 0.00014266666666666667,
-      "loss": 0.2,
+      "learning_rate": 0.0002783448275862069,
+      "loss": 0.3024,
       "step": 2111
     },
     {
       "epoch": 38.4,
-      "learning_rate": 0.00014244444444444447,
-      "loss": 0.3029,
+      "learning_rate": 0.00027820689655172416,
+      "loss": 0.3736,
       "step": 2112
     },
     {
       "epoch": 38.42,
-      "learning_rate": 0.0001422222222222222,
-      "loss": 0.2668,
+      "learning_rate": 0.0002780689655172414,
+      "loss": 0.338,
       "step": 2113
     },
     {
       "epoch": 38.44,
-      "learning_rate": 0.00014199999999999998,
-      "loss": 0.2551,
+      "learning_rate": 0.00027793103448275864,
+      "loss": 0.3171,
       "step": 2114
     },
     {
       "epoch": 38.45,
-      "learning_rate": 0.00014177777777777778,
-      "loss": 0.3223,
+      "learning_rate": 0.00027779310344827585,
+      "loss": 0.3542,
       "step": 2115
     },
     {
       "epoch": 38.47,
-      "learning_rate": 0.00014155555555555555,
-      "loss": 0.2701,
+      "learning_rate": 0.0002776551724137931,
+      "loss": 0.3364,
       "step": 2116
     },
     {
       "epoch": 38.49,
-      "learning_rate": 0.00014133333333333334,
-      "loss": 0.2577,
+      "learning_rate": 0.00027751724137931033,
+      "loss": 0.3452,
       "step": 2117
     },
     {
       "epoch": 38.51,
-      "learning_rate": 0.00014111111111111111,
-      "loss": 0.2807,
+      "learning_rate": 0.0002773793103448276,
+      "loss": 0.3323,
       "step": 2118
     },
     {
       "epoch": 38.53,
-      "learning_rate": 0.0001408888888888889,
-      "loss": 0.2171,
+      "learning_rate": 0.0002772413793103448,
+      "loss": 0.2994,
       "step": 2119
     },
     {
       "epoch": 38.55,
-      "learning_rate": 0.00014066666666666665,
-      "loss": 0.2694,
+      "learning_rate": 0.0002771034482758621,
+      "loss": 0.3805,
       "step": 2120
     },
     {
       "epoch": 38.56,
-      "learning_rate": 0.00014044444444444445,
-      "loss": 0.2101,
+      "learning_rate": 0.0002769655172413793,
+      "loss": 0.3239,
       "step": 2121
     },
     {
       "epoch": 38.58,
-      "learning_rate": 0.00014022222222222222,
-      "loss": 0.2913,
+      "learning_rate": 0.00027682758620689655,
+      "loss": 0.3854,
       "step": 2122
     },
     {
       "epoch": 38.6,
-      "learning_rate": 0.00014000000000000001,
-      "loss": 0.2479,
+      "learning_rate": 0.00027668965517241376,
+      "loss": 0.35,
       "step": 2123
     },
     {
       "epoch": 38.62,
-      "learning_rate": 0.00013977777777777778,
-      "loss": 0.2332,
+      "learning_rate": 0.0002765517241379311,
+      "loss": 0.3215,
       "step": 2124
     },
     {
       "epoch": 38.64,
-      "learning_rate": 0.00013955555555555555,
-      "loss": 0.2483,
+      "learning_rate": 0.0002764137931034483,
+      "loss": 0.2934,
       "step": 2125
     },
     {
       "epoch": 38.65,
-      "learning_rate": 0.00013933333333333335,
-      "loss": 0.2895,
+      "learning_rate": 0.00027627586206896556,
+      "loss": 0.4105,
       "step": 2126
     },
     {
       "epoch": 38.67,
-      "learning_rate": 0.0001391111111111111,
-      "loss": 0.2817,
+      "learning_rate": 0.0002761379310344828,
+      "loss": 0.3082,
       "step": 2127
     },
     {
       "epoch": 38.69,
-      "learning_rate": 0.0001388888888888889,
-      "loss": 0.3187,
+      "learning_rate": 0.00027600000000000004,
+      "loss": 0.4386,
       "step": 2128
     },
     {
       "epoch": 38.71,
-      "learning_rate": 0.00013866666666666666,
-      "loss": 0.3315,
+      "learning_rate": 0.00027586206896551725,
+      "loss": 0.3839,
       "step": 2129
     },
     {
       "epoch": 38.73,
-      "learning_rate": 0.00013844444444444445,
-      "loss": 0.295,
+      "learning_rate": 0.0002757241379310345,
+      "loss": 0.3284,
       "step": 2130
     },
     {
       "epoch": 38.75,
-      "learning_rate": 0.00013822222222222222,
-      "loss": 0.2704,
+      "learning_rate": 0.00027558620689655173,
+      "loss": 0.3736,
       "step": 2131
     },
     {
       "epoch": 38.76,
-      "learning_rate": 0.00013800000000000002,
-      "loss": 0.2067,
+      "learning_rate": 0.000275448275862069,
+      "loss": 0.2752,
       "step": 2132
     },
     {
       "epoch": 38.78,
-      "learning_rate": 0.0001377777777777778,
-      "loss": 0.2579,
+      "learning_rate": 0.0002753103448275862,
+      "loss": 0.3397,
       "step": 2133
     },
     {
       "epoch": 38.8,
-      "learning_rate": 0.00013755555555555553,
-      "loss": 0.2794,
+      "learning_rate": 0.0002751724137931034,
+      "loss": 0.3777,
       "step": 2134
     },
     {
       "epoch": 38.82,
-      "learning_rate": 0.00013733333333333333,
-      "loss": 0.2576,
+      "learning_rate": 0.0002750344827586207,
+      "loss": 0.3211,
       "step": 2135
     },
     {
       "epoch": 38.84,
-      "learning_rate": 0.0001371111111111111,
-      "loss": 0.2566,
+      "learning_rate": 0.0002748965517241379,
+      "loss": 0.3258,
       "step": 2136
     },
     {
       "epoch": 38.85,
-      "learning_rate": 0.0001368888888888889,
-      "loss": 0.2768,
+      "learning_rate": 0.00027475862068965516,
+      "loss": 0.3168,
       "step": 2137
     },
     {
       "epoch": 38.87,
-      "learning_rate": 0.00013666666666666666,
-      "loss": 0.3329,
+      "learning_rate": 0.0002746206896551724,
+      "loss": 0.3877,
       "step": 2138
     },
     {
       "epoch": 38.89,
-      "learning_rate": 0.00013644444444444446,
-      "loss": 0.279,
+      "learning_rate": 0.00027448275862068964,
+      "loss": 0.3778,
       "step": 2139
     },
     {
       "epoch": 38.91,
-      "learning_rate": 0.00013622222222222223,
-      "loss": 0.2681,
+      "learning_rate": 0.00027434482758620685,
+      "loss": 0.3432,
       "step": 2140
     },
     {
       "epoch": 38.93,
-      "learning_rate": 0.00013600000000000003,
-      "loss": 0.3474,
+      "learning_rate": 0.0002742068965517242,
+      "loss": 0.3351,
       "step": 2141
     },
     {
       "epoch": 38.95,
-      "learning_rate": 0.00013577777777777777,
-      "loss": 0.3703,
+      "learning_rate": 0.0002740689655172414,
+      "loss": 0.4146,
       "step": 2142
     },
     {
       "epoch": 38.96,
-      "learning_rate": 0.00013555555555555556,
-      "loss": 0.2882,
+      "learning_rate": 0.00027393103448275865,
+      "loss": 0.3815,
       "step": 2143
     },
     {
       "epoch": 38.98,
-      "learning_rate": 0.00013533333333333333,
-      "loss": 0.2593,
+      "learning_rate": 0.00027379310344827586,
+      "loss": 0.3001,
       "step": 2144
     },
     {
       "epoch": 39.0,
-      "learning_rate": 0.0001351111111111111,
-      "loss": 0.3236,
+      "learning_rate": 0.00027365517241379313,
+      "loss": 0.3978,
       "step": 2145
     },
     {
       "epoch": 39.02,
-      "learning_rate": 0.0001348888888888889,
-      "loss": 0.2972,
+      "learning_rate": 0.00027351724137931034,
+      "loss": 0.3462,
       "step": 2146
     },
     {
       "epoch": 39.04,
-      "learning_rate": 0.00013466666666666667,
-      "loss": 0.2629,
+      "learning_rate": 0.0002733793103448276,
+      "loss": 0.3018,
       "step": 2147
     },
     {
       "epoch": 39.05,
-      "learning_rate": 0.00013444444444444447,
-      "loss": 0.2645,
+      "learning_rate": 0.0002732413793103448,
+      "loss": 0.3231,
       "step": 2148
     },
     {
       "epoch": 39.07,
-      "learning_rate": 0.0001342222222222222,
-      "loss": 0.2688,
+      "learning_rate": 0.0002731034482758621,
+      "loss": 0.3518,
       "step": 2149
     },
     {
       "epoch": 39.09,
-      "learning_rate": 0.000134,
-      "loss": 0.3579,
+      "learning_rate": 0.0002729655172413793,
+      "loss": 0.4083,
       "step": 2150
     },
     {
       "epoch": 39.11,
-      "learning_rate": 0.00013377777777777777,
-      "loss": 0.2714,
+      "learning_rate": 0.00027282758620689656,
+      "loss": 0.3521,
       "step": 2151
     },
     {
       "epoch": 39.13,
-      "learning_rate": 0.00013355555555555557,
-      "loss": 0.2476,
+      "learning_rate": 0.0002726896551724138,
+      "loss": 0.3414,
       "step": 2152
     },
     {
       "epoch": 39.15,
-      "learning_rate": 0.00013333333333333334,
-      "loss": 0.2703,
+      "learning_rate": 0.00027255172413793104,
+      "loss": 0.3534,
       "step": 2153
     },
     {
       "epoch": 39.16,
-      "learning_rate": 0.00013311111111111114,
-      "loss": 0.2582,
+      "learning_rate": 0.00027241379310344825,
+      "loss": 0.3271,
       "step": 2154
     },
     {
       "epoch": 39.18,
-      "learning_rate": 0.0001328888888888889,
-      "loss": 0.2743,
+      "learning_rate": 0.0002722758620689655,
+      "loss": 0.3666,
       "step": 2155
     },
     {
       "epoch": 39.2,
-      "learning_rate": 0.00013266666666666665,
-      "loss": 0.2928,
+      "learning_rate": 0.00027213793103448273,
+      "loss": 0.3734,
       "step": 2156
     },
     {
       "epoch": 39.22,
-      "learning_rate": 0.00013244444444444444,
-      "loss": 0.2679,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 0.3524,
       "step": 2157
     },
     {
       "epoch": 39.24,
-      "learning_rate": 0.00013222222222222221,
-      "loss": 0.2076,
+      "learning_rate": 0.00027186206896551726,
+      "loss": 0.2832,
       "step": 2158
     },
     {
       "epoch": 39.25,
-      "learning_rate": 0.000132,
-      "loss": 0.221,
+      "learning_rate": 0.00027172413793103453,
+      "loss": 0.292,
       "step": 2159
     },
     {
       "epoch": 39.27,
-      "learning_rate": 0.00013177777777777778,
-      "loss": 0.2426,
+      "learning_rate": 0.00027158620689655174,
+      "loss": 0.2936,
       "step": 2160
     },
     {
       "epoch": 39.29,
-      "learning_rate": 0.00013155555555555558,
-      "loss": 0.2937,
+      "learning_rate": 0.000271448275862069,
+      "loss": 0.3388,
       "step": 2161
     },
     {
       "epoch": 39.31,
-      "learning_rate": 0.00013133333333333335,
-      "loss": 0.2696,
+      "learning_rate": 0.0002713103448275862,
+      "loss": 0.3533,
       "step": 2162
     },
     {
       "epoch": 39.33,
-      "learning_rate": 0.00013111111111111111,
-      "loss": 0.2505,
+      "learning_rate": 0.0002711724137931035,
+      "loss": 0.3237,
       "step": 2163
     },
     {
       "epoch": 39.35,
-      "learning_rate": 0.00013088888888888888,
-      "loss": 0.2283,
+      "learning_rate": 0.0002710344827586207,
+      "loss": 0.3125,
       "step": 2164
     },
     {
       "epoch": 39.36,
-      "learning_rate": 0.00013066666666666665,
-      "loss": 0.3122,
+      "learning_rate": 0.00027089655172413796,
+      "loss": 0.3642,
       "step": 2165
     },
     {
       "epoch": 39.38,
-      "learning_rate": 0.00013044444444444445,
-      "loss": 0.2673,
+      "learning_rate": 0.0002707586206896552,
+      "loss": 0.3079,
       "step": 2166
     },
     {
       "epoch": 39.4,
-      "learning_rate": 0.00013022222222222222,
-      "loss": 0.2737,
+      "learning_rate": 0.0002706206896551724,
+      "loss": 0.3051,
       "step": 2167
     },
     {
       "epoch": 39.42,
-      "learning_rate": 0.00013000000000000002,
-      "loss": 0.3326,
+      "learning_rate": 0.00027048275862068965,
+      "loss": 0.3899,
       "step": 2168
     },
     {
       "epoch": 39.44,
-      "learning_rate": 0.00012977777777777779,
-      "loss": 0.3363,
+      "learning_rate": 0.00027034482758620687,
+      "loss": 0.4303,
       "step": 2169
     },
     {
       "epoch": 39.45,
-      "learning_rate": 0.00012955555555555555,
-      "loss": 0.2058,
+      "learning_rate": 0.00027020689655172413,
+      "loss": 0.3077,
       "step": 2170
     },
     {
       "epoch": 39.47,
-      "learning_rate": 0.00012933333333333332,
-      "loss": 0.2149,
+      "learning_rate": 0.00027006896551724134,
+      "loss": 0.2959,
       "step": 2171
     },
     {
       "epoch": 39.49,
-      "learning_rate": 0.00012911111111111112,
-      "loss": 0.2789,
+      "learning_rate": 0.0002699310344827586,
+      "loss": 0.3769,
       "step": 2172
     },
     {
       "epoch": 39.51,
-      "learning_rate": 0.0001288888888888889,
-      "loss": 0.3027,
+      "learning_rate": 0.0002697931034482758,
+      "loss": 0.3774,
       "step": 2173
     },
     {
       "epoch": 39.53,
-      "learning_rate": 0.00012866666666666669,
-      "loss": 0.2563,
+      "learning_rate": 0.00026965517241379314,
+      "loss": 0.3104,
       "step": 2174
     },
     {
       "epoch": 39.55,
-      "learning_rate": 0.00012844444444444446,
-      "loss": 0.3092,
+      "learning_rate": 0.00026951724137931035,
+      "loss": 0.372,
       "step": 2175
     },
     {
       "epoch": 39.56,
-      "learning_rate": 0.00012822222222222222,
-      "loss": 0.2632,
+      "learning_rate": 0.0002693793103448276,
+      "loss": 0.3279,
       "step": 2176
     },
     {
       "epoch": 39.58,
-      "learning_rate": 0.000128,
-      "loss": 0.294,
+      "learning_rate": 0.00026924137931034483,
+      "loss": 0.3673,
       "step": 2177
     },
     {
       "epoch": 39.6,
-      "learning_rate": 0.00012777777777777776,
-      "loss": 0.2676,
+      "learning_rate": 0.0002691034482758621,
+      "loss": 0.3266,
       "step": 2178
     },
     {
       "epoch": 39.62,
-      "learning_rate": 0.00012755555555555556,
-      "loss": 0.2457,
+      "learning_rate": 0.0002689655172413793,
+      "loss": 0.3341,
       "step": 2179
     },
     {
       "epoch": 39.64,
-      "learning_rate": 0.00012733333333333333,
-      "loss": 0.2549,
+      "learning_rate": 0.0002688275862068966,
+      "loss": 0.3489,
       "step": 2180
     },
     {
       "epoch": 39.65,
-      "learning_rate": 0.00012711111111111113,
-      "loss": 0.2876,
+      "learning_rate": 0.0002686896551724138,
+      "loss": 0.3543,
       "step": 2181
     },
     {
       "epoch": 39.67,
-      "learning_rate": 0.0001268888888888889,
-      "loss": 0.2367,
+      "learning_rate": 0.00026855172413793105,
+      "loss": 0.3469,
       "step": 2182
     },
     {
       "epoch": 39.69,
-      "learning_rate": 0.0001266666666666667,
-      "loss": 0.2401,
+      "learning_rate": 0.00026841379310344827,
+      "loss": 0.3077,
       "step": 2183
     },
     {
       "epoch": 39.71,
-      "learning_rate": 0.00012644444444444443,
-      "loss": 0.2841,
+      "learning_rate": 0.00026827586206896553,
+      "loss": 0.3672,
       "step": 2184
     },
     {
       "epoch": 39.73,
-      "learning_rate": 0.00012622222222222223,
-      "loss": 0.2325,
+      "learning_rate": 0.00026813793103448274,
+      "loss": 0.3181,
       "step": 2185
     },
     {
       "epoch": 39.75,
-      "learning_rate": 0.000126,
-      "loss": 0.276,
+      "learning_rate": 0.000268,
+      "loss": 0.3183,
       "step": 2186
     },
     {
       "epoch": 39.76,
-      "learning_rate": 0.00012577777777777777,
-      "loss": 0.2604,
+      "learning_rate": 0.0002678620689655172,
+      "loss": 0.3433,
       "step": 2187
     },
     {
       "epoch": 39.78,
-      "learning_rate": 0.00012555555555555557,
-      "loss": 0.279,
+      "learning_rate": 0.0002677241379310345,
+      "loss": 0.3875,
       "step": 2188
     },
     {
       "epoch": 39.8,
-      "learning_rate": 0.00012533333333333334,
-      "loss": 0.2652,
+      "learning_rate": 0.0002675862068965517,
+      "loss": 0.3107,
       "step": 2189
     },
     {
       "epoch": 39.82,
-      "learning_rate": 0.00012511111111111113,
-      "loss": 0.2504,
+      "learning_rate": 0.00026744827586206896,
+      "loss": 0.3411,
       "step": 2190
     },
     {
       "epoch": 39.84,
-      "learning_rate": 0.00012488888888888887,
-      "loss": 0.2486,
+      "learning_rate": 0.00026731034482758623,
+      "loss": 0.3193,
       "step": 2191
     },
     {
       "epoch": 39.85,
-      "learning_rate": 0.00012466666666666667,
-      "loss": 0.2958,
+      "learning_rate": 0.0002671724137931035,
+      "loss": 0.3922,
       "step": 2192
     },
     {
       "epoch": 39.87,
-      "learning_rate": 0.00012444444444444444,
-      "loss": 0.2517,
+      "learning_rate": 0.0002670344827586207,
+      "loss": 0.3452,
       "step": 2193
     },
     {
       "epoch": 39.89,
-      "learning_rate": 0.00012422222222222224,
-      "loss": 0.2803,
+      "learning_rate": 0.000266896551724138,
+      "loss": 0.3881,
       "step": 2194
     },
     {
       "epoch": 39.91,
-      "learning_rate": 0.000124,
-      "loss": 0.2249,
+      "learning_rate": 0.0002667586206896552,
+      "loss": 0.2758,
       "step": 2195
     },
     {
       "epoch": 39.93,
-      "learning_rate": 0.00012377777777777777,
-      "loss": 0.261,
+      "learning_rate": 0.00026662068965517245,
+      "loss": 0.329,
       "step": 2196
     },
     {
       "epoch": 39.95,
-      "learning_rate": 0.00012355555555555557,
-      "loss": 0.246,
+      "learning_rate": 0.00026648275862068966,
+      "loss": 0.3412,
       "step": 2197
     },
     {
       "epoch": 39.96,
-      "learning_rate": 0.00012333333333333334,
-      "loss": 0.2755,
+      "learning_rate": 0.00026634482758620693,
+      "loss": 0.3547,
       "step": 2198
     },
     {
       "epoch": 39.98,
-      "learning_rate": 0.0001231111111111111,
-      "loss": 0.2422,
+      "learning_rate": 0.00026620689655172414,
+      "loss": 0.3359,
       "step": 2199
     },
     {
       "epoch": 40.0,
-      "learning_rate": 0.0001228888888888889,
-      "loss": 0.2799,
+      "learning_rate": 0.0002660689655172414,
+      "loss": 0.3241,
       "step": 2200
     },
     {
       "epoch": 40.02,
-      "learning_rate": 0.00012266666666666668,
-      "loss": 0.2979,
+      "learning_rate": 0.0002659310344827586,
+      "loss": 0.3299,
       "step": 2201
     },
     {
       "epoch": 40.04,
-      "learning_rate": 0.00012244444444444445,
-      "loss": 0.25,
+      "learning_rate": 0.00026579310344827583,
+      "loss": 0.314,
       "step": 2202
     },
     {
       "epoch": 40.05,
-      "learning_rate": 0.00012222222222222221,
-      "loss": 0.2546,
+      "learning_rate": 0.0002656551724137931,
+      "loss": 0.3253,
       "step": 2203
     },
     {
       "epoch": 40.07,
-      "learning_rate": 0.000122,
-      "loss": 0.2788,
+      "learning_rate": 0.0002655172413793103,
+      "loss": 0.3569,
       "step": 2204
     },
     {
       "epoch": 40.09,
-      "learning_rate": 0.00012177777777777778,
-      "loss": 0.2833,
+      "learning_rate": 0.0002653793103448276,
+      "loss": 0.3693,
       "step": 2205
     },
     {
       "epoch": 40.11,
-      "learning_rate": 0.00012155555555555556,
-      "loss": 0.1849,
+      "learning_rate": 0.0002652413793103448,
+      "loss": 0.2778,
       "step": 2206
     },
     {
       "epoch": 40.13,
-      "learning_rate": 0.00012133333333333333,
-      "loss": 0.2564,
+      "learning_rate": 0.00026510344827586205,
+      "loss": 0.3567,
       "step": 2207
     },
     {
       "epoch": 40.15,
-      "learning_rate": 0.00012111111111111112,
-      "loss": 0.291,
+      "learning_rate": 0.0002649655172413793,
+      "loss": 0.3349,
       "step": 2208
     },
     {
       "epoch": 40.16,
-      "learning_rate": 0.0001208888888888889,
-      "loss": 0.2765,
+      "learning_rate": 0.0002648275862068966,
+      "loss": 0.3278,
       "step": 2209
     },
     {
       "epoch": 40.18,
-      "learning_rate": 0.00012066666666666667,
-      "loss": 0.261,
+      "learning_rate": 0.0002646896551724138,
+      "loss": 0.3345,
       "step": 2210
     },
     {
       "epoch": 40.2,
-      "learning_rate": 0.00012044444444444445,
-      "loss": 0.2741,
+      "learning_rate": 0.00026455172413793106,
+      "loss": 0.3042,
       "step": 2211
     },
     {
       "epoch": 40.22,
-      "learning_rate": 0.00012022222222222222,
-      "loss": 0.2066,
+      "learning_rate": 0.0002644137931034483,
+      "loss": 0.257,
       "step": 2212
     },
     {
       "epoch": 40.24,
-      "learning_rate": 0.00012,
-      "loss": 0.3223,
+      "learning_rate": 0.00026427586206896554,
+      "loss": 0.3712,
       "step": 2213
     },
     {
       "epoch": 40.25,
-      "learning_rate": 0.00011977777777777777,
-      "loss": 0.242,
+      "learning_rate": 0.00026413793103448275,
+      "loss": 0.3013,
       "step": 2214
     },
     {
       "epoch": 40.27,
-      "learning_rate": 0.00011955555555555556,
-      "loss": 0.2412,
+      "learning_rate": 0.000264,
+      "loss": 0.298,
       "step": 2215
     },
     {
       "epoch": 40.29,
-      "learning_rate": 0.00011933333333333334,
-      "loss": 0.2351,
+      "learning_rate": 0.00026386206896551723,
+      "loss": 0.3179,
       "step": 2216
     },
     {
       "epoch": 40.31,
-      "learning_rate": 0.00011911111111111112,
-      "loss": 0.2651,
+      "learning_rate": 0.0002637241379310345,
+      "loss": 0.328,
       "step": 2217
     },
     {
       "epoch": 40.33,
-      "learning_rate": 0.00011888888888888889,
-      "loss": 0.3165,
+      "learning_rate": 0.0002635862068965517,
+      "loss": 0.4012,
       "step": 2218
     },
     {
       "epoch": 40.35,
-      "learning_rate": 0.00011866666666666667,
-      "loss": 0.2434,
+      "learning_rate": 0.000263448275862069,
+      "loss": 0.2635,
       "step": 2219
     },
     {
       "epoch": 40.36,
-      "learning_rate": 0.00011844444444444446,
-      "loss": 0.291,
+      "learning_rate": 0.0002633103448275862,
+      "loss": 0.3499,
       "step": 2220
     },
     {
       "epoch": 40.38,
-      "learning_rate": 0.00011822222222222223,
-      "loss": 0.2508,
+      "learning_rate": 0.00026317241379310345,
+      "loss": 0.3015,
       "step": 2221
     },
     {
       "epoch": 40.4,
-      "learning_rate": 0.000118,
-      "loss": 0.2423,
+      "learning_rate": 0.00026303448275862067,
+      "loss": 0.298,
       "step": 2222
     },
     {
       "epoch": 40.42,
-      "learning_rate": 0.00011777777777777778,
-      "loss": 0.2409,
+      "learning_rate": 0.00026289655172413793,
+      "loss": 0.3406,
       "step": 2223
     },
     {
       "epoch": 40.44,
-      "learning_rate": 0.00011755555555555556,
-      "loss": 0.2423,
+      "learning_rate": 0.00026275862068965514,
+      "loss": 0.306,
       "step": 2224
     },
     {
       "epoch": 40.45,
-      "learning_rate": 0.00011733333333333333,
-      "loss": 0.2937,
+      "learning_rate": 0.00026262068965517246,
+      "loss": 0.3924,
       "step": 2225
     },
     {
       "epoch": 40.47,
-      "learning_rate": 0.00011711111111111111,
-      "loss": 0.2554,
+      "learning_rate": 0.0002624827586206897,
+      "loss": 0.3452,
       "step": 2226
     },
     {
       "epoch": 40.49,
-      "learning_rate": 0.0001168888888888889,
-      "loss": 0.2403,
+      "learning_rate": 0.00026234482758620694,
+      "loss": 0.3538,
       "step": 2227
     },
     {
       "epoch": 40.51,
-      "learning_rate": 0.00011666666666666667,
-      "loss": 0.2018,
+      "learning_rate": 0.00026220689655172415,
+      "loss": 0.2749,
       "step": 2228
     },
     {
       "epoch": 40.53,
-      "learning_rate": 0.00011644444444444445,
-      "loss": 0.2503,
+      "learning_rate": 0.0002620689655172414,
+      "loss": 0.3473,
       "step": 2229
     },
     {
       "epoch": 40.55,
-      "learning_rate": 0.00011622222222222223,
-      "loss": 0.2389,
+      "learning_rate": 0.00026193103448275863,
+      "loss": 0.3387,
       "step": 2230
     },
     {
       "epoch": 40.56,
-      "learning_rate": 0.00011600000000000001,
-      "loss": 0.3546,
+      "learning_rate": 0.0002617931034482759,
+      "loss": 0.3935,
       "step": 2231
     },
     {
       "epoch": 40.58,
-      "learning_rate": 0.00011577777777777777,
-      "loss": 0.2404,
+      "learning_rate": 0.0002616551724137931,
+      "loss": 0.3314,
       "step": 2232
     },
     {
       "epoch": 40.6,
-      "learning_rate": 0.00011555555555555555,
-      "loss": 0.3226,
+      "learning_rate": 0.0002615172413793104,
+      "loss": 0.3927,
       "step": 2233
     },
     {
       "epoch": 40.62,
-      "learning_rate": 0.00011533333333333334,
-      "loss": 0.2367,
+      "learning_rate": 0.0002613793103448276,
+      "loss": 0.3136,
       "step": 2234
     },
     {
       "epoch": 40.64,
-      "learning_rate": 0.0001151111111111111,
-      "loss": 0.2978,
+      "learning_rate": 0.0002612413793103448,
+      "loss": 0.3816,
       "step": 2235
     },
     {
       "epoch": 40.65,
-      "learning_rate": 0.00011488888888888889,
-      "loss": 0.2684,
+      "learning_rate": 0.00026110344827586207,
+      "loss": 0.3853,
       "step": 2236
     },
     {
       "epoch": 40.67,
-      "learning_rate": 0.00011466666666666667,
-      "loss": 0.2739,
+      "learning_rate": 0.0002609655172413793,
+      "loss": 0.3705,
       "step": 2237
     },
     {
       "epoch": 40.69,
-      "learning_rate": 0.00011444444444444445,
-      "loss": 0.2965,
+      "learning_rate": 0.00026082758620689654,
+      "loss": 0.3738,
       "step": 2238
     },
     {
       "epoch": 40.71,
-      "learning_rate": 0.00011422222222222222,
-      "loss": 0.2645,
+      "learning_rate": 0.00026068965517241376,
+      "loss": 0.3333,
       "step": 2239
     },
     {
       "epoch": 40.73,
-      "learning_rate": 0.000114,
-      "loss": 0.2995,
+      "learning_rate": 0.000260551724137931,
+      "loss": 0.3704,
       "step": 2240
     },
     {
       "epoch": 40.75,
-      "learning_rate": 0.00011377777777777779,
-      "loss": 0.2356,
+      "learning_rate": 0.00026041379310344823,
+      "loss": 0.3428,
       "step": 2241
     },
     {
       "epoch": 40.76,
-      "learning_rate": 0.00011355555555555556,
-      "loss": 0.2389,
+      "learning_rate": 0.00026027586206896555,
+      "loss": 0.2877,
       "step": 2242
     },
     {
       "epoch": 40.78,
-      "learning_rate": 0.00011333333333333333,
-      "loss": 0.2127,
+      "learning_rate": 0.00026013793103448277,
+      "loss": 0.2872,
       "step": 2243
     },
     {
       "epoch": 40.8,
-      "learning_rate": 0.00011311111111111111,
-      "loss": 0.2462,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 0.3277,
       "step": 2244
     },
     {
       "epoch": 40.82,
-      "learning_rate": 0.0001128888888888889,
-      "loss": 0.2919,
+      "learning_rate": 0.00025986206896551724,
+      "loss": 0.39,
       "step": 2245
     },
     {
       "epoch": 40.84,
-      "learning_rate": 0.00011266666666666666,
-      "loss": 0.2728,
+      "learning_rate": 0.0002597241379310345,
+      "loss": 0.3685,
       "step": 2246
     },
     {
       "epoch": 40.85,
-      "learning_rate": 0.00011244444444444445,
-      "loss": 0.2573,
+      "learning_rate": 0.0002595862068965517,
+      "loss": 0.3084,
       "step": 2247
     },
     {
       "epoch": 40.87,
-      "learning_rate": 0.00011222222222222223,
-      "loss": 0.2909,
+      "learning_rate": 0.000259448275862069,
+      "loss": 0.3688,
       "step": 2248
     },
     {
       "epoch": 40.89,
-      "learning_rate": 0.000112,
-      "loss": 0.2984,
+      "learning_rate": 0.0002593103448275862,
+      "loss": 0.4008,
       "step": 2249
     },
     {
       "epoch": 40.91,
-      "learning_rate": 0.00011177777777777778,
-      "loss": 0.2195,
+      "learning_rate": 0.00025917241379310347,
+      "loss": 0.2794,
       "step": 2250
     },
     {
       "epoch": 40.93,
-      "learning_rate": 0.00011155555555555556,
-      "loss": 0.2941,
+      "learning_rate": 0.0002590344827586207,
+      "loss": 0.3737,
       "step": 2251
     },
     {
       "epoch": 40.95,
-      "learning_rate": 0.00011133333333333335,
-      "loss": 0.3084,
+      "learning_rate": 0.00025889655172413794,
+      "loss": 0.4072,
       "step": 2252
     },
     {
       "epoch": 40.96,
-      "learning_rate": 0.0001111111111111111,
-      "loss": 0.3348,
+      "learning_rate": 0.00025875862068965515,
+      "loss": 0.3938,
       "step": 2253
     },
     {
       "epoch": 40.98,
-      "learning_rate": 0.00011088888888888889,
-      "loss": 0.2479,
+      "learning_rate": 0.0002586206896551724,
+      "loss": 0.3495,
       "step": 2254
     },
     {
       "epoch": 41.0,
-      "learning_rate": 0.00011066666666666667,
-      "loss": 0.1862,
+      "learning_rate": 0.00025848275862068963,
+      "loss": 0.2464,
       "step": 2255
     },
     {
       "epoch": 41.02,
-      "learning_rate": 0.00011044444444444444,
-      "loss": 0.2558,
+      "learning_rate": 0.0002583448275862069,
+      "loss": 0.3235,
       "step": 2256
     },
     {
       "epoch": 41.04,
-      "learning_rate": 0.00011022222222222222,
-      "loss": 0.261,
+      "learning_rate": 0.0002582068965517241,
+      "loss": 0.3445,
       "step": 2257
     },
     {
       "epoch": 41.05,
-      "learning_rate": 0.00011,
-      "loss": 0.2822,
+      "learning_rate": 0.00025806896551724143,
+      "loss": 0.3732,
       "step": 2258
     },
     {
       "epoch": 41.07,
-      "learning_rate": 0.00010977777777777779,
-      "loss": 0.2473,
+      "learning_rate": 0.00025793103448275864,
+      "loss": 0.3327,
       "step": 2259
     },
     {
       "epoch": 41.09,
-      "learning_rate": 0.00010955555555555556,
-      "loss": 0.2223,
+      "learning_rate": 0.0002577931034482759,
+      "loss": 0.304,
       "step": 2260
     },
     {
       "epoch": 41.11,
-      "learning_rate": 0.00010933333333333334,
-      "loss": 0.2131,
+      "learning_rate": 0.0002576551724137931,
+      "loss": 0.311,
       "step": 2261
     },
     {
       "epoch": 41.13,
-      "learning_rate": 0.00010911111111111112,
-      "loss": 0.2365,
+      "learning_rate": 0.0002575172413793104,
+      "loss": 0.2913,
       "step": 2262
     },
     {
       "epoch": 41.15,
-      "learning_rate": 0.00010888888888888888,
-      "loss": 0.2497,
+      "learning_rate": 0.0002573793103448276,
+      "loss": 0.3194,
       "step": 2263
     },
     {
       "epoch": 41.16,
-      "learning_rate": 0.00010866666666666666,
-      "loss": 0.2116,
+      "learning_rate": 0.00025724137931034486,
+      "loss": 0.2619,
       "step": 2264
     },
     {
       "epoch": 41.18,
-      "learning_rate": 0.00010844444444444444,
-      "loss": 0.2519,
+      "learning_rate": 0.0002571034482758621,
+      "loss": 0.3456,
       "step": 2265
     },
     {
       "epoch": 41.2,
-      "learning_rate": 0.00010822222222222223,
-      "loss": 0.2726,
+      "learning_rate": 0.00025696551724137934,
+      "loss": 0.3525,
       "step": 2266
     },
     {
       "epoch": 41.22,
-      "learning_rate": 0.000108,
-      "loss": 0.23,
+      "learning_rate": 0.00025682758620689655,
+      "loss": 0.2736,
       "step": 2267
     },
     {
       "epoch": 41.24,
-      "learning_rate": 0.00010777777777777778,
-      "loss": 0.2279,
+      "learning_rate": 0.00025668965517241377,
+      "loss": 0.2759,
       "step": 2268
     },
     {
       "epoch": 41.25,
-      "learning_rate": 0.00010755555555555556,
-      "loss": 0.2933,
+      "learning_rate": 0.00025655172413793103,
+      "loss": 0.3577,
       "step": 2269
     },
     {
       "epoch": 41.27,
-      "learning_rate": 0.00010733333333333334,
-      "loss": 0.2387,
+      "learning_rate": 0.00025641379310344824,
+      "loss": 0.3432,
       "step": 2270
     },
     {
       "epoch": 41.29,
-      "learning_rate": 0.00010711111111111111,
-      "loss": 0.2563,
+      "learning_rate": 0.0002562758620689655,
+      "loss": 0.3525,
       "step": 2271
     },
     {
       "epoch": 41.31,
-      "learning_rate": 0.0001068888888888889,
-      "loss": 0.2709,
+      "learning_rate": 0.0002561379310344827,
+      "loss": 0.349,
       "step": 2272
     },
     {
       "epoch": 41.33,
-      "learning_rate": 0.00010666666666666668,
-      "loss": 0.2022,
+      "learning_rate": 0.000256,
+      "loss": 0.2739,
       "step": 2273
     },
     {
       "epoch": 41.35,
-      "learning_rate": 0.00010644444444444444,
-      "loss": 0.2053,
+      "learning_rate": 0.0002558620689655172,
+      "loss": 0.2942,
       "step": 2274
     },
     {
       "epoch": 41.36,
-      "learning_rate": 0.00010622222222222222,
-      "loss": 0.2268,
+      "learning_rate": 0.0002557241379310345,
+      "loss": 0.3423,
       "step": 2275
     },
     {
       "epoch": 41.38,
-      "learning_rate": 0.000106,
-      "loss": 0.2477,
+      "learning_rate": 0.00025558620689655173,
+      "loss": 0.3478,
       "step": 2276
     },
     {
       "epoch": 41.4,
-      "learning_rate": 0.00010577777777777778,
-      "loss": 0.235,
+      "learning_rate": 0.000255448275862069,
+      "loss": 0.3134,
       "step": 2277
     },
     {
       "epoch": 41.42,
-      "learning_rate": 0.00010555555555555555,
-      "loss": 0.2467,
+      "learning_rate": 0.0002553103448275862,
+      "loss": 0.3014,
       "step": 2278
     },
     {
       "epoch": 41.44,
-      "learning_rate": 0.00010533333333333334,
-      "loss": 0.2138,
+      "learning_rate": 0.0002551724137931035,
+      "loss": 0.2932,
       "step": 2279
     },
     {
       "epoch": 41.45,
-      "learning_rate": 0.00010511111111111112,
-      "loss": 0.2446,
+      "learning_rate": 0.0002550344827586207,
+      "loss": 0.3235,
       "step": 2280
     },
     {
       "epoch": 41.47,
-      "learning_rate": 0.00010488888888888889,
-      "loss": 0.3009,
+      "learning_rate": 0.00025489655172413795,
+      "loss": 0.3667,
       "step": 2281
     },
     {
       "epoch": 41.49,
-      "learning_rate": 0.00010466666666666667,
-      "loss": 0.2683,
+      "learning_rate": 0.00025475862068965517,
+      "loss": 0.3157,
       "step": 2282
     },
     {
       "epoch": 41.51,
-      "learning_rate": 0.00010444444444444445,
-      "loss": 0.2741,
+      "learning_rate": 0.00025462068965517243,
+      "loss": 0.343,
       "step": 2283
     },
     {
       "epoch": 41.53,
-      "learning_rate": 0.00010422222222222222,
-      "loss": 0.266,
+      "learning_rate": 0.00025448275862068964,
+      "loss": 0.3471,
       "step": 2284
     },
     {
       "epoch": 41.55,
-      "learning_rate": 0.000104,
-      "loss": 0.1986,
+      "learning_rate": 0.0002543448275862069,
+      "loss": 0.2731,
       "step": 2285
     },
     {
       "epoch": 41.56,
-      "learning_rate": 0.00010377777777777778,
-      "loss": 0.2173,
+      "learning_rate": 0.0002542068965517241,
+      "loss": 0.3464,
       "step": 2286
     },
     {
       "epoch": 41.58,
-      "learning_rate": 0.00010355555555555556,
-      "loss": 0.2809,
+      "learning_rate": 0.0002540689655172414,
+      "loss": 0.3644,
       "step": 2287
     },
     {
       "epoch": 41.6,
-      "learning_rate": 0.00010333333333333333,
-      "loss": 0.2245,
+      "learning_rate": 0.0002539310344827586,
+      "loss": 0.3192,
       "step": 2288
     },
     {
       "epoch": 41.62,
-      "learning_rate": 0.00010311111111111111,
-      "loss": 0.2541,
+      "learning_rate": 0.00025379310344827587,
+      "loss": 0.3543,
       "step": 2289
     },
     {
       "epoch": 41.64,
-      "learning_rate": 0.0001028888888888889,
-      "loss": 0.2498,
+      "learning_rate": 0.0002536551724137931,
+      "loss": 0.3618,
       "step": 2290
     },
     {
       "epoch": 41.65,
-      "learning_rate": 0.00010266666666666668,
-      "loss": 0.2495,
+      "learning_rate": 0.00025351724137931034,
+      "loss": 0.3291,
       "step": 2291
     },
     {
       "epoch": 41.67,
-      "learning_rate": 0.00010244444444444445,
-      "loss": 0.2433,
+      "learning_rate": 0.0002533793103448276,
+      "loss": 0.3131,
       "step": 2292
     },
     {
       "epoch": 41.69,
-      "learning_rate": 0.00010222222222222223,
-      "loss": 0.2326,
+      "learning_rate": 0.0002532413793103449,
+      "loss": 0.3121,
       "step": 2293
     },
     {
       "epoch": 41.71,
-      "learning_rate": 0.000102,
-      "loss": 0.246,
+      "learning_rate": 0.0002531034482758621,
+      "loss": 0.2836,
       "step": 2294
     },
     {
       "epoch": 41.73,
-      "learning_rate": 0.00010177777777777777,
-      "loss": 0.2736,
+      "learning_rate": 0.00025296551724137935,
+      "loss": 0.3575,
       "step": 2295
     },
     {
       "epoch": 41.75,
-      "learning_rate": 0.00010155555555555555,
-      "loss": 0.2461,
+      "learning_rate": 0.00025282758620689657,
+      "loss": 0.3414,
       "step": 2296
     },
     {
       "epoch": 41.76,
-      "learning_rate": 0.00010133333333333333,
-      "loss": 0.2713,
+      "learning_rate": 0.00025268965517241383,
+      "loss": 0.3181,
       "step": 2297
     },
     {
       "epoch": 41.78,
-      "learning_rate": 0.00010111111111111112,
-      "loss": 0.2702,
+      "learning_rate": 0.00025255172413793104,
+      "loss": 0.3497,
       "step": 2298
     },
     {
       "epoch": 41.8,
-      "learning_rate": 0.00010088888888888889,
-      "loss": 0.2511,
+      "learning_rate": 0.0002524137931034483,
+      "loss": 0.3293,
       "step": 2299
     },
     {
       "epoch": 41.82,
-      "learning_rate": 0.00010066666666666667,
-      "loss": 0.1909,
+      "learning_rate": 0.0002522758620689655,
+      "loss": 0.2863,
       "step": 2300
     },
     {
       "epoch": 41.84,
-      "learning_rate": 0.00010044444444444445,
-      "loss": 0.2225,
+      "learning_rate": 0.00025213793103448273,
+      "loss": 0.3216,
       "step": 2301
     },
     {
       "epoch": 41.85,
-      "learning_rate": 0.00010022222222222222,
-      "loss": 0.2185,
+      "learning_rate": 0.000252,
+      "loss": 0.3091,
       "step": 2302
     },
     {
       "epoch": 41.87,
-      "learning_rate": 0.0001,
-      "loss": 0.2227,
+      "learning_rate": 0.0002518620689655172,
+      "loss": 0.3809,
       "step": 2303
     },
     {
       "epoch": 41.89,
-      "learning_rate": 9.977777777777779e-05,
-      "loss": 0.2637,
+      "learning_rate": 0.0002517241379310345,
+      "loss": 0.3376,
       "step": 2304
     },
     {
       "epoch": 41.91,
-      "learning_rate": 9.955555555555556e-05,
-      "loss": 0.2078,
+      "learning_rate": 0.0002515862068965517,
+      "loss": 0.2947,
       "step": 2305
     },
     {
       "epoch": 41.93,
-      "learning_rate": 9.933333333333333e-05,
-      "loss": 0.2713,
+      "learning_rate": 0.00025144827586206896,
+      "loss": 0.3845,
       "step": 2306
     },
     {
       "epoch": 41.95,
-      "learning_rate": 9.911111111111111e-05,
-      "loss": 0.2291,
+      "learning_rate": 0.00025131034482758617,
+      "loss": 0.2866,
       "step": 2307
     },
     {
       "epoch": 41.96,
-      "learning_rate": 9.888888888888889e-05,
-      "loss": 0.3386,
+      "learning_rate": 0.00025117241379310343,
+      "loss": 0.3945,
       "step": 2308
     },
     {
       "epoch": 41.98,
-      "learning_rate": 9.866666666666668e-05,
-      "loss": 0.2262,
+      "learning_rate": 0.0002510344827586207,
+      "loss": 0.3209,
       "step": 2309
     },
     {
       "epoch": 42.0,
-      "learning_rate": 9.844444444444444e-05,
-      "loss": 0.1529,
+      "learning_rate": 0.00025089655172413797,
+      "loss": 0.2216,
       "step": 2310
     },
     {
       "epoch": 42.02,
-      "learning_rate": 9.822222222222223e-05,
-      "loss": 0.2427,
+      "learning_rate": 0.0002507586206896552,
+      "loss": 0.2836,
       "step": 2311
     },
     {
       "epoch": 42.04,
-      "learning_rate": 9.800000000000001e-05,
-      "loss": 0.1961,
+      "learning_rate": 0.00025062068965517244,
+      "loss": 0.2647,
       "step": 2312
     },
     {
       "epoch": 42.05,
-      "learning_rate": 9.777777777777778e-05,
-      "loss": 0.1779,
+      "learning_rate": 0.00025048275862068966,
+      "loss": 0.2526,
       "step": 2313
     },
     {
       "epoch": 42.07,
-      "learning_rate": 9.755555555555556e-05,
-      "loss": 0.2038,
+      "learning_rate": 0.0002503448275862069,
+      "loss": 0.2866,
       "step": 2314
     },
     {
       "epoch": 42.09,
-      "learning_rate": 9.733333333333333e-05,
-      "loss": 0.2424,
+      "learning_rate": 0.00025020689655172413,
+      "loss": 0.2988,
       "step": 2315
     },
     {
       "epoch": 42.11,
-      "learning_rate": 9.711111111111111e-05,
-      "loss": 0.2749,
+      "learning_rate": 0.0002500689655172414,
+      "loss": 0.3545,
       "step": 2316
     },
     {
       "epoch": 42.13,
-      "learning_rate": 9.688888888888888e-05,
-      "loss": 0.2464,
+      "learning_rate": 0.0002499310344827586,
+      "loss": 0.3691,
       "step": 2317
     },
     {
       "epoch": 42.15,
-      "learning_rate": 9.666666666666667e-05,
-      "loss": 0.2596,
+      "learning_rate": 0.0002497931034482759,
+      "loss": 0.3265,
       "step": 2318
     },
     {
       "epoch": 42.16,
-      "learning_rate": 9.644444444444445e-05,
-      "loss": 0.2185,
+      "learning_rate": 0.0002496551724137931,
+      "loss": 0.2781,
       "step": 2319
     },
     {
       "epoch": 42.18,
-      "learning_rate": 9.622222222222222e-05,
-      "loss": 0.2341,
+      "learning_rate": 0.00024951724137931036,
+      "loss": 0.3012,
       "step": 2320
     },
     {
       "epoch": 42.2,
-      "learning_rate": 9.6e-05,
-      "loss": 0.2361,
+      "learning_rate": 0.00024937931034482757,
+      "loss": 0.3342,
       "step": 2321
     },
     {
       "epoch": 42.22,
-      "learning_rate": 9.577777777777779e-05,
-      "loss": 0.2362,
+      "learning_rate": 0.00024924137931034483,
+      "loss": 0.3119,
       "step": 2322
     },
     {
       "epoch": 42.24,
-      "learning_rate": 9.555555555555557e-05,
-      "loss": 0.2092,
+      "learning_rate": 0.00024910344827586204,
+      "loss": 0.2848,
       "step": 2323
     },
     {
       "epoch": 42.25,
-      "learning_rate": 9.533333333333334e-05,
-      "loss": 0.3038,
+      "learning_rate": 0.0002489655172413793,
+      "loss": 0.3924,
       "step": 2324
     },
     {
       "epoch": 42.27,
-      "learning_rate": 9.511111111111111e-05,
-      "loss": 0.2303,
+      "learning_rate": 0.0002488275862068966,
+      "loss": 0.3103,
       "step": 2325
     },
     {
       "epoch": 42.29,
-      "learning_rate": 9.488888888888889e-05,
-      "loss": 0.2891,
+      "learning_rate": 0.0002486896551724138,
+      "loss": 0.3916,
       "step": 2326
     },
     {
       "epoch": 42.31,
-      "learning_rate": 9.466666666666666e-05,
-      "loss": 0.2463,
+      "learning_rate": 0.00024855172413793105,
+      "loss": 0.3329,
       "step": 2327
     },
     {
       "epoch": 42.33,
-      "learning_rate": 9.444444444444444e-05,
-      "loss": 0.216,
+      "learning_rate": 0.00024841379310344827,
+      "loss": 0.2826,
       "step": 2328
     },
     {
       "epoch": 42.35,
-      "learning_rate": 9.422222222222223e-05,
-      "loss": 0.2281,
+      "learning_rate": 0.00024827586206896553,
+      "loss": 0.3128,
       "step": 2329
     },
     {
       "epoch": 42.36,
-      "learning_rate": 9.400000000000001e-05,
-      "loss": 0.2799,
+      "learning_rate": 0.00024813793103448274,
+      "loss": 0.3696,
       "step": 2330
     },
     {
       "epoch": 42.38,
-      "learning_rate": 9.377777777777778e-05,
-      "loss": 0.243,
+      "learning_rate": 0.000248,
+      "loss": 0.2855,
       "step": 2331
     },
     {
       "epoch": 42.4,
-      "learning_rate": 9.355555555555556e-05,
-      "loss": 0.2549,
+      "learning_rate": 0.0002478620689655172,
+      "loss": 0.2896,
       "step": 2332
     },
     {
       "epoch": 42.42,
-      "learning_rate": 9.333333333333334e-05,
-      "loss": 0.2589,
+      "learning_rate": 0.0002477241379310345,
+      "loss": 0.3402,
       "step": 2333
     },
     {
       "epoch": 42.44,
-      "learning_rate": 9.311111111111111e-05,
-      "loss": 0.2519,
+      "learning_rate": 0.00024758620689655175,
+      "loss": 0.3281,
       "step": 2334
     },
     {
       "epoch": 42.45,
-      "learning_rate": 9.288888888888888e-05,
-      "loss": 0.2187,
+      "learning_rate": 0.00024744827586206897,
+      "loss": 0.3236,
       "step": 2335
     },
     {
       "epoch": 42.47,
-      "learning_rate": 9.266666666666666e-05,
-      "loss": 0.2575,
+      "learning_rate": 0.00024731034482758623,
+      "loss": 0.331,
       "step": 2336
     },
     {
       "epoch": 42.49,
-      "learning_rate": 9.244444444444445e-05,
-      "loss": 0.2504,
+      "learning_rate": 0.00024717241379310344,
+      "loss": 0.2993,
       "step": 2337
     },
     {
       "epoch": 42.51,
-      "learning_rate": 9.222222222222222e-05,
-      "loss": 0.1973,
+      "learning_rate": 0.0002470344827586207,
+      "loss": 0.2941,
       "step": 2338
     },
     {
       "epoch": 42.53,
-      "learning_rate": 9.2e-05,
-      "loss": 0.2848,
+      "learning_rate": 0.0002468965517241379,
+      "loss": 0.3576,
       "step": 2339
     },
     {
       "epoch": 42.55,
-      "learning_rate": 9.177777777777778e-05,
-      "loss": 0.2194,
+      "learning_rate": 0.0002467586206896552,
+      "loss": 0.2819,
       "step": 2340
     },
     {
       "epoch": 42.56,
-      "learning_rate": 9.155555555555555e-05,
-      "loss": 0.3058,
+      "learning_rate": 0.0002466206896551724,
+      "loss": 0.3482,
       "step": 2341
     },
     {
       "epoch": 42.58,
-      "learning_rate": 9.133333333333334e-05,
-      "loss": 0.2199,
+      "learning_rate": 0.00024648275862068967,
+      "loss": 0.3133,
       "step": 2342
     },
     {
       "epoch": 42.6,
-      "learning_rate": 9.111111111111112e-05,
-      "loss": 0.242,
+      "learning_rate": 0.00024634482758620693,
+      "loss": 0.3526,
       "step": 2343
     },
     {
       "epoch": 42.62,
-      "learning_rate": 9.08888888888889e-05,
-      "loss": 0.1893,
+      "learning_rate": 0.00024620689655172414,
+      "loss": 0.307,
       "step": 2344
     },
     {
       "epoch": 42.64,
-      "learning_rate": 9.066666666666667e-05,
-      "loss": 0.2016,
+      "learning_rate": 0.0002460689655172414,
+      "loss": 0.268,
       "step": 2345
     },
     {
       "epoch": 42.65,
-      "learning_rate": 9.044444444444444e-05,
-      "loss": 0.2357,
+      "learning_rate": 0.0002459310344827586,
+      "loss": 0.2894,
       "step": 2346
     },
     {
       "epoch": 42.67,
-      "learning_rate": 9.022222222222222e-05,
-      "loss": 0.2705,
+      "learning_rate": 0.0002457931034482759,
+      "loss": 0.3661,
       "step": 2347
     },
     {
       "epoch": 42.69,
-      "learning_rate": 8.999999999999999e-05,
-      "loss": 0.2184,
+      "learning_rate": 0.0002456551724137931,
+      "loss": 0.2954,
       "step": 2348
     },
     {
       "epoch": 42.71,
-      "learning_rate": 8.977777777777778e-05,
-      "loss": 0.214,
+      "learning_rate": 0.00024551724137931037,
+      "loss": 0.313,
       "step": 2349
     },
     {
       "epoch": 42.73,
-      "learning_rate": 8.955555555555556e-05,
-      "loss": 0.2008,
+      "learning_rate": 0.0002453793103448276,
+      "loss": 0.2683,
       "step": 2350
     },
     {
       "epoch": 42.75,
-      "learning_rate": 8.933333333333334e-05,
-      "loss": 0.305,
+      "learning_rate": 0.00024524137931034484,
+      "loss": 0.3919,
       "step": 2351
     },
     {
       "epoch": 42.76,
-      "learning_rate": 8.911111111111111e-05,
-      "loss": 0.255,
+      "learning_rate": 0.00024510344827586206,
+      "loss": 0.3252,
       "step": 2352
     },
     {
       "epoch": 42.78,
-      "learning_rate": 8.888888888888889e-05,
-      "loss": 0.2376,
+      "learning_rate": 0.0002449655172413793,
+      "loss": 0.305,
       "step": 2353
     },
     {
       "epoch": 42.8,
-      "learning_rate": 8.866666666666668e-05,
-      "loss": 0.2244,
+      "learning_rate": 0.00024482758620689653,
+      "loss": 0.2699,
       "step": 2354
     },
     {
       "epoch": 42.82,
-      "learning_rate": 8.844444444444446e-05,
-      "loss": 0.2967,
+      "learning_rate": 0.0002446896551724138,
+      "loss": 0.3744,
       "step": 2355
     },
     {
       "epoch": 42.84,
-      "learning_rate": 8.822222222222221e-05,
-      "loss": 0.2118,
+      "learning_rate": 0.000244551724137931,
+      "loss": 0.301,
       "step": 2356
     },
     {
       "epoch": 42.85,
-      "learning_rate": 8.8e-05,
-      "loss": 0.2566,
+      "learning_rate": 0.0002444137931034483,
+      "loss": 0.3543,
       "step": 2357
     },
     {
       "epoch": 42.87,
-      "learning_rate": 8.777777777777778e-05,
-      "loss": 0.2286,
+      "learning_rate": 0.0002442758620689655,
+      "loss": 0.3074,
       "step": 2358
     },
     {
       "epoch": 42.89,
-      "learning_rate": 8.755555555555555e-05,
-      "loss": 0.2432,
+      "learning_rate": 0.00024413793103448276,
+      "loss": 0.3387,
       "step": 2359
     },
     {
       "epoch": 42.91,
-      "learning_rate": 8.733333333333333e-05,
-      "loss": 0.1968,
+      "learning_rate": 0.000244,
+      "loss": 0.2062,
       "step": 2360
     },
     {
       "epoch": 42.93,
-      "learning_rate": 8.711111111111112e-05,
-      "loss": 0.2033,
+      "learning_rate": 0.00024386206896551723,
+      "loss": 0.2782,
       "step": 2361
     },
     {
       "epoch": 42.95,
-      "learning_rate": 8.68888888888889e-05,
-      "loss": 0.2066,
+      "learning_rate": 0.00024372413793103447,
+      "loss": 0.2612,
       "step": 2362
     },
     {
       "epoch": 42.96,
-      "learning_rate": 8.666666666666667e-05,
-      "loss": 0.2226,
+      "learning_rate": 0.00024358620689655174,
+      "loss": 0.3078,
       "step": 2363
     },
     {
       "epoch": 42.98,
-      "learning_rate": 8.644444444444445e-05,
-      "loss": 0.2957,
+      "learning_rate": 0.00024344827586206898,
+      "loss": 0.3665,
       "step": 2364
     },
     {
       "epoch": 43.0,
-      "learning_rate": 8.644444444444445e-05,
-      "loss": 0.3031,
+      "learning_rate": 0.00024331034482758622,
+      "loss": 0.3427,
       "step": 2365
     },
     {
       "epoch": 43.02,
-      "learning_rate": 8.622222222222223e-05,
-      "loss": 0.2482,
+      "learning_rate": 0.00024317241379310346,
+      "loss": 0.3397,
       "step": 2366
     },
     {
       "epoch": 43.04,
-      "learning_rate": 8.599999999999999e-05,
-      "loss": 0.284,
+      "learning_rate": 0.0002430344827586207,
+      "loss": 0.3467,
       "step": 2367
     },
     {
       "epoch": 43.05,
-      "learning_rate": 8.577777777777777e-05,
-      "loss": 0.2182,
+      "learning_rate": 0.00024289655172413793,
+      "loss": 0.2799,
       "step": 2368
     },
     {
       "epoch": 43.07,
-      "learning_rate": 8.555555555555556e-05,
-      "loss": 0.2275,
+      "learning_rate": 0.00024275862068965517,
+      "loss": 0.311,
       "step": 2369
     },
     {
       "epoch": 43.09,
-      "learning_rate": 8.533333333333334e-05,
-      "loss": 0.2518,
+      "learning_rate": 0.0002426206896551724,
+      "loss": 0.3043,
       "step": 2370
     },
     {
       "epoch": 43.11,
-      "learning_rate": 8.511111111111111e-05,
-      "loss": 0.2164,
+      "learning_rate": 0.00024248275862068965,
+      "loss": 0.2958,
       "step": 2371
     },
     {
       "epoch": 43.13,
-      "learning_rate": 8.488888888888889e-05,
-      "loss": 0.3022,
+      "learning_rate": 0.00024234482758620692,
+      "loss": 0.3873,
       "step": 2372
     },
     {
       "epoch": 43.15,
-      "learning_rate": 8.466666666666667e-05,
-      "loss": 0.2312,
+      "learning_rate": 0.00024220689655172416,
+      "loss": 0.2829,
       "step": 2373
     },
     {
       "epoch": 43.16,
-      "learning_rate": 8.444444444444444e-05,
-      "loss": 0.1996,
+      "learning_rate": 0.0002420689655172414,
+      "loss": 0.2838,
       "step": 2374
     },
     {
       "epoch": 43.18,
-      "learning_rate": 8.422222222222223e-05,
-      "loss": 0.276,
+      "learning_rate": 0.00024193103448275863,
+      "loss": 0.3395,
       "step": 2375
     },
     {
       "epoch": 43.2,
-      "learning_rate": 8.400000000000001e-05,
-      "loss": 0.2593,
+      "learning_rate": 0.00024179310344827587,
+      "loss": 0.3155,
       "step": 2376
     },
     {
       "epoch": 43.22,
-      "learning_rate": 8.377777777777779e-05,
-      "loss": 0.2574,
+      "learning_rate": 0.0002416551724137931,
+      "loss": 0.327,
       "step": 2377
     },
     {
       "epoch": 43.24,
-      "learning_rate": 8.355555555555555e-05,
-      "loss": 0.191,
+      "learning_rate": 0.00024151724137931035,
+      "loss": 0.2862,
       "step": 2378
     },
     {
       "epoch": 43.25,
-      "learning_rate": 8.333333333333333e-05,
-      "loss": 0.2095,
+      "learning_rate": 0.0002413793103448276,
+      "loss": 0.3052,
       "step": 2379
     },
     {
       "epoch": 43.27,
-      "learning_rate": 8.311111111111111e-05,
-      "loss": 0.2586,
+      "learning_rate": 0.00024124137931034486,
+      "loss": 0.3379,
       "step": 2380
     },
     {
       "epoch": 43.29,
-      "learning_rate": 8.288888888888888e-05,
-      "loss": 0.2167,
+      "learning_rate": 0.0002411034482758621,
+      "loss": 0.2865,
       "step": 2381
     },
     {
       "epoch": 43.31,
-      "learning_rate": 8.266666666666667e-05,
-      "loss": 0.2292,
+      "learning_rate": 0.00024096551724137933,
+      "loss": 0.2832,
       "step": 2382
     },
     {
       "epoch": 43.33,
-      "learning_rate": 8.244444444444445e-05,
-      "loss": 0.2228,
+      "learning_rate": 0.00024082758620689657,
+      "loss": 0.315,
       "step": 2383
     },
     {
       "epoch": 43.35,
-      "learning_rate": 8.222222222222223e-05,
-      "loss": 0.2062,
+      "learning_rate": 0.00024068965517241378,
+      "loss": 0.2995,
       "step": 2384
     },
     {
       "epoch": 43.36,
-      "learning_rate": 8.2e-05,
-      "loss": 0.2627,
+      "learning_rate": 0.00024055172413793102,
+      "loss": 0.3447,
       "step": 2385
     },
     {
       "epoch": 43.38,
-      "learning_rate": 8.177777777777778e-05,
-      "loss": 0.2735,
+      "learning_rate": 0.00024041379310344826,
+      "loss": 0.3024,
       "step": 2386
     },
     {
       "epoch": 43.4,
-      "learning_rate": 8.155555555555557e-05,
-      "loss": 0.2299,
+      "learning_rate": 0.0002402758620689655,
+      "loss": 0.326,
       "step": 2387
     },
     {
       "epoch": 43.42,
-      "learning_rate": 8.133333333333332e-05,
-      "loss": 0.2058,
+      "learning_rate": 0.00024013793103448277,
+      "loss": 0.2872,
       "step": 2388
     },
     {
       "epoch": 43.44,
-      "learning_rate": 8.11111111111111e-05,
-      "loss": 0.2672,
+      "learning_rate": 0.00024,
+      "loss": 0.374,
       "step": 2389
     },
     {
       "epoch": 43.45,
-      "learning_rate": 8.088888888888889e-05,
-      "loss": 0.2624,
+      "learning_rate": 0.00023986206896551725,
+      "loss": 0.3168,
       "step": 2390
     },
     {
       "epoch": 43.47,
-      "learning_rate": 8.066666666666667e-05,
-      "loss": 0.2114,
+      "learning_rate": 0.00023972413793103448,
+      "loss": 0.2892,
       "step": 2391
     },
     {
       "epoch": 43.49,
-      "learning_rate": 8.044444444444444e-05,
-      "loss": 0.2349,
+      "learning_rate": 0.00023958620689655172,
+      "loss": 0.3302,
       "step": 2392
     },
     {
       "epoch": 43.51,
-      "learning_rate": 8.022222222222222e-05,
-      "loss": 0.2277,
+      "learning_rate": 0.00023944827586206896,
+      "loss": 0.3214,
       "step": 2393
     },
     {
       "epoch": 43.53,
-      "learning_rate": 8e-05,
-      "loss": 0.2227,
+      "learning_rate": 0.0002393103448275862,
+      "loss": 0.3025,
       "step": 2394
     },
     {
       "epoch": 43.55,
-      "learning_rate": 7.977777777777778e-05,
-      "loss": 0.2196,
+      "learning_rate": 0.00023917241379310344,
+      "loss": 0.3276,
       "step": 2395
     },
     {
       "epoch": 43.56,
-      "learning_rate": 7.955555555555556e-05,
-      "loss": 0.2627,
+      "learning_rate": 0.00023903448275862068,
+      "loss": 0.3332,
       "step": 2396
     },
     {
       "epoch": 43.58,
-      "learning_rate": 7.933333333333334e-05,
-      "loss": 0.209,
+      "learning_rate": 0.00023889655172413794,
+      "loss": 0.2907,
       "step": 2397
     },
     {
       "epoch": 43.6,
-      "learning_rate": 7.911111111111111e-05,
-      "loss": 0.1914,
+      "learning_rate": 0.00023875862068965518,
+      "loss": 0.3115,
       "step": 2398
     },
     {
       "epoch": 43.62,
-      "learning_rate": 7.888888888888888e-05,
-      "loss": 0.224,
+      "learning_rate": 0.00023862068965517242,
+      "loss": 0.306,
       "step": 2399
     },
     {
       "epoch": 43.64,
-      "learning_rate": 7.866666666666666e-05,
-      "loss": 0.2475,
+      "learning_rate": 0.00023848275862068966,
+      "loss": 0.3239,
       "step": 2400
     },
     {
       "epoch": 43.65,
-      "learning_rate": 7.844444444444445e-05,
-      "loss": 0.3144,
+      "learning_rate": 0.0002383448275862069,
+      "loss": 0.4017,
       "step": 2401
     },
     {
       "epoch": 43.67,
-      "learning_rate": 7.822222222222222e-05,
-      "loss": 0.1941,
+      "learning_rate": 0.00023820689655172414,
+      "loss": 0.2601,
       "step": 2402
     },
     {
       "epoch": 43.69,
-      "learning_rate": 7.8e-05,
-      "loss": 0.2593,
+      "learning_rate": 0.00023806896551724138,
+      "loss": 0.3455,
       "step": 2403
     },
     {
       "epoch": 43.71,
-      "learning_rate": 7.777777777777778e-05,
-      "loss": 0.2541,
+      "learning_rate": 0.00023793103448275862,
+      "loss": 0.3242,
       "step": 2404
     },
     {
       "epoch": 43.73,
-      "learning_rate": 7.755555555555556e-05,
-      "loss": 0.2656,
+      "learning_rate": 0.00023779310344827588,
+      "loss": 0.3403,
       "step": 2405
     },
     {
       "epoch": 43.75,
-      "learning_rate": 7.733333333333333e-05,
-      "loss": 0.2356,
+      "learning_rate": 0.00023765517241379312,
+      "loss": 0.3006,
       "step": 2406
     },
     {
       "epoch": 43.76,
-      "learning_rate": 7.711111111111112e-05,
-      "loss": 0.2436,
+      "learning_rate": 0.00023751724137931036,
+      "loss": 0.3397,
       "step": 2407
     },
     {
       "epoch": 43.78,
-      "learning_rate": 7.68888888888889e-05,
-      "loss": 0.2094,
+      "learning_rate": 0.0002373793103448276,
+      "loss": 0.2921,
       "step": 2408
     },
     {
       "epoch": 43.8,
-      "learning_rate": 7.666666666666667e-05,
-      "loss": 0.229,
+      "learning_rate": 0.00023724137931034484,
+      "loss": 0.2971,
       "step": 2409
     },
     {
       "epoch": 43.82,
-      "learning_rate": 7.644444444444444e-05,
-      "loss": 0.1859,
+      "learning_rate": 0.00023710344827586208,
+      "loss": 0.2851,
       "step": 2410
     },
     {
       "epoch": 43.84,
-      "learning_rate": 7.622222222222222e-05,
-      "loss": 0.2757,
+      "learning_rate": 0.00023696551724137932,
+      "loss": 0.3855,
       "step": 2411
     },
     {
       "epoch": 43.85,
-      "learning_rate": 7.6e-05,
-      "loss": 0.196,
+      "learning_rate": 0.00023682758620689656,
+      "loss": 0.2632,
       "step": 2412
     },
     {
       "epoch": 43.87,
-      "learning_rate": 7.577777777777777e-05,
-      "loss": 0.2877,
+      "learning_rate": 0.00023668965517241382,
+      "loss": 0.3579,
       "step": 2413
     },
     {
       "epoch": 43.89,
-      "learning_rate": 7.555555555555556e-05,
-      "loss": 0.2117,
+      "learning_rate": 0.00023655172413793106,
+      "loss": 0.2913,
       "step": 2414
     },
     {
       "epoch": 43.91,
-      "learning_rate": 7.533333333333334e-05,
-      "loss": 0.1856,
+      "learning_rate": 0.0002364137931034483,
+      "loss": 0.248,
       "step": 2415
     },
     {
       "epoch": 43.93,
-      "learning_rate": 7.511111111111112e-05,
-      "loss": 0.2431,
+      "learning_rate": 0.00023627586206896554,
+      "loss": 0.3521,
       "step": 2416
     },
     {
       "epoch": 43.95,
-      "learning_rate": 7.488888888888889e-05,
-      "loss": 0.2394,
+      "learning_rate": 0.00023613793103448275,
+      "loss": 0.3122,
       "step": 2417
     },
     {
       "epoch": 43.96,
-      "learning_rate": 7.466666666666667e-05,
-      "loss": 0.2527,
+      "learning_rate": 0.000236,
+      "loss": 0.3391,
       "step": 2418
     },
     {
       "epoch": 43.98,
-      "learning_rate": 7.444444444444444e-05,
-      "loss": 0.2234,
+      "learning_rate": 0.00023586206896551723,
+      "loss": 0.3128,
       "step": 2419
     },
     {
       "epoch": 44.0,
-      "learning_rate": 7.422222222222221e-05,
-      "loss": 0.2095,
+      "learning_rate": 0.00023572413793103447,
+      "loss": 0.2354,
       "step": 2420
     },
     {
       "epoch": 44.02,
-      "learning_rate": 7.4e-05,
-      "loss": 0.2639,
+      "learning_rate": 0.0002355862068965517,
+      "loss": 0.3251,
       "step": 2421
     },
     {
       "epoch": 44.04,
-      "learning_rate": 7.377777777777778e-05,
-      "loss": 0.2353,
+      "learning_rate": 0.00023544827586206897,
+      "loss": 0.3195,
       "step": 2422
     },
     {
       "epoch": 44.05,
-      "learning_rate": 7.355555555555556e-05,
-      "loss": 0.2006,
+      "learning_rate": 0.0002353103448275862,
+      "loss": 0.2948,
       "step": 2423
     },
     {
       "epoch": 44.07,
-      "learning_rate": 7.333333333333333e-05,
-      "loss": 0.2032,
+      "learning_rate": 0.00023517241379310345,
+      "loss": 0.2865,
       "step": 2424
     },
     {
       "epoch": 44.09,
-      "learning_rate": 7.311111111111111e-05,
-      "loss": 0.2522,
+      "learning_rate": 0.0002350344827586207,
+      "loss": 0.3495,
       "step": 2425
     },
     {
       "epoch": 44.11,
-      "learning_rate": 7.28888888888889e-05,
-      "loss": 0.2493,
+      "learning_rate": 0.00023489655172413793,
+      "loss": 0.3201,
       "step": 2426
     },
     {
       "epoch": 44.13,
-      "learning_rate": 7.266666666666667e-05,
-      "loss": 0.2537,
+      "learning_rate": 0.00023475862068965517,
+      "loss": 0.2744,
       "step": 2427
     },
     {
       "epoch": 44.15,
-      "learning_rate": 7.244444444444445e-05,
-      "loss": 0.2437,
+      "learning_rate": 0.0002346206896551724,
+      "loss": 0.3102,
       "step": 2428
     },
     {
       "epoch": 44.16,
-      "learning_rate": 7.222222222222222e-05,
-      "loss": 0.2002,
+      "learning_rate": 0.00023448275862068965,
+      "loss": 0.2765,
       "step": 2429
     },
     {
       "epoch": 44.18,
-      "learning_rate": 7.2e-05,
-      "loss": 0.2837,
+      "learning_rate": 0.0002343448275862069,
+      "loss": 0.3436,
       "step": 2430
     },
     {
       "epoch": 44.2,
-      "learning_rate": 7.177777777777777e-05,
-      "loss": 0.2133,
+      "learning_rate": 0.00023420689655172415,
+      "loss": 0.2946,
       "step": 2431
     },
     {
       "epoch": 44.22,
-      "learning_rate": 7.155555555555555e-05,
-      "loss": 0.209,
+      "learning_rate": 0.0002340689655172414,
+      "loss": 0.3106,
       "step": 2432
     },
     {
       "epoch": 44.24,
-      "learning_rate": 7.133333333333334e-05,
-      "loss": 0.2328,
+      "learning_rate": 0.00023393103448275863,
+      "loss": 0.313,
       "step": 2433
     },
     {
       "epoch": 44.25,
-      "learning_rate": 7.11111111111111e-05,
-      "loss": 0.2419,
+      "learning_rate": 0.00023379310344827587,
+      "loss": 0.2893,
       "step": 2434
     },
     {
       "epoch": 44.27,
-      "learning_rate": 7.088888888888889e-05,
-      "loss": 0.2505,
+      "learning_rate": 0.0002336551724137931,
+      "loss": 0.3558,
       "step": 2435
     },
     {
       "epoch": 44.29,
-      "learning_rate": 7.066666666666667e-05,
-      "loss": 0.2289,
+      "learning_rate": 0.00023351724137931035,
+      "loss": 0.3059,
       "step": 2436
     },
     {
       "epoch": 44.31,
-      "learning_rate": 7.044444444444446e-05,
-      "loss": 0.2088,
+      "learning_rate": 0.00023337931034482758,
+      "loss": 0.2812,
       "step": 2437
     },
     {
       "epoch": 44.33,
-      "learning_rate": 7.022222222222222e-05,
-      "loss": 0.228,
+      "learning_rate": 0.00023324137931034482,
+      "loss": 0.31,
       "step": 2438
     },
     {
       "epoch": 44.35,
-      "learning_rate": 7.000000000000001e-05,
-      "loss": 0.2078,
+      "learning_rate": 0.0002331034482758621,
+      "loss": 0.3107,
       "step": 2439
     },
     {
       "epoch": 44.36,
-      "learning_rate": 6.977777777777778e-05,
-      "loss": 0.2495,
+      "learning_rate": 0.00023296551724137933,
+      "loss": 0.3107,
       "step": 2440
     },
     {
       "epoch": 44.38,
-      "learning_rate": 6.955555555555555e-05,
-      "loss": 0.2629,
+      "learning_rate": 0.00023282758620689657,
+      "loss": 0.3131,
       "step": 2441
     },
     {
       "epoch": 44.4,
-      "learning_rate": 6.933333333333333e-05,
-      "loss": 0.281,
+      "learning_rate": 0.0002326896551724138,
+      "loss": 0.3374,
       "step": 2442
     },
     {
       "epoch": 44.42,
-      "learning_rate": 6.911111111111111e-05,
-      "loss": 0.2469,
+      "learning_rate": 0.00023255172413793105,
+      "loss": 0.3488,
       "step": 2443
     },
     {
       "epoch": 44.44,
-      "learning_rate": 6.88888888888889e-05,
-      "loss": 0.2425,
+      "learning_rate": 0.00023241379310344828,
+      "loss": 0.3309,
       "step": 2444
     },
     {
       "epoch": 44.45,
-      "learning_rate": 6.866666666666666e-05,
-      "loss": 0.1485,
+      "learning_rate": 0.00023227586206896552,
+      "loss": 0.2559,
       "step": 2445
     },
     {
       "epoch": 44.47,
-      "learning_rate": 6.844444444444445e-05,
-      "loss": 0.2521,
+      "learning_rate": 0.00023213793103448276,
+      "loss": 0.3561,
       "step": 2446
     },
     {
       "epoch": 44.49,
-      "learning_rate": 6.822222222222223e-05,
-      "loss": 0.2644,
+      "learning_rate": 0.00023200000000000003,
+      "loss": 0.3349,
       "step": 2447
     },
     {
       "epoch": 44.51,
-      "learning_rate": 6.800000000000001e-05,
-      "loss": 0.2039,
+      "learning_rate": 0.00023186206896551727,
+      "loss": 0.2638,
       "step": 2448
     },
     {
       "epoch": 44.53,
-      "learning_rate": 6.777777777777778e-05,
-      "loss": 0.1918,
+      "learning_rate": 0.0002317241379310345,
+      "loss": 0.309,
       "step": 2449
     },
     {
       "epoch": 44.55,
-      "learning_rate": 6.755555555555555e-05,
-      "loss": 0.233,
+      "learning_rate": 0.00023158620689655172,
+      "loss": 0.2613,
       "step": 2450
     },
     {
       "epoch": 44.56,
-      "learning_rate": 6.733333333333333e-05,
-      "loss": 0.263,
+      "learning_rate": 0.00023144827586206896,
+      "loss": 0.3059,
       "step": 2451
     },
     {
       "epoch": 44.58,
-      "learning_rate": 6.71111111111111e-05,
-      "loss": 0.2723,
+      "learning_rate": 0.0002313103448275862,
+      "loss": 0.2979,
       "step": 2452
     },
     {
       "epoch": 44.6,
-      "learning_rate": 6.688888888888889e-05,
-      "loss": 0.2581,
+      "learning_rate": 0.00023117241379310344,
+      "loss": 0.3323,
       "step": 2453
     },
     {
       "epoch": 44.62,
-      "learning_rate": 6.666666666666667e-05,
-      "loss": 0.2267,
+      "learning_rate": 0.00023103448275862067,
+      "loss": 0.2712,
       "step": 2454
     },
     {
       "epoch": 44.64,
-      "learning_rate": 6.644444444444445e-05,
-      "loss": 0.2061,
+      "learning_rate": 0.0002308965517241379,
+      "loss": 0.3008,
       "step": 2455
     },
     {
       "epoch": 44.65,
-      "learning_rate": 6.622222222222222e-05,
-      "loss": 0.2351,
+      "learning_rate": 0.00023075862068965518,
+      "loss": 0.2814,
       "step": 2456
     },
     {
       "epoch": 44.67,
-      "learning_rate": 6.6e-05,
-      "loss": 0.2081,
+      "learning_rate": 0.00023062068965517242,
+      "loss": 0.2883,
       "step": 2457
     },
     {
       "epoch": 44.69,
-      "learning_rate": 6.577777777777779e-05,
-      "loss": 0.2198,
+      "learning_rate": 0.00023048275862068966,
+      "loss": 0.2784,
       "step": 2458
     },
     {
       "epoch": 44.71,
-      "learning_rate": 6.555555555555556e-05,
-      "loss": 0.2416,
+      "learning_rate": 0.0002303448275862069,
+      "loss": 0.3248,
       "step": 2459
     },
     {
       "epoch": 44.73,
-      "learning_rate": 6.533333333333333e-05,
-      "loss": 0.2174,
+      "learning_rate": 0.00023020689655172413,
+      "loss": 0.2602,
       "step": 2460
     },
     {
       "epoch": 44.75,
-      "learning_rate": 6.511111111111111e-05,
-      "loss": 0.2561,
+      "learning_rate": 0.00023006896551724137,
+      "loss": 0.3074,
       "step": 2461
     },
     {
       "epoch": 44.76,
-      "learning_rate": 6.488888888888889e-05,
-      "loss": 0.1753,
+      "learning_rate": 0.0002299310344827586,
+      "loss": 0.2466,
       "step": 2462
     },
     {
       "epoch": 44.78,
-      "learning_rate": 6.466666666666666e-05,
-      "loss": 0.2171,
+      "learning_rate": 0.00022979310344827585,
+      "loss": 0.3059,
       "step": 2463
     },
     {
       "epoch": 44.8,
-      "learning_rate": 6.444444444444444e-05,
-      "loss": 0.24,
+      "learning_rate": 0.00022965517241379312,
+      "loss": 0.2944,
       "step": 2464
     },
     {
       "epoch": 44.82,
-      "learning_rate": 6.422222222222223e-05,
-      "loss": 0.2511,
+      "learning_rate": 0.00022951724137931036,
+      "loss": 0.3118,
       "step": 2465
     },
     {
       "epoch": 44.84,
-      "learning_rate": 6.4e-05,
-      "loss": 0.2272,
+      "learning_rate": 0.0002293793103448276,
+      "loss": 0.3069,
       "step": 2466
     },
     {
       "epoch": 44.85,
-      "learning_rate": 6.377777777777778e-05,
-      "loss": 0.2189,
+      "learning_rate": 0.00022924137931034483,
+      "loss": 0.2654,
       "step": 2467
     },
     {
       "epoch": 44.87,
-      "learning_rate": 6.355555555555556e-05,
-      "loss": 0.1887,
+      "learning_rate": 0.00022910344827586207,
+      "loss": 0.2955,
       "step": 2468
     },
     {
       "epoch": 44.89,
-      "learning_rate": 6.333333333333335e-05,
-      "loss": 0.2351,
+      "learning_rate": 0.0002289655172413793,
+      "loss": 0.3426,
       "step": 2469
     },
     {
       "epoch": 44.91,
-      "learning_rate": 6.311111111111112e-05,
-      "loss": 0.181,
+      "learning_rate": 0.00022882758620689655,
+      "loss": 0.2456,
       "step": 2470
     },
     {
       "epoch": 44.93,
-      "learning_rate": 6.288888888888888e-05,
-      "loss": 0.2867,
+      "learning_rate": 0.0002286896551724138,
+      "loss": 0.3907,
       "step": 2471
     },
     {
       "epoch": 44.95,
-      "learning_rate": 6.266666666666667e-05,
-      "loss": 0.2065,
+      "learning_rate": 0.00022855172413793106,
+      "loss": 0.2712,
       "step": 2472
     },
     {
       "epoch": 44.96,
-      "learning_rate": 6.244444444444444e-05,
-      "loss": 0.2172,
+      "learning_rate": 0.0002284137931034483,
+      "loss": 0.2995,
       "step": 2473
     },
     {
       "epoch": 44.98,
-      "learning_rate": 6.222222222222222e-05,
-      "loss": 0.1947,
+      "learning_rate": 0.00022827586206896553,
+      "loss": 0.2745,
       "step": 2474
     },
     {
       "epoch": 45.0,
-      "learning_rate": 6.2e-05,
-      "loss": 0.1765,
+      "learning_rate": 0.00022827586206896553,
+      "loss": 0.2158,
       "step": 2475
     },
     {
       "epoch": 45.02,
-      "learning_rate": 6.177777777777779e-05,
-      "loss": 0.238,
+      "learning_rate": 0.00022813793103448277,
+      "loss": 0.3115,
       "step": 2476
     },
     {
       "epoch": 45.04,
-      "learning_rate": 6.155555555555555e-05,
-      "loss": 0.2002,
+      "learning_rate": 0.000228,
+      "loss": 0.2858,
       "step": 2477
     },
     {
       "epoch": 45.05,
-      "learning_rate": 6.133333333333334e-05,
-      "loss": 0.2349,
+      "learning_rate": 0.00022786206896551725,
+      "loss": 0.3175,
       "step": 2478
     },
     {
       "epoch": 45.07,
-      "learning_rate": 6.111111111111111e-05,
-      "loss": 0.2302,
+      "learning_rate": 0.0002277241379310345,
+      "loss": 0.317,
       "step": 2479
     },
     {
       "epoch": 45.09,
-      "learning_rate": 6.088888888888889e-05,
-      "loss": 0.2432,
+      "learning_rate": 0.00022758620689655173,
+      "loss": 0.293,
       "step": 2480
     },
     {
       "epoch": 45.11,
-      "learning_rate": 6.0666666666666666e-05,
-      "loss": 0.1969,
+      "learning_rate": 0.00022744827586206897,
+      "loss": 0.2779,
       "step": 2481
     },
     {
       "epoch": 45.13,
-      "learning_rate": 6.044444444444445e-05,
-      "loss": 0.223,
+      "learning_rate": 0.00022731034482758623,
+      "loss": 0.3088,
       "step": 2482
     },
     {
       "epoch": 45.15,
-      "learning_rate": 6.0222222222222225e-05,
-      "loss": 0.2513,
+      "learning_rate": 0.00022717241379310347,
+      "loss": 0.3505,
       "step": 2483
     },
     {
       "epoch": 45.16,
-      "learning_rate": 6e-05,
-      "loss": 0.2573,
+      "learning_rate": 0.00022703448275862069,
+      "loss": 0.313,
       "step": 2484
     },
     {
       "epoch": 45.18,
-      "learning_rate": 5.977777777777778e-05,
-      "loss": 0.186,
+      "learning_rate": 0.00022689655172413792,
+      "loss": 0.2747,
       "step": 2485
     },
     {
       "epoch": 45.2,
-      "learning_rate": 5.955555555555556e-05,
-      "loss": 0.2183,
+      "learning_rate": 0.00022675862068965516,
+      "loss": 0.2828,
       "step": 2486
     },
     {
       "epoch": 45.22,
-      "learning_rate": 5.933333333333334e-05,
-      "loss": 0.2114,
+      "learning_rate": 0.0002266206896551724,
+      "loss": 0.2618,
       "step": 2487
     },
     {
       "epoch": 45.24,
-      "learning_rate": 5.911111111111111e-05,
-      "loss": 0.2558,
+      "learning_rate": 0.00022648275862068964,
+      "loss": 0.3209,
       "step": 2488
     },
     {
       "epoch": 45.25,
-      "learning_rate": 5.888888888888889e-05,
-      "loss": 0.2514,
+      "learning_rate": 0.00022634482758620688,
+      "loss": 0.3397,
       "step": 2489
     },
     {
       "epoch": 45.27,
-      "learning_rate": 5.8666666666666665e-05,
-      "loss": 0.1645,
+      "learning_rate": 0.00022620689655172415,
+      "loss": 0.211,
       "step": 2490
     },
     {
       "epoch": 45.29,
-      "learning_rate": 5.844444444444445e-05,
-      "loss": 0.218,
+      "learning_rate": 0.00022606896551724139,
+      "loss": 0.2854,
       "step": 2491
     },
     {
       "epoch": 45.31,
-      "learning_rate": 5.8222222222222224e-05,
-      "loss": 0.2432,
+      "learning_rate": 0.00022593103448275862,
+      "loss": 0.3363,
       "step": 2492
     },
     {
       "epoch": 45.33,
-      "learning_rate": 5.800000000000001e-05,
-      "loss": 0.2182,
+      "learning_rate": 0.00022579310344827586,
+      "loss": 0.2809,
       "step": 2493
     },
     {
       "epoch": 45.35,
-      "learning_rate": 5.7777777777777776e-05,
-      "loss": 0.2758,
+      "learning_rate": 0.0002256551724137931,
+      "loss": 0.3249,
       "step": 2494
     },
     {
       "epoch": 45.36,
-      "learning_rate": 5.755555555555555e-05,
-      "loss": 0.2215,
+      "learning_rate": 0.00022551724137931034,
+      "loss": 0.2761,
       "step": 2495
     },
     {
       "epoch": 45.38,
-      "learning_rate": 5.7333333333333336e-05,
-      "loss": 0.2215,
+      "learning_rate": 0.00022537931034482758,
+      "loss": 0.3093,
       "step": 2496
     },
     {
       "epoch": 45.4,
-      "learning_rate": 5.711111111111111e-05,
-      "loss": 0.2258,
+      "learning_rate": 0.00022524137931034482,
+      "loss": 0.293,
       "step": 2497
     },
     {
       "epoch": 45.42,
-      "learning_rate": 5.6888888888888895e-05,
-      "loss": 0.2071,
+      "learning_rate": 0.00022510344827586208,
+      "loss": 0.3047,
       "step": 2498
     },
     {
       "epoch": 45.44,
-      "learning_rate": 5.6666666666666664e-05,
-      "loss": 0.1857,
+      "learning_rate": 0.00022496551724137932,
+      "loss": 0.2721,
       "step": 2499
     },
     {
       "epoch": 45.45,
-      "learning_rate": 5.644444444444445e-05,
-      "loss": 0.2212,
+      "learning_rate": 0.00022482758620689656,
+      "loss": 0.2647,
       "step": 2500
     },
     {
       "epoch": 45.45,
-      "eval_cer": 0.07662783549978261,
-      "eval_loss": 0.35795867443084717,
-      "eval_runtime": 13.9558,
-      "eval_samples_per_second": 118.015,
-      "eval_steps_per_second": 1.863,
-      "eval_wer": 0.32723930139924423,
+      "eval_cer": 0.09142565855560053,
+      "eval_loss": 0.39010944962501526,
+      "eval_runtime": 14.0766,
+      "eval_samples_per_second": 117.003,
+      "eval_steps_per_second": 1.847,
+      "eval_wer": 0.3689102236748034,
       "step": 2500
     },
     {
       "epoch": 45.47,
-      "learning_rate": 5.622222222222222e-05,
-      "loss": 0.2582,
+      "learning_rate": 0.0002246896551724138,
+      "loss": 0.3409,
       "step": 2501
     },
     {
       "epoch": 45.49,
-      "learning_rate": 5.6e-05,
-      "loss": 0.2907,
+      "learning_rate": 0.00022455172413793104,
+      "loss": 0.3613,
       "step": 2502
     },
     {
       "epoch": 45.51,
-      "learning_rate": 5.577777777777778e-05,
-      "loss": 0.1927,
+      "learning_rate": 0.00022441379310344828,
+      "loss": 0.2853,
       "step": 2503
     },
     {
       "epoch": 45.53,
-      "learning_rate": 5.555555555555555e-05,
-      "loss": 0.2308,
+      "learning_rate": 0.00022427586206896552,
+      "loss": 0.2963,
       "step": 2504
     },
     {
       "epoch": 45.55,
-      "learning_rate": 5.5333333333333334e-05,
-      "loss": 0.237,
+      "learning_rate": 0.00022413793103448276,
+      "loss": 0.2997,
       "step": 2505
     },
     {
       "epoch": 45.56,
-      "learning_rate": 5.511111111111111e-05,
-      "loss": 0.246,
+      "learning_rate": 0.000224,
+      "loss": 0.2974,
       "step": 2506
     },
     {
       "epoch": 45.58,
-      "learning_rate": 5.4888888888888893e-05,
-      "loss": 0.2758,
+      "learning_rate": 0.00022386206896551726,
+      "loss": 0.3504,
       "step": 2507
     },
     {
       "epoch": 45.6,
-      "learning_rate": 5.466666666666667e-05,
-      "loss": 0.2534,
+      "learning_rate": 0.0002237241379310345,
+      "loss": 0.3046,
       "step": 2508
     },
     {
       "epoch": 45.62,
-      "learning_rate": 5.444444444444444e-05,
-      "loss": 0.2519,
+      "learning_rate": 0.00022358620689655174,
+      "loss": 0.3405,
       "step": 2509
     },
     {
       "epoch": 45.64,
-      "learning_rate": 5.422222222222222e-05,
-      "loss": 0.2757,
+      "learning_rate": 0.00022344827586206898,
+      "loss": 0.3741,
       "step": 2510
     },
     {
       "epoch": 45.65,
-      "learning_rate": 5.4e-05,
-      "loss": 0.2077,
+      "learning_rate": 0.00022331034482758622,
+      "loss": 0.306,
       "step": 2511
     },
     {
       "epoch": 45.67,
-      "learning_rate": 5.377777777777778e-05,
-      "loss": 0.2132,
+      "learning_rate": 0.00022317241379310346,
+      "loss": 0.3098,
       "step": 2512
     },
     {
       "epoch": 45.69,
-      "learning_rate": 5.355555555555556e-05,
-      "loss": 0.234,
+      "learning_rate": 0.0002230344827586207,
+      "loss": 0.2864,
       "step": 2513
     },
     {
       "epoch": 45.71,
-      "learning_rate": 5.333333333333334e-05,
-      "loss": 0.221,
+      "learning_rate": 0.00022289655172413794,
+      "loss": 0.2916,
       "step": 2514
     },
     {
       "epoch": 45.73,
-      "learning_rate": 5.311111111111111e-05,
-      "loss": 0.2128,
+      "learning_rate": 0.0002227586206896552,
+      "loss": 0.2964,
       "step": 2515
     },
     {
       "epoch": 45.75,
-      "learning_rate": 5.288888888888889e-05,
-      "loss": 0.2122,
+      "learning_rate": 0.00022262068965517244,
+      "loss": 0.298,
       "step": 2516
     },
     {
       "epoch": 45.76,
-      "learning_rate": 5.266666666666667e-05,
-      "loss": 0.2295,
+      "learning_rate": 0.00022248275862068968,
+      "loss": 0.3115,
       "step": 2517
     },
     {
       "epoch": 45.78,
-      "learning_rate": 5.2444444444444445e-05,
-      "loss": 0.2452,
+      "learning_rate": 0.0002223448275862069,
+      "loss": 0.2778,
       "step": 2518
     },
     {
       "epoch": 45.8,
-      "learning_rate": 5.222222222222223e-05,
-      "loss": 0.226,
+      "learning_rate": 0.00022220689655172413,
+      "loss": 0.3377,
       "step": 2519
     },
     {
       "epoch": 45.82,
-      "learning_rate": 5.2e-05,
-      "loss": 0.1796,
+      "learning_rate": 0.00022206896551724137,
+      "loss": 0.2478,
       "step": 2520
     },
     {
       "epoch": 45.84,
-      "learning_rate": 5.177777777777778e-05,
-      "loss": 0.2114,
+      "learning_rate": 0.0002219310344827586,
+      "loss": 0.2782,
       "step": 2521
     },
     {
       "epoch": 45.85,
-      "learning_rate": 5.1555555555555556e-05,
-      "loss": 0.2439,
+      "learning_rate": 0.00022179310344827585,
+      "loss": 0.3163,
       "step": 2522
     },
     {
       "epoch": 45.87,
-      "learning_rate": 5.133333333333334e-05,
-      "loss": 0.2505,
+      "learning_rate": 0.00022165517241379309,
+      "loss": 0.33,
       "step": 2523
     },
     {
       "epoch": 45.89,
-      "learning_rate": 5.1111111111111115e-05,
-      "loss": 0.1977,
+      "learning_rate": 0.00022151724137931035,
+      "loss": 0.2853,
       "step": 2524
     },
     {
       "epoch": 45.91,
-      "learning_rate": 5.0888888888888884e-05,
-      "loss": 0.1895,
+      "learning_rate": 0.0002213793103448276,
+      "loss": 0.272,
       "step": 2525
     },
     {
       "epoch": 45.93,
-      "learning_rate": 5.066666666666667e-05,
-      "loss": 0.2433,
+      "learning_rate": 0.00022124137931034483,
+      "loss": 0.3097,
       "step": 2526
     },
     {
       "epoch": 45.95,
-      "learning_rate": 5.044444444444444e-05,
-      "loss": 0.1748,
+      "learning_rate": 0.00022110344827586207,
+      "loss": 0.2195,
       "step": 2527
     },
     {
       "epoch": 45.96,
-      "learning_rate": 5.0222222222222226e-05,
-      "loss": 0.1977,
+      "learning_rate": 0.0002209655172413793,
+      "loss": 0.2672,
       "step": 2528
     },
     {
       "epoch": 45.98,
-      "learning_rate": 5e-05,
-      "loss": 0.2541,
+      "learning_rate": 0.00022082758620689655,
+      "loss": 0.3372,
       "step": 2529
     },
     {
       "epoch": 46.0,
-      "learning_rate": 4.977777777777778e-05,
-      "loss": 0.2051,
+      "learning_rate": 0.00022068965517241379,
+      "loss": 0.2193,
       "step": 2530
     },
     {
       "epoch": 46.02,
-      "learning_rate": 4.9555555555555555e-05,
-      "loss": 0.2141,
+      "learning_rate": 0.00022055172413793102,
+      "loss": 0.3226,
       "step": 2531
     },
     {
       "epoch": 46.04,
-      "learning_rate": 4.933333333333334e-05,
-      "loss": 0.2845,
+      "learning_rate": 0.0002204137931034483,
+      "loss": 0.341,
       "step": 2532
     },
     {
       "epoch": 46.05,
-      "learning_rate": 4.9111111111111114e-05,
-      "loss": 0.2142,
+      "learning_rate": 0.00022027586206896553,
+      "loss": 0.2613,
       "step": 2533
     },
     {
       "epoch": 46.07,
-      "learning_rate": 4.888888888888889e-05,
-      "loss": 0.2343,
+      "learning_rate": 0.00022013793103448277,
+      "loss": 0.2691,
       "step": 2534
     },
     {
       "epoch": 46.09,
-      "learning_rate": 4.8666666666666666e-05,
-      "loss": 0.2603,
+      "learning_rate": 0.00022,
+      "loss": 0.3219,
       "step": 2535
     },
     {
       "epoch": 46.11,
-      "learning_rate": 4.844444444444444e-05,
-      "loss": 0.2615,
+      "learning_rate": 0.00021986206896551725,
+      "loss": 0.2752,
       "step": 2536
     },
     {
       "epoch": 46.13,
-      "learning_rate": 4.8222222222222225e-05,
-      "loss": 0.203,
+      "learning_rate": 0.00021972413793103449,
+      "loss": 0.2949,
       "step": 2537
     },
     {
       "epoch": 46.15,
-      "learning_rate": 4.8e-05,
-      "loss": 0.2363,
+      "learning_rate": 0.00021958620689655172,
+      "loss": 0.3156,
       "step": 2538
     },
     {
       "epoch": 46.16,
-      "learning_rate": 4.7777777777777784e-05,
-      "loss": 0.2471,
+      "learning_rate": 0.00021944827586206896,
+      "loss": 0.2954,
       "step": 2539
     },
     {
       "epoch": 46.18,
-      "learning_rate": 4.7555555555555553e-05,
-      "loss": 0.2475,
+      "learning_rate": 0.00021931034482758623,
+      "loss": 0.2864,
       "step": 2540
     },
     {
       "epoch": 46.2,
-      "learning_rate": 4.733333333333333e-05,
-      "loss": 0.2092,
+      "learning_rate": 0.00021917241379310347,
+      "loss": 0.2785,
       "step": 2541
     },
     {
       "epoch": 46.22,
-      "learning_rate": 4.711111111111111e-05,
-      "loss": 0.2362,
+      "learning_rate": 0.0002190344827586207,
+      "loss": 0.3228,
       "step": 2542
     },
     {
       "epoch": 46.24,
-      "learning_rate": 4.688888888888889e-05,
-      "loss": 0.2098,
+      "learning_rate": 0.00021889655172413795,
+      "loss": 0.2816,
       "step": 2543
     },
     {
       "epoch": 46.25,
-      "learning_rate": 4.666666666666667e-05,
-      "loss": 0.1863,
+      "learning_rate": 0.00021875862068965519,
+      "loss": 0.2865,
       "step": 2544
     },
     {
       "epoch": 46.27,
-      "learning_rate": 4.644444444444444e-05,
-      "loss": 0.2327,
+      "learning_rate": 0.00021862068965517242,
+      "loss": 0.3048,
       "step": 2545
     },
     {
       "epoch": 46.29,
-      "learning_rate": 4.6222222222222224e-05,
-      "loss": 0.2298,
+      "learning_rate": 0.00021848275862068966,
+      "loss": 0.3293,
       "step": 2546
     },
     {
       "epoch": 46.31,
-      "learning_rate": 4.6e-05,
-      "loss": 0.2807,
+      "learning_rate": 0.0002183448275862069,
+      "loss": 0.3771,
       "step": 2547
     },
     {
       "epoch": 46.33,
-      "learning_rate": 4.5777777777777776e-05,
-      "loss": 0.2126,
+      "learning_rate": 0.00021820689655172414,
+      "loss": 0.2889,
       "step": 2548
     },
     {
       "epoch": 46.35,
-      "learning_rate": 4.555555555555556e-05,
-      "loss": 0.1888,
+      "learning_rate": 0.0002180689655172414,
+      "loss": 0.26,
       "step": 2549
     },
     {
       "epoch": 46.36,
-      "learning_rate": 4.5333333333333335e-05,
-      "loss": 0.2515,
+      "learning_rate": 0.00021793103448275865,
+      "loss": 0.3539,
       "step": 2550
     },
     {
       "epoch": 46.38,
-      "learning_rate": 4.511111111111111e-05,
-      "loss": 0.1946,
+      "learning_rate": 0.00021779310344827586,
+      "loss": 0.2349,
       "step": 2551
     },
     {
       "epoch": 46.4,
-      "learning_rate": 4.488888888888889e-05,
-      "loss": 0.2196,
+      "learning_rate": 0.0002176551724137931,
+      "loss": 0.3046,
       "step": 2552
     },
     {
       "epoch": 46.42,
-      "learning_rate": 4.466666666666667e-05,
-      "loss": 0.2202,
+      "learning_rate": 0.00021751724137931034,
+      "loss": 0.2721,
       "step": 2553
     },
     {
       "epoch": 46.44,
-      "learning_rate": 4.4444444444444447e-05,
-      "loss": 0.2176,
+      "learning_rate": 0.00021737931034482758,
+      "loss": 0.2974,
       "step": 2554
     },
     {
       "epoch": 46.45,
-      "learning_rate": 4.422222222222223e-05,
-      "loss": 0.1966,
+      "learning_rate": 0.00021724137931034481,
+      "loss": 0.279,
       "step": 2555
     },
     {
       "epoch": 46.47,
-      "learning_rate": 4.4e-05,
-      "loss": 0.2154,
+      "learning_rate": 0.00021710344827586205,
+      "loss": 0.2587,
       "step": 2556
     },
     {
       "epoch": 46.49,
-      "learning_rate": 4.3777777777777775e-05,
-      "loss": 0.2186,
+      "learning_rate": 0.00021696551724137932,
+      "loss": 0.2648,
       "step": 2557
     },
     {
       "epoch": 46.51,
-      "learning_rate": 4.355555555555556e-05,
-      "loss": 0.2162,
+      "learning_rate": 0.00021682758620689656,
+      "loss": 0.2788,
       "step": 2558
     },
     {
       "epoch": 46.53,
-      "learning_rate": 4.3333333333333334e-05,
-      "loss": 0.2535,
+      "learning_rate": 0.0002166896551724138,
+      "loss": 0.3269,
       "step": 2559
     },
     {
       "epoch": 46.55,
-      "learning_rate": 4.311111111111112e-05,
-      "loss": 0.2794,
+      "learning_rate": 0.00021655172413793104,
+      "loss": 0.3426,
       "step": 2560
     },
     {
       "epoch": 46.56,
-      "learning_rate": 4.2888888888888886e-05,
-      "loss": 0.1949,
+      "learning_rate": 0.00021641379310344828,
+      "loss": 0.2703,
       "step": 2561
     },
     {
       "epoch": 46.58,
-      "learning_rate": 4.266666666666667e-05,
-      "loss": 0.1945,
+      "learning_rate": 0.00021627586206896551,
+      "loss": 0.3133,
       "step": 2562
     },
     {
       "epoch": 46.6,
-      "learning_rate": 4.2444444444444445e-05,
-      "loss": 0.2501,
+      "learning_rate": 0.00021613793103448275,
+      "loss": 0.3377,
       "step": 2563
     },
     {
       "epoch": 46.62,
-      "learning_rate": 4.222222222222222e-05,
-      "loss": 0.2411,
+      "learning_rate": 0.000216,
+      "loss": 0.3097,
       "step": 2564
     },
     {
       "epoch": 46.64,
-      "learning_rate": 4.2000000000000004e-05,
-      "loss": 0.2424,
+      "learning_rate": 0.00021586206896551723,
+      "loss": 0.3159,
       "step": 2565
     },
     {
       "epoch": 46.65,
-      "learning_rate": 4.1777777777777774e-05,
-      "loss": 0.2447,
+      "learning_rate": 0.0002157241379310345,
+      "loss": 0.3286,
       "step": 2566
     },
     {
       "epoch": 46.67,
-      "learning_rate": 4.155555555555556e-05,
-      "loss": 0.2228,
+      "learning_rate": 0.00021558620689655174,
+      "loss": 0.283,
       "step": 2567
     },
     {
       "epoch": 46.69,
-      "learning_rate": 4.133333333333333e-05,
-      "loss": 0.2685,
+      "learning_rate": 0.00021544827586206897,
+      "loss": 0.3262,
       "step": 2568
     },
     {
       "epoch": 46.71,
-      "learning_rate": 4.1111111111111116e-05,
-      "loss": 0.2007,
+      "learning_rate": 0.00021531034482758621,
+      "loss": 0.2824,
       "step": 2569
     },
     {
       "epoch": 46.73,
-      "learning_rate": 4.088888888888889e-05,
-      "loss": 0.2456,
+      "learning_rate": 0.00021517241379310345,
+      "loss": 0.2963,
       "step": 2570
     },
     {
       "epoch": 46.75,
-      "learning_rate": 4.066666666666666e-05,
-      "loss": 0.1662,
+      "learning_rate": 0.0002150344827586207,
+      "loss": 0.2409,
       "step": 2571
     },
     {
       "epoch": 46.76,
-      "learning_rate": 4.0444444444444444e-05,
-      "loss": 0.1967,
+      "learning_rate": 0.00021489655172413793,
+      "loss": 0.25,
       "step": 2572
     },
     {
       "epoch": 46.78,
-      "learning_rate": 4.022222222222222e-05,
-      "loss": 0.1832,
+      "learning_rate": 0.00021475862068965517,
+      "loss": 0.2845,
       "step": 2573
     },
     {
       "epoch": 46.8,
-      "learning_rate": 4e-05,
-      "loss": 0.1818,
+      "learning_rate": 0.00021462068965517244,
+      "loss": 0.2573,
       "step": 2574
     },
     {
       "epoch": 46.82,
-      "learning_rate": 3.977777777777778e-05,
-      "loss": 0.2436,
+      "learning_rate": 0.00021448275862068967,
+      "loss": 0.3441,
       "step": 2575
     },
     {
       "epoch": 46.84,
-      "learning_rate": 3.9555555555555556e-05,
-      "loss": 0.2181,
+      "learning_rate": 0.00021434482758620691,
+      "loss": 0.2945,
       "step": 2576
     },
     {
       "epoch": 46.85,
-      "learning_rate": 3.933333333333333e-05,
-      "loss": 0.1932,
+      "learning_rate": 0.00021420689655172415,
+      "loss": 0.2965,
       "step": 2577
     },
     {
       "epoch": 46.87,
-      "learning_rate": 3.911111111111111e-05,
-      "loss": 0.224,
+      "learning_rate": 0.0002140689655172414,
+      "loss": 0.2692,
       "step": 2578
     },
     {
       "epoch": 46.89,
-      "learning_rate": 3.888888888888889e-05,
-      "loss": 0.2214,
+      "learning_rate": 0.00021393103448275863,
+      "loss": 0.3058,
       "step": 2579
     },
     {
       "epoch": 46.91,
-      "learning_rate": 3.866666666666667e-05,
-      "loss": 0.1769,
+      "learning_rate": 0.00021379310344827587,
+      "loss": 0.257,
       "step": 2580
     },
     {
       "epoch": 46.93,
-      "learning_rate": 3.844444444444445e-05,
-      "loss": 0.2556,
+      "learning_rate": 0.0002136551724137931,
+      "loss": 0.3325,
       "step": 2581
     },
     {
       "epoch": 46.95,
-      "learning_rate": 3.822222222222222e-05,
-      "loss": 0.1824,
+      "learning_rate": 0.00021351724137931037,
+      "loss": 0.2678,
       "step": 2582
     },
     {
       "epoch": 46.96,
-      "learning_rate": 3.8e-05,
-      "loss": 0.2403,
+      "learning_rate": 0.0002133793103448276,
+      "loss": 0.3068,
       "step": 2583
     },
     {
       "epoch": 46.98,
-      "learning_rate": 3.777777777777778e-05,
-      "loss": 0.2196,
+      "learning_rate": 0.00021324137931034483,
+      "loss": 0.2747,
       "step": 2584
     },
     {
       "epoch": 47.0,
-      "learning_rate": 3.755555555555556e-05,
-      "loss": 0.2077,
+      "learning_rate": 0.00021310344827586206,
+      "loss": 0.2457,
       "step": 2585
     },
     {
       "epoch": 47.02,
-      "learning_rate": 3.733333333333334e-05,
-      "loss": 0.2442,
+      "learning_rate": 0.0002129655172413793,
+      "loss": 0.3217,
       "step": 2586
     },
     {
       "epoch": 47.04,
-      "learning_rate": 3.711111111111111e-05,
-      "loss": 0.2416,
+      "learning_rate": 0.00021282758620689654,
+      "loss": 0.3253,
       "step": 2587
     },
     {
       "epoch": 47.05,
-      "learning_rate": 3.688888888888889e-05,
-      "loss": 0.2214,
+      "learning_rate": 0.00021268965517241378,
+      "loss": 0.2843,
       "step": 2588
     },
     {
       "epoch": 47.07,
-      "learning_rate": 3.6666666666666666e-05,
-      "loss": 0.2364,
+      "learning_rate": 0.00021255172413793102,
+      "loss": 0.3248,
       "step": 2589
     },
     {
       "epoch": 47.09,
-      "learning_rate": 3.644444444444445e-05,
-      "loss": 0.2178,
+      "learning_rate": 0.00021241379310344826,
+      "loss": 0.3087,
       "step": 2590
     },
     {
       "epoch": 47.11,
-      "learning_rate": 3.6222222222222225e-05,
-      "loss": 0.2209,
+      "learning_rate": 0.00021227586206896553,
+      "loss": 0.302,
       "step": 2591
     },
     {
       "epoch": 47.13,
-      "learning_rate": 3.6e-05,
-      "loss": 0.249,
+      "learning_rate": 0.00021213793103448276,
+      "loss": 0.332,
       "step": 2592
     },
     {
       "epoch": 47.15,
-      "learning_rate": 3.577777777777778e-05,
-      "loss": 0.2345,
+      "learning_rate": 0.000212,
+      "loss": 0.3334,
       "step": 2593
     },
     {
       "epoch": 47.16,
-      "learning_rate": 3.555555555555555e-05,
-      "loss": 0.268,
+      "learning_rate": 0.00021186206896551724,
+      "loss": 0.3552,
       "step": 2594
     },
     {
       "epoch": 47.18,
-      "learning_rate": 3.5333333333333336e-05,
-      "loss": 0.1918,
+      "learning_rate": 0.00021172413793103448,
+      "loss": 0.3016,
       "step": 2595
     },
     {
       "epoch": 47.2,
-      "learning_rate": 3.511111111111111e-05,
-      "loss": 0.2444,
+      "learning_rate": 0.00021158620689655172,
+      "loss": 0.31,
       "step": 2596
     },
     {
       "epoch": 47.22,
-      "learning_rate": 3.488888888888889e-05,
-      "loss": 0.269,
+      "learning_rate": 0.00021144827586206896,
+      "loss": 0.3261,
       "step": 2597
     },
     {
       "epoch": 47.24,
-      "learning_rate": 3.4666666666666665e-05,
-      "loss": 0.157,
+      "learning_rate": 0.0002113103448275862,
+      "loss": 0.2581,
       "step": 2598
     },
     {
       "epoch": 47.25,
-      "learning_rate": 3.444444444444445e-05,
-      "loss": 0.2181,
+      "learning_rate": 0.00021117241379310346,
+      "loss": 0.3094,
       "step": 2599
     },
     {
       "epoch": 47.27,
-      "learning_rate": 3.4222222222222224e-05,
-      "loss": 0.2326,
+      "learning_rate": 0.0002110344827586207,
+      "loss": 0.3104,
       "step": 2600
     },
     {
       "epoch": 47.29,
-      "learning_rate": 3.4000000000000007e-05,
-      "loss": 0.1719,
+      "learning_rate": 0.00021089655172413794,
+      "loss": 0.2634,
       "step": 2601
     },
     {
       "epoch": 47.31,
-      "learning_rate": 3.3777777777777776e-05,
-      "loss": 0.2245,
+      "learning_rate": 0.00021075862068965518,
+      "loss": 0.2858,
       "step": 2602
     },
     {
       "epoch": 47.33,
-      "learning_rate": 3.355555555555555e-05,
-      "loss": 0.2121,
+      "learning_rate": 0.00021062068965517242,
+      "loss": 0.2746,
       "step": 2603
     },
     {
       "epoch": 47.35,
-      "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.1638,
+      "learning_rate": 0.00021048275862068966,
+      "loss": 0.2523,
       "step": 2604
     },
     {
       "epoch": 47.36,
-      "learning_rate": 3.311111111111111e-05,
-      "loss": 0.234,
+      "learning_rate": 0.0002103448275862069,
+      "loss": 0.3128,
       "step": 2605
     },
     {
       "epoch": 47.38,
-      "learning_rate": 3.2888888888888894e-05,
-      "loss": 0.1966,
+      "learning_rate": 0.00021020689655172414,
+      "loss": 0.2681,
       "step": 2606
     },
     {
       "epoch": 47.4,
-      "learning_rate": 3.266666666666666e-05,
-      "loss": 0.1837,
+      "learning_rate": 0.0002100689655172414,
+      "loss": 0.2304,
       "step": 2607
     },
     {
       "epoch": 47.42,
-      "learning_rate": 3.2444444444444446e-05,
-      "loss": 0.2135,
+      "learning_rate": 0.00020993103448275864,
+      "loss": 0.3229,
       "step": 2608
     },
     {
       "epoch": 47.44,
-      "learning_rate": 3.222222222222222e-05,
-      "loss": 0.2165,
+      "learning_rate": 0.00020979310344827588,
+      "loss": 0.306,
       "step": 2609
     },
     {
       "epoch": 47.45,
-      "learning_rate": 3.2e-05,
-      "loss": 0.2344,
+      "learning_rate": 0.00020965517241379312,
+      "loss": 0.2946,
       "step": 2610
     },
     {
       "epoch": 47.47,
-      "learning_rate": 3.177777777777778e-05,
-      "loss": 0.2339,
+      "learning_rate": 0.00020951724137931036,
+      "loss": 0.3376,
       "step": 2611
     },
     {
       "epoch": 47.49,
-      "learning_rate": 3.155555555555556e-05,
-      "loss": 0.2154,
+      "learning_rate": 0.0002093793103448276,
+      "loss": 0.3113,
       "step": 2612
     },
     {
       "epoch": 47.51,
-      "learning_rate": 3.1333333333333334e-05,
-      "loss": 0.2033,
+      "learning_rate": 0.00020924137931034484,
+      "loss": 0.2995,
       "step": 2613
     },
     {
       "epoch": 47.53,
-      "learning_rate": 3.111111111111111e-05,
-      "loss": 0.1608,
+      "learning_rate": 0.00020910344827586208,
+      "loss": 0.2516,
       "step": 2614
     },
     {
       "epoch": 47.55,
-      "learning_rate": 3.088888888888889e-05,
-      "loss": 0.1913,
+      "learning_rate": 0.00020896551724137931,
+      "loss": 0.2865,
       "step": 2615
     },
     {
       "epoch": 47.56,
-      "learning_rate": 3.066666666666667e-05,
-      "loss": 0.2214,
+      "learning_rate": 0.00020882758620689658,
+      "loss": 0.2947,
       "step": 2616
     },
     {
       "epoch": 47.58,
-      "learning_rate": 3.0444444444444445e-05,
-      "loss": 0.2244,
+      "learning_rate": 0.0002086896551724138,
+      "loss": 0.309,
       "step": 2617
     },
     {
       "epoch": 47.6,
-      "learning_rate": 3.0222222222222225e-05,
-      "loss": 0.239,
+      "learning_rate": 0.00020855172413793103,
+      "loss": 0.3259,
       "step": 2618
     },
     {
       "epoch": 47.62,
-      "learning_rate": 3e-05,
-      "loss": 0.1806,
+      "learning_rate": 0.00020841379310344827,
+      "loss": 0.2469,
       "step": 2619
     },
     {
       "epoch": 47.64,
-      "learning_rate": 2.977777777777778e-05,
-      "loss": 0.2386,
+      "learning_rate": 0.0002082758620689655,
+      "loss": 0.3041,
       "step": 2620
     },
     {
       "epoch": 47.65,
-      "learning_rate": 2.9555555555555556e-05,
-      "loss": 0.2532,
+      "learning_rate": 0.00020813793103448275,
+      "loss": 0.3185,
       "step": 2621
     },
     {
       "epoch": 47.67,
-      "learning_rate": 2.9333333333333333e-05,
-      "loss": 0.2154,
+      "learning_rate": 0.000208,
+      "loss": 0.2903,
       "step": 2622
     },
     {
       "epoch": 47.69,
-      "learning_rate": 2.9111111111111112e-05,
-      "loss": 0.2592,
+      "learning_rate": 0.00020786206896551723,
+      "loss": 0.3242,
       "step": 2623
     },
     {
       "epoch": 47.71,
-      "learning_rate": 2.8888888888888888e-05,
-      "loss": 0.2664,
+      "learning_rate": 0.0002077241379310345,
+      "loss": 0.3501,
       "step": 2624
     },
     {
       "epoch": 47.73,
-      "learning_rate": 2.8666666666666668e-05,
-      "loss": 0.2617,
+      "learning_rate": 0.00020758620689655173,
+      "loss": 0.3395,
       "step": 2625
     },
     {
       "epoch": 47.75,
-      "learning_rate": 2.8444444444444447e-05,
-      "loss": 0.2545,
+      "learning_rate": 0.00020744827586206897,
+      "loss": 0.3013,
       "step": 2626
     },
     {
       "epoch": 47.76,
-      "learning_rate": 2.8222222222222223e-05,
-      "loss": 0.2635,
+      "learning_rate": 0.0002073103448275862,
+      "loss": 0.3326,
       "step": 2627
     },
     {
       "epoch": 47.78,
-      "learning_rate": 2.8e-05,
-      "loss": 0.2062,
+      "learning_rate": 0.00020717241379310345,
+      "loss": 0.2814,
       "step": 2628
     },
     {
       "epoch": 47.8,
-      "learning_rate": 2.7777777777777776e-05,
-      "loss": 0.1859,
+      "learning_rate": 0.0002070344827586207,
+      "loss": 0.2664,
       "step": 2629
     },
     {
       "epoch": 47.82,
-      "learning_rate": 2.7555555555555555e-05,
-      "loss": 0.2164,
+      "learning_rate": 0.00020689655172413793,
+      "loss": 0.3054,
       "step": 2630
     },
     {
       "epoch": 47.84,
-      "learning_rate": 2.7333333333333335e-05,
-      "loss": 0.2011,
+      "learning_rate": 0.00020675862068965516,
+      "loss": 0.2636,
       "step": 2631
     },
     {
       "epoch": 47.85,
-      "learning_rate": 2.711111111111111e-05,
-      "loss": 0.2209,
+      "learning_rate": 0.0002066206896551724,
+      "loss": 0.2922,
       "step": 2632
     },
     {
       "epoch": 47.87,
-      "learning_rate": 2.688888888888889e-05,
-      "loss": 0.2945,
+      "learning_rate": 0.00020648275862068967,
+      "loss": 0.3313,
       "step": 2633
     },
     {
       "epoch": 47.89,
-      "learning_rate": 2.666666666666667e-05,
-      "loss": 0.2132,
+      "learning_rate": 0.0002063448275862069,
+      "loss": 0.3003,
       "step": 2634
     },
     {
       "epoch": 47.91,
-      "learning_rate": 2.6444444444444446e-05,
-      "loss": 0.18,
+      "learning_rate": 0.00020620689655172415,
+      "loss": 0.2894,
       "step": 2635
     },
     {
       "epoch": 47.93,
-      "learning_rate": 2.6222222222222222e-05,
-      "loss": 0.2265,
+      "learning_rate": 0.0002060689655172414,
+      "loss": 0.2743,
       "step": 2636
     },
     {
       "epoch": 47.95,
-      "learning_rate": 2.6e-05,
-      "loss": 0.2071,
+      "learning_rate": 0.00020593103448275863,
+      "loss": 0.2691,
       "step": 2637
     },
     {
       "epoch": 47.96,
-      "learning_rate": 2.5777777777777778e-05,
-      "loss": 0.2066,
+      "learning_rate": 0.00020579310344827586,
+      "loss": 0.2545,
       "step": 2638
     },
     {
       "epoch": 47.98,
-      "learning_rate": 2.5555555555555557e-05,
-      "loss": 0.2516,
+      "learning_rate": 0.0002056551724137931,
+      "loss": 0.3424,
       "step": 2639
     },
     {
       "epoch": 48.0,
-      "learning_rate": 2.5333333333333334e-05,
-      "loss": 0.1752,
+      "learning_rate": 0.00020551724137931034,
+      "loss": 0.2241,
       "step": 2640
     },
     {
       "epoch": 48.02,
-      "learning_rate": 2.5111111111111113e-05,
-      "loss": 0.2517,
+      "learning_rate": 0.0002053793103448276,
+      "loss": 0.3236,
       "step": 2641
     },
     {
       "epoch": 48.04,
-      "learning_rate": 2.488888888888889e-05,
-      "loss": 0.2087,
+      "learning_rate": 0.00020524137931034485,
+      "loss": 0.2702,
       "step": 2642
     },
     {
       "epoch": 48.05,
-      "learning_rate": 2.466666666666667e-05,
-      "loss": 0.2145,
+      "learning_rate": 0.0002051034482758621,
+      "loss": 0.2688,
       "step": 2643
     },
     {
       "epoch": 48.07,
-      "learning_rate": 2.4444444444444445e-05,
-      "loss": 0.2155,
+      "learning_rate": 0.00020496551724137933,
+      "loss": 0.2726,
       "step": 2644
     },
     {
       "epoch": 48.09,
-      "learning_rate": 2.422222222222222e-05,
-      "loss": 0.2211,
+      "learning_rate": 0.00020482758620689656,
+      "loss": 0.2627,
       "step": 2645
     },
     {
       "epoch": 48.11,
-      "learning_rate": 2.4e-05,
-      "loss": 0.218,
+      "learning_rate": 0.0002046896551724138,
+      "loss": 0.2809,
       "step": 2646
     },
     {
       "epoch": 48.13,
-      "learning_rate": 2.3777777777777777e-05,
-      "loss": 0.2296,
+      "learning_rate": 0.00020455172413793104,
+      "loss": 0.2721,
       "step": 2647
     },
     {
       "epoch": 48.15,
-      "learning_rate": 2.3555555555555556e-05,
-      "loss": 0.2291,
+      "learning_rate": 0.00020441379310344828,
+      "loss": 0.3448,
       "step": 2648
     },
     {
       "epoch": 48.16,
-      "learning_rate": 2.3333333333333336e-05,
-      "loss": 0.2182,
+      "learning_rate": 0.00020427586206896555,
+      "loss": 0.3046,
       "step": 2649
     },
     {
       "epoch": 48.18,
-      "learning_rate": 2.3111111111111112e-05,
-      "loss": 0.198,
+      "learning_rate": 0.0002041379310344828,
+      "loss": 0.2881,
       "step": 2650
     },
     {
       "epoch": 48.2,
-      "learning_rate": 2.2888888888888888e-05,
-      "loss": 0.1929,
+      "learning_rate": 0.000204,
+      "loss": 0.2691,
       "step": 2651
     },
     {
       "epoch": 48.22,
-      "learning_rate": 2.2666666666666668e-05,
-      "loss": 0.1893,
+      "learning_rate": 0.00020386206896551724,
+      "loss": 0.2768,
       "step": 2652
     },
     {
       "epoch": 48.24,
-      "learning_rate": 2.2444444444444444e-05,
-      "loss": 0.1902,
+      "learning_rate": 0.00020372413793103448,
+      "loss": 0.2431,
       "step": 2653
     },
     {
       "epoch": 48.25,
-      "learning_rate": 2.2222222222222223e-05,
-      "loss": 0.2057,
+      "learning_rate": 0.00020358620689655172,
+      "loss": 0.2671,
       "step": 2654
     },
     {
       "epoch": 48.27,
-      "learning_rate": 2.2e-05,
-      "loss": 0.2048,
+      "learning_rate": 0.00020344827586206895,
+      "loss": 0.279,
       "step": 2655
     },
     {
       "epoch": 48.29,
-      "learning_rate": 2.177777777777778e-05,
-      "loss": 0.1787,
+      "learning_rate": 0.0002033103448275862,
+      "loss": 0.2424,
       "step": 2656
     },
     {
       "epoch": 48.31,
-      "learning_rate": 2.155555555555556e-05,
-      "loss": 0.2263,
+      "learning_rate": 0.00020317241379310343,
+      "loss": 0.3172,
       "step": 2657
     },
     {
       "epoch": 48.33,
-      "learning_rate": 2.1333333333333335e-05,
-      "loss": 0.2014,
+      "learning_rate": 0.0002030344827586207,
+      "loss": 0.2669,
       "step": 2658
     },
     {
       "epoch": 48.35,
-      "learning_rate": 2.111111111111111e-05,
-      "loss": 0.243,
+      "learning_rate": 0.00020289655172413794,
+      "loss": 0.3404,
       "step": 2659
     },
     {
       "epoch": 48.36,
-      "learning_rate": 2.0888888888888887e-05,
-      "loss": 0.2334,
+      "learning_rate": 0.00020275862068965518,
+      "loss": 0.2751,
       "step": 2660
     },
     {
       "epoch": 48.38,
-      "learning_rate": 2.0666666666666666e-05,
-      "loss": 0.2111,
+      "learning_rate": 0.00020262068965517242,
+      "loss": 0.2778,
       "step": 2661
     },
     {
       "epoch": 48.4,
-      "learning_rate": 2.0444444444444446e-05,
-      "loss": 0.2244,
+      "learning_rate": 0.00020248275862068965,
+      "loss": 0.2847,
       "step": 2662
     },
     {
       "epoch": 48.42,
-      "learning_rate": 2.0222222222222222e-05,
-      "loss": 0.1912,
+      "learning_rate": 0.0002023448275862069,
+      "loss": 0.2382,
       "step": 2663
     },
     {
       "epoch": 48.44,
-      "learning_rate": 2e-05,
-      "loss": 0.2112,
+      "learning_rate": 0.00020220689655172413,
+      "loss": 0.2831,
       "step": 2664
     },
     {
       "epoch": 48.45,
-      "learning_rate": 1.9777777777777778e-05,
-      "loss": 0.2275,
+      "learning_rate": 0.00020206896551724137,
+      "loss": 0.2823,
       "step": 2665
     },
     {
       "epoch": 48.47,
-      "learning_rate": 1.9555555555555554e-05,
-      "loss": 0.2544,
+      "learning_rate": 0.00020193103448275864,
+      "loss": 0.3124,
       "step": 2666
     },
     {
       "epoch": 48.49,
-      "learning_rate": 1.9333333333333333e-05,
-      "loss": 0.2237,
+      "learning_rate": 0.00020179310344827588,
+      "loss": 0.2702,
       "step": 2667
     },
     {
       "epoch": 48.51,
-      "learning_rate": 1.911111111111111e-05,
-      "loss": 0.1838,
+      "learning_rate": 0.00020165517241379311,
+      "loss": 0.2664,
       "step": 2668
     },
     {
       "epoch": 48.53,
-      "learning_rate": 1.888888888888889e-05,
-      "loss": 0.2287,
+      "learning_rate": 0.00020151724137931035,
+      "loss": 0.2878,
       "step": 2669
     },
     {
       "epoch": 48.55,
-      "learning_rate": 1.866666666666667e-05,
-      "loss": 0.1823,
+      "learning_rate": 0.0002013793103448276,
+      "loss": 0.2237,
       "step": 2670
     },
     {
       "epoch": 48.56,
-      "learning_rate": 1.8444444444444445e-05,
-      "loss": 0.1333,
+      "learning_rate": 0.00020124137931034483,
+      "loss": 0.1948,
       "step": 2671
     },
     {
       "epoch": 48.58,
-      "learning_rate": 1.8222222222222224e-05,
-      "loss": 0.2189,
+      "learning_rate": 0.00020110344827586207,
+      "loss": 0.2928,
       "step": 2672
     },
     {
       "epoch": 48.6,
-      "learning_rate": 1.8e-05,
-      "loss": 0.2188,
+      "learning_rate": 0.0002009655172413793,
+      "loss": 0.2788,
       "step": 2673
     },
     {
       "epoch": 48.62,
-      "learning_rate": 1.7777777777777777e-05,
-      "loss": 0.2385,
+      "learning_rate": 0.00020082758620689655,
+      "loss": 0.2773,
       "step": 2674
     },
     {
       "epoch": 48.64,
-      "learning_rate": 1.7555555555555556e-05,
-      "loss": 0.2159,
+      "learning_rate": 0.00020068965517241381,
+      "loss": 0.2798,
       "step": 2675
     },
     {
       "epoch": 48.65,
-      "learning_rate": 1.7333333333333332e-05,
-      "loss": 0.2196,
+      "learning_rate": 0.00020055172413793105,
+      "loss": 0.2963,
       "step": 2676
     },
     {
       "epoch": 48.67,
-      "learning_rate": 1.7111111111111112e-05,
-      "loss": 0.1965,
+      "learning_rate": 0.0002004137931034483,
+      "loss": 0.2437,
       "step": 2677
     },
     {
       "epoch": 48.69,
-      "learning_rate": 1.6888888888888888e-05,
-      "loss": 0.203,
+      "learning_rate": 0.00020027586206896553,
+      "loss": 0.2826,
       "step": 2678
     },
     {
       "epoch": 48.71,
-      "learning_rate": 1.6666666666666667e-05,
-      "loss": 0.2378,
+      "learning_rate": 0.00020013793103448277,
+      "loss": 0.3113,
       "step": 2679
     },
     {
       "epoch": 48.73,
-      "learning_rate": 1.6444444444444447e-05,
-      "loss": 0.2155,
+      "learning_rate": 0.0002,
+      "loss": 0.2817,
       "step": 2680
     },
     {
       "epoch": 48.75,
-      "learning_rate": 1.6222222222222223e-05,
-      "loss": 0.2281,
+      "learning_rate": 0.00019986206896551725,
+      "loss": 0.2897,
       "step": 2681
     },
     {
       "epoch": 48.76,
-      "learning_rate": 1.6e-05,
-      "loss": 0.1594,
+      "learning_rate": 0.0001997241379310345,
+      "loss": 0.2128,
       "step": 2682
     },
     {
       "epoch": 48.78,
-      "learning_rate": 1.577777777777778e-05,
-      "loss": 0.2088,
+      "learning_rate": 0.00019958620689655175,
+      "loss": 0.2631,
       "step": 2683
     },
     {
       "epoch": 48.8,
-      "learning_rate": 1.5555555555555555e-05,
-      "loss": 0.2286,
+      "learning_rate": 0.00019944827586206897,
+      "loss": 0.3037,
       "step": 2684
     },
     {
       "epoch": 48.82,
-      "learning_rate": 1.5333333333333334e-05,
-      "loss": 0.2282,
+      "learning_rate": 0.0001993103448275862,
+      "loss": 0.2786,
       "step": 2685
     },
     {
       "epoch": 48.84,
-      "learning_rate": 1.5111111111111112e-05,
-      "loss": 0.1929,
+      "learning_rate": 0.00019917241379310344,
+      "loss": 0.2496,
       "step": 2686
     },
     {
       "epoch": 48.85,
-      "learning_rate": 1.488888888888889e-05,
-      "loss": 0.2506,
+      "learning_rate": 0.00019903448275862068,
+      "loss": 0.3085,
       "step": 2687
     },
     {
       "epoch": 48.87,
-      "learning_rate": 1.4666666666666666e-05,
-      "loss": 0.153,
+      "learning_rate": 0.00019889655172413792,
+      "loss": 0.2301,
       "step": 2688
     },
     {
       "epoch": 48.89,
-      "learning_rate": 1.4444444444444444e-05,
-      "loss": 0.2424,
+      "learning_rate": 0.00019875862068965516,
+      "loss": 0.3397,
       "step": 2689
     },
     {
       "epoch": 48.91,
-      "learning_rate": 1.4222222222222224e-05,
-      "loss": 0.1728,
+      "learning_rate": 0.0001986206896551724,
+      "loss": 0.2318,
       "step": 2690
     },
     {
       "epoch": 48.93,
-      "learning_rate": 1.4e-05,
-      "loss": 0.2029,
+      "learning_rate": 0.00019848275862068967,
+      "loss": 0.2916,
       "step": 2691
     },
     {
       "epoch": 48.95,
-      "learning_rate": 1.3777777777777778e-05,
-      "loss": 0.2139,
+      "learning_rate": 0.0001983448275862069,
+      "loss": 0.2524,
       "step": 2692
     },
     {
       "epoch": 48.96,
-      "learning_rate": 1.3555555555555555e-05,
-      "loss": 0.1705,
+      "learning_rate": 0.00019820689655172414,
+      "loss": 0.239,
       "step": 2693
     },
     {
       "epoch": 48.98,
-      "learning_rate": 1.3333333333333335e-05,
-      "loss": 0.1655,
+      "learning_rate": 0.00019806896551724138,
+      "loss": 0.2732,
       "step": 2694
     },
     {
       "epoch": 49.0,
-      "learning_rate": 1.3111111111111111e-05,
-      "loss": 0.2094,
+      "learning_rate": 0.00019793103448275862,
+      "loss": 0.2065,
       "step": 2695
     },
     {
       "epoch": 49.02,
-      "learning_rate": 1.2888888888888889e-05,
-      "loss": 0.2007,
+      "learning_rate": 0.00019779310344827586,
+      "loss": 0.2746,
       "step": 2696
     },
     {
       "epoch": 49.04,
-      "learning_rate": 1.2666666666666667e-05,
-      "loss": 0.2344,
+      "learning_rate": 0.0001976551724137931,
+      "loss": 0.346,
       "step": 2697
     },
     {
       "epoch": 49.05,
-      "learning_rate": 1.2444444444444445e-05,
-      "loss": 0.1848,
+      "learning_rate": 0.00019751724137931034,
+      "loss": 0.2662,
       "step": 2698
     },
     {
       "epoch": 49.07,
-      "learning_rate": 1.2222222222222222e-05,
-      "loss": 0.2546,
+      "learning_rate": 0.00019737931034482758,
+      "loss": 0.3156,
       "step": 2699
     },
     {
       "epoch": 49.09,
-      "learning_rate": 1.2e-05,
-      "loss": 0.1728,
+      "learning_rate": 0.00019724137931034484,
+      "loss": 0.2549,
       "step": 2700
     },
     {
       "epoch": 49.11,
-      "learning_rate": 1.1777777777777778e-05,
-      "loss": 0.1892,
+      "learning_rate": 0.00019710344827586208,
+      "loss": 0.2515,
       "step": 2701
     },
     {
       "epoch": 49.13,
-      "learning_rate": 1.1555555555555556e-05,
-      "loss": 0.2536,
+      "learning_rate": 0.00019696551724137932,
+      "loss": 0.3011,
       "step": 2702
     },
     {
       "epoch": 49.15,
-      "learning_rate": 1.1333333333333334e-05,
-      "loss": 0.2374,
+      "learning_rate": 0.00019682758620689656,
+      "loss": 0.3199,
       "step": 2703
     },
     {
       "epoch": 49.16,
-      "learning_rate": 1.1111111111111112e-05,
-      "loss": 0.2107,
+      "learning_rate": 0.0001966896551724138,
+      "loss": 0.275,
       "step": 2704
     },
     {
       "epoch": 49.18,
-      "learning_rate": 1.088888888888889e-05,
-      "loss": 0.2016,
+      "learning_rate": 0.00019655172413793104,
+      "loss": 0.2533,
       "step": 2705
     },
     {
       "epoch": 49.2,
-      "learning_rate": 1.0666666666666667e-05,
-      "loss": 0.2073,
+      "learning_rate": 0.00019641379310344828,
+      "loss": 0.264,
       "step": 2706
     },
     {
       "epoch": 49.22,
-      "learning_rate": 1.0444444444444443e-05,
-      "loss": 0.1872,
+      "learning_rate": 0.00019627586206896552,
+      "loss": 0.2669,
       "step": 2707
     },
     {
       "epoch": 49.24,
-      "learning_rate": 1.0222222222222223e-05,
-      "loss": 0.1586,
+      "learning_rate": 0.00019613793103448278,
+      "loss": 0.2116,
       "step": 2708
     },
     {
       "epoch": 49.25,
-      "learning_rate": 1e-05,
-      "loss": 0.1931,
+      "learning_rate": 0.00019600000000000002,
+      "loss": 0.2427,
       "step": 2709
     },
     {
       "epoch": 49.27,
-      "learning_rate": 9.777777777777777e-06,
-      "loss": 0.2185,
+      "learning_rate": 0.00019586206896551726,
+      "loss": 0.2811,
       "step": 2710
     },
     {
       "epoch": 49.29,
-      "learning_rate": 9.555555555555555e-06,
-      "loss": 0.1904,
+      "learning_rate": 0.0001957241379310345,
+      "loss": 0.2642,
       "step": 2711
     },
     {
       "epoch": 49.31,
-      "learning_rate": 9.333333333333334e-06,
-      "loss": 0.204,
+      "learning_rate": 0.00019558620689655174,
+      "loss": 0.2792,
       "step": 2712
     },
     {
       "epoch": 49.33,
-      "learning_rate": 9.111111111111112e-06,
-      "loss": 0.2426,
+      "learning_rate": 0.00019544827586206898,
+      "loss": 0.3007,
       "step": 2713
     },
     {
       "epoch": 49.35,
-      "learning_rate": 8.888888888888888e-06,
-      "loss": 0.2577,
+      "learning_rate": 0.00019531034482758622,
+      "loss": 0.3246,
       "step": 2714
     },
     {
       "epoch": 49.36,
-      "learning_rate": 8.666666666666666e-06,
-      "loss": 0.2654,
+      "learning_rate": 0.00019517241379310345,
+      "loss": 0.3703,
       "step": 2715
     },
     {
       "epoch": 49.38,
-      "learning_rate": 8.444444444444444e-06,
-      "loss": 0.1843,
+      "learning_rate": 0.00019503448275862072,
+      "loss": 0.2253,
       "step": 2716
     },
     {
       "epoch": 49.4,
-      "learning_rate": 8.222222222222223e-06,
-      "loss": 0.2435,
+      "learning_rate": 0.00019489655172413793,
+      "loss": 0.3132,
       "step": 2717
     },
     {
       "epoch": 49.42,
-      "learning_rate": 8e-06,
-      "loss": 0.2246,
+      "learning_rate": 0.00019475862068965517,
+      "loss": 0.2555,
       "step": 2718
     },
     {
       "epoch": 49.44,
-      "learning_rate": 7.777777777777777e-06,
-      "loss": 0.189,
+      "learning_rate": 0.0001946206896551724,
+      "loss": 0.2615,
       "step": 2719
     },
     {
       "epoch": 49.45,
-      "learning_rate": 7.555555555555556e-06,
-      "loss": 0.1748,
+      "learning_rate": 0.00019448275862068965,
+      "loss": 0.2533,
       "step": 2720
     },
     {
       "epoch": 49.47,
-      "learning_rate": 7.333333333333333e-06,
-      "loss": 0.226,
+      "learning_rate": 0.0001943448275862069,
+      "loss": 0.2771,
       "step": 2721
     },
     {
       "epoch": 49.49,
-      "learning_rate": 7.111111111111112e-06,
-      "loss": 0.2503,
+      "learning_rate": 0.00019420689655172413,
+      "loss": 0.2822,
       "step": 2722
     },
     {
       "epoch": 49.51,
-      "learning_rate": 6.888888888888889e-06,
-      "loss": 0.1823,
+      "learning_rate": 0.00019406896551724137,
+      "loss": 0.2699,
       "step": 2723
     },
     {
       "epoch": 49.53,
-      "learning_rate": 6.6666666666666675e-06,
-      "loss": 0.1618,
+      "learning_rate": 0.0001939310344827586,
+      "loss": 0.2253,
       "step": 2724
     },
     {
       "epoch": 49.55,
-      "learning_rate": 6.4444444444444445e-06,
-      "loss": 0.2396,
+      "learning_rate": 0.00019379310344827587,
+      "loss": 0.3027,
       "step": 2725
     },
     {
       "epoch": 49.56,
-      "learning_rate": 6.222222222222222e-06,
-      "loss": 0.1829,
+      "learning_rate": 0.0001936551724137931,
+      "loss": 0.2583,
       "step": 2726
     },
     {
       "epoch": 49.58,
-      "learning_rate": 6e-06,
-      "loss": 0.211,
+      "learning_rate": 0.00019351724137931035,
+      "loss": 0.2986,
       "step": 2727
     },
     {
       "epoch": 49.6,
-      "learning_rate": 5.777777777777778e-06,
-      "loss": 0.2055,
+      "learning_rate": 0.0001933793103448276,
+      "loss": 0.2649,
       "step": 2728
     },
     {
       "epoch": 49.62,
-      "learning_rate": 5.555555555555556e-06,
-      "loss": 0.2388,
+      "learning_rate": 0.00019324137931034483,
+      "loss": 0.3356,
       "step": 2729
     },
     {
       "epoch": 49.64,
-      "learning_rate": 5.333333333333334e-06,
-      "loss": 0.1925,
+      "learning_rate": 0.00019310344827586207,
+      "loss": 0.2666,
       "step": 2730
     },
     {
       "epoch": 49.65,
-      "learning_rate": 5.1111111111111115e-06,
-      "loss": 0.1779,
+      "learning_rate": 0.0001929655172413793,
+      "loss": 0.2305,
       "step": 2731
     },
     {
       "epoch": 49.67,
-      "learning_rate": 4.8888888888888885e-06,
-      "loss": 0.2195,
+      "learning_rate": 0.00019282758620689654,
+      "loss": 0.2657,
       "step": 2732
     },
     {
       "epoch": 49.69,
-      "learning_rate": 4.666666666666667e-06,
-      "loss": 0.2376,
+      "learning_rate": 0.0001926896551724138,
+      "loss": 0.3061,
       "step": 2733
     },
     {
       "epoch": 49.71,
-      "learning_rate": 4.444444444444444e-06,
-      "loss": 0.1738,
+      "learning_rate": 0.00019255172413793105,
+      "loss": 0.2458,
       "step": 2734
     },
     {
       "epoch": 49.73,
-      "learning_rate": 4.222222222222222e-06,
-      "loss": 0.2451,
+      "learning_rate": 0.0001924137931034483,
+      "loss": 0.3097,
       "step": 2735
     },
     {
       "epoch": 49.75,
-      "learning_rate": 4e-06,
-      "loss": 0.2171,
+      "learning_rate": 0.00019227586206896553,
+      "loss": 0.2823,
       "step": 2736
     },
     {
       "epoch": 49.76,
-      "learning_rate": 3.777777777777778e-06,
-      "loss": 0.2027,
+      "learning_rate": 0.00019213793103448277,
+      "loss": 0.2644,
       "step": 2737
     },
     {
       "epoch": 49.78,
-      "learning_rate": 3.555555555555556e-06,
-      "loss": 0.1634,
+      "learning_rate": 0.000192,
+      "loss": 0.2516,
       "step": 2738
     },
     {
       "epoch": 49.8,
-      "learning_rate": 3.3333333333333337e-06,
-      "loss": 0.2337,
+      "learning_rate": 0.00019186206896551724,
+      "loss": 0.2822,
       "step": 2739
     },
     {
       "epoch": 49.82,
-      "learning_rate": 3.111111111111111e-06,
-      "loss": 0.2448,
+      "learning_rate": 0.00019172413793103448,
+      "loss": 0.3048,
       "step": 2740
     },
     {
       "epoch": 49.84,
-      "learning_rate": 2.888888888888889e-06,
-      "loss": 0.2048,
+      "learning_rate": 0.00019158620689655172,
+      "loss": 0.2759,
       "step": 2741
     },
     {
       "epoch": 49.85,
-      "learning_rate": 2.666666666666667e-06,
-      "loss": 0.2138,
+      "learning_rate": 0.000191448275862069,
+      "loss": 0.2847,
       "step": 2742
     },
     {
       "epoch": 49.87,
-      "learning_rate": 2.4444444444444442e-06,
-      "loss": 0.1679,
+      "learning_rate": 0.00019131034482758623,
+      "loss": 0.2544,
       "step": 2743
     },
     {
       "epoch": 49.89,
-      "learning_rate": 2.222222222222222e-06,
-      "loss": 0.2228,
+      "learning_rate": 0.00019117241379310347,
+      "loss": 0.2796,
       "step": 2744
     },
     {
       "epoch": 49.91,
-      "learning_rate": 2e-06,
-      "loss": 0.2032,
+      "learning_rate": 0.0001910344827586207,
+      "loss": 0.2485,
       "step": 2745
     },
     {
       "epoch": 49.93,
-      "learning_rate": 1.777777777777778e-06,
-      "loss": 0.2309,
+      "learning_rate": 0.00019089655172413794,
+      "loss": 0.3075,
       "step": 2746
     },
     {
       "epoch": 49.95,
-      "learning_rate": 1.5555555555555556e-06,
-      "loss": 0.2616,
+      "learning_rate": 0.00019075862068965518,
+      "loss": 0.3134,
       "step": 2747
     },
     {
       "epoch": 49.96,
-      "learning_rate": 1.3333333333333334e-06,
-      "loss": 0.1936,
+      "learning_rate": 0.00019062068965517242,
+      "loss": 0.2963,
       "step": 2748
     },
     {
       "epoch": 49.98,
-      "learning_rate": 1.111111111111111e-06,
-      "loss": 0.2173,
+      "learning_rate": 0.00019048275862068966,
+      "loss": 0.2806,
       "step": 2749
     },
     {
       "epoch": 50.0,
-      "learning_rate": 8.88888888888889e-07,
-      "loss": 0.1733,
+      "learning_rate": 0.0001903448275862069,
+      "loss": 0.2411,
       "step": 2750
     },
     {
-      "epoch": 50.0,
-      "step": 2750,
-      "total_flos": 2.4190366418747785e+19,
-      "train_loss": 0.8772053740783171,
-      "train_runtime": 2063.7987,
-      "train_samples_per_second": 84.262,
-      "train_steps_per_second": 1.332
+      "epoch": 50.02,
+      "learning_rate": 0.00019020689655172414,
+      "loss": 0.2574,
+      "step": 2751
+    },
+    {
+      "epoch": 50.04,
+      "learning_rate": 0.00019006896551724138,
+      "loss": 0.2544,
+      "step": 2752
+    },
+    {
+      "epoch": 50.05,
+      "learning_rate": 0.00018993103448275862,
+      "loss": 0.2141,
+      "step": 2753
+    },
+    {
+      "epoch": 50.07,
+      "learning_rate": 0.00018979310344827586,
+      "loss": 0.2433,
+      "step": 2754
+    },
+    {
+      "epoch": 50.09,
+      "learning_rate": 0.0001896551724137931,
+      "loss": 0.2587,
+      "step": 2755
+    },
+    {
+      "epoch": 50.11,
+      "learning_rate": 0.00018951724137931033,
+      "loss": 0.2565,
+      "step": 2756
+    },
+    {
+      "epoch": 50.13,
+      "learning_rate": 0.00018937931034482757,
+      "loss": 0.2521,
+      "step": 2757
+    },
+    {
+      "epoch": 50.15,
+      "learning_rate": 0.0001892413793103448,
+      "loss": 0.2893,
+      "step": 2758
+    },
+    {
+      "epoch": 50.16,
+      "learning_rate": 0.00018910344827586208,
+      "loss": 0.2454,
+      "step": 2759
+    },
+    {
+      "epoch": 50.18,
+      "learning_rate": 0.00018896551724137932,
+      "loss": 0.2835,
+      "step": 2760
+    },
+    {
+      "epoch": 50.2,
+      "learning_rate": 0.00018882758620689656,
+      "loss": 0.2714,
+      "step": 2761
+    },
+    {
+      "epoch": 50.22,
+      "learning_rate": 0.0001886896551724138,
+      "loss": 0.2748,
+      "step": 2762
+    },
+    {
+      "epoch": 50.24,
+      "learning_rate": 0.00018855172413793103,
+      "loss": 0.2797,
+      "step": 2763
+    },
+    {
+      "epoch": 50.25,
+      "learning_rate": 0.00018841379310344827,
+      "loss": 0.2419,
+      "step": 2764
+    },
+    {
+      "epoch": 50.27,
+      "learning_rate": 0.0001882758620689655,
+      "loss": 0.3022,
+      "step": 2765
+    },
+    {
+      "epoch": 50.29,
+      "learning_rate": 0.00018813793103448275,
+      "loss": 0.229,
+      "step": 2766
+    },
+    {
+      "epoch": 50.31,
+      "learning_rate": 0.00018800000000000002,
+      "loss": 0.2917,
+      "step": 2767
+    },
+    {
+      "epoch": 50.33,
+      "learning_rate": 0.00018786206896551726,
+      "loss": 0.2612,
+      "step": 2768
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 0.0001877241379310345,
+      "loss": 0.3055,
+      "step": 2769
+    },
+    {
+      "epoch": 50.36,
+      "learning_rate": 0.00018758620689655173,
+      "loss": 0.295,
+      "step": 2770
+    },
+    {
+      "epoch": 50.38,
+      "learning_rate": 0.00018744827586206897,
+      "loss": 0.2901,
+      "step": 2771
+    },
+    {
+      "epoch": 50.4,
+      "learning_rate": 0.0001873103448275862,
+      "loss": 0.3311,
+      "step": 2772
+    },
+    {
+      "epoch": 50.42,
+      "learning_rate": 0.00018717241379310345,
+      "loss": 0.2802,
+      "step": 2773
+    },
+    {
+      "epoch": 50.44,
+      "learning_rate": 0.0001870344827586207,
+      "loss": 0.2504,
+      "step": 2774
+    },
+    {
+      "epoch": 50.45,
+      "learning_rate": 0.00018689655172413795,
+      "loss": 0.2375,
+      "step": 2775
+    },
+    {
+      "epoch": 50.47,
+      "learning_rate": 0.0001867586206896552,
+      "loss": 0.2633,
+      "step": 2776
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 0.00018662068965517243,
+      "loss": 0.3031,
+      "step": 2777
+    },
+    {
+      "epoch": 50.51,
+      "learning_rate": 0.00018648275862068967,
+      "loss": 0.297,
+      "step": 2778
+    },
+    {
+      "epoch": 50.53,
+      "learning_rate": 0.0001863448275862069,
+      "loss": 0.2651,
+      "step": 2779
+    },
+    {
+      "epoch": 50.55,
+      "learning_rate": 0.00018620689655172415,
+      "loss": 0.2716,
+      "step": 2780
+    },
+    {
+      "epoch": 50.56,
+      "learning_rate": 0.0001860689655172414,
+      "loss": 0.274,
+      "step": 2781
+    },
+    {
+      "epoch": 50.58,
+      "learning_rate": 0.00018593103448275863,
+      "loss": 0.2751,
+      "step": 2782
+    },
+    {
+      "epoch": 50.6,
+      "learning_rate": 0.00018579310344827587,
+      "loss": 0.294,
+      "step": 2783
+    },
+    {
+      "epoch": 50.62,
+      "learning_rate": 0.0001856551724137931,
+      "loss": 0.2765,
+      "step": 2784
+    },
+    {
+      "epoch": 50.64,
+      "learning_rate": 0.00018551724137931034,
+      "loss": 0.2435,
+      "step": 2785
+    },
+    {
+      "epoch": 50.65,
+      "learning_rate": 0.00018537931034482758,
+      "loss": 0.3064,
+      "step": 2786
+    },
+    {
+      "epoch": 50.67,
+      "learning_rate": 0.00018524137931034482,
+      "loss": 0.2378,
+      "step": 2787
+    },
+    {
+      "epoch": 50.69,
+      "learning_rate": 0.00018510344827586206,
+      "loss": 0.2586,
+      "step": 2788
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 0.0001849655172413793,
+      "loss": 0.2754,
+      "step": 2789
+    },
+    {
+      "epoch": 50.73,
+      "learning_rate": 0.00018482758620689654,
+      "loss": 0.2767,
+      "step": 2790
+    },
+    {
+      "epoch": 50.75,
+      "learning_rate": 0.00018468965517241378,
+      "loss": 0.237,
+      "step": 2791
+    },
+    {
+      "epoch": 50.76,
+      "learning_rate": 0.00018455172413793104,
+      "loss": 0.297,
+      "step": 2792
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 0.00018441379310344828,
+      "loss": 0.2862,
+      "step": 2793
+    },
+    {
+      "epoch": 50.8,
+      "learning_rate": 0.00018427586206896552,
+      "loss": 0.2041,
+      "step": 2794
+    },
+    {
+      "epoch": 50.82,
+      "learning_rate": 0.00018413793103448276,
+      "loss": 0.2865,
+      "step": 2795
+    },
+    {
+      "epoch": 50.84,
+      "learning_rate": 0.000184,
+      "loss": 0.2395,
+      "step": 2796
+    },
+    {
+      "epoch": 50.85,
+      "learning_rate": 0.00018386206896551724,
+      "loss": 0.2317,
+      "step": 2797
+    },
+    {
+      "epoch": 50.87,
+      "learning_rate": 0.00018372413793103448,
+      "loss": 0.305,
+      "step": 2798
+    },
+    {
+      "epoch": 50.89,
+      "learning_rate": 0.00018358620689655172,
+      "loss": 0.253,
+      "step": 2799
+    },
+    {
+      "epoch": 50.91,
+      "learning_rate": 0.00018344827586206898,
+      "loss": 0.2914,
+      "step": 2800
+    },
+    {
+      "epoch": 50.93,
+      "learning_rate": 0.00018331034482758622,
+      "loss": 0.2962,
+      "step": 2801
+    },
+    {
+      "epoch": 50.95,
+      "learning_rate": 0.00018317241379310346,
+      "loss": 0.2536,
+      "step": 2802
+    },
+    {
+      "epoch": 50.96,
+      "learning_rate": 0.0001830344827586207,
+      "loss": 0.2861,
+      "step": 2803
+    },
+    {
+      "epoch": 50.98,
+      "learning_rate": 0.00018289655172413794,
+      "loss": 0.2626,
+      "step": 2804
+    },
+    {
+      "epoch": 51.0,
+      "learning_rate": 0.00018275862068965518,
+      "loss": 0.1481,
+      "step": 2805
+    },
+    {
+      "epoch": 51.02,
+      "learning_rate": 0.00018262068965517242,
+      "loss": 0.2425,
+      "step": 2806
+    },
+    {
+      "epoch": 51.04,
+      "learning_rate": 0.00018248275862068966,
+      "loss": 0.2889,
+      "step": 2807
+    },
+    {
+      "epoch": 51.05,
+      "learning_rate": 0.0001823448275862069,
+      "loss": 0.2601,
+      "step": 2808
+    },
+    {
+      "epoch": 51.07,
+      "learning_rate": 0.00018220689655172416,
+      "loss": 0.2903,
+      "step": 2809
+    },
+    {
+      "epoch": 51.09,
+      "learning_rate": 0.0001820689655172414,
+      "loss": 0.3356,
+      "step": 2810
+    },
+    {
+      "epoch": 51.11,
+      "learning_rate": 0.00018193103448275864,
+      "loss": 0.2982,
+      "step": 2811
+    },
+    {
+      "epoch": 51.13,
+      "learning_rate": 0.00018179310344827588,
+      "loss": 0.2792,
+      "step": 2812
+    },
+    {
+      "epoch": 51.15,
+      "learning_rate": 0.00018165517241379312,
+      "loss": 0.3058,
+      "step": 2813
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 0.00018151724137931036,
+      "loss": 0.2344,
+      "step": 2814
+    },
+    {
+      "epoch": 51.18,
+      "learning_rate": 0.0001813793103448276,
+      "loss": 0.2673,
+      "step": 2815
+    },
+    {
+      "epoch": 51.2,
+      "learning_rate": 0.00018124137931034483,
+      "loss": 0.235,
+      "step": 2816
+    },
+    {
+      "epoch": 51.22,
+      "learning_rate": 0.00018110344827586207,
+      "loss": 0.2189,
+      "step": 2817
+    },
+    {
+      "epoch": 51.24,
+      "learning_rate": 0.0001809655172413793,
+      "loss": 0.2472,
+      "step": 2818
+    },
+    {
+      "epoch": 51.25,
+      "learning_rate": 0.00018082758620689655,
+      "loss": 0.2295,
+      "step": 2819
+    },
+    {
+      "epoch": 51.27,
+      "learning_rate": 0.0001806896551724138,
+      "loss": 0.2629,
+      "step": 2820
+    },
+    {
+      "epoch": 51.29,
+      "learning_rate": 0.00018055172413793103,
+      "loss": 0.2829,
+      "step": 2821
+    },
+    {
+      "epoch": 51.31,
+      "learning_rate": 0.00018041379310344827,
+      "loss": 0.2691,
+      "step": 2822
+    },
+    {
+      "epoch": 51.33,
+      "learning_rate": 0.0001802758620689655,
+      "loss": 0.2974,
+      "step": 2823
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 0.00018013793103448275,
+      "loss": 0.2638,
+      "step": 2824
+    },
+    {
+      "epoch": 51.36,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.2376,
+      "step": 2825
+    },
+    {
+      "epoch": 51.38,
+      "learning_rate": 0.00017986206896551725,
+      "loss": 0.2357,
+      "step": 2826
+    },
+    {
+      "epoch": 51.4,
+      "learning_rate": 0.0001797241379310345,
+      "loss": 0.2595,
+      "step": 2827
+    },
+    {
+      "epoch": 51.42,
+      "learning_rate": 0.00017958620689655173,
+      "loss": 0.2749,
+      "step": 2828
+    },
+    {
+      "epoch": 51.44,
+      "learning_rate": 0.00017944827586206897,
+      "loss": 0.3021,
+      "step": 2829
+    },
+    {
+      "epoch": 51.45,
+      "learning_rate": 0.0001793103448275862,
+      "loss": 0.2104,
+      "step": 2830
+    },
+    {
+      "epoch": 51.47,
+      "learning_rate": 0.00017917241379310345,
+      "loss": 0.2089,
+      "step": 2831
+    },
+    {
+      "epoch": 51.49,
+      "learning_rate": 0.00017903448275862068,
+      "loss": 0.2376,
+      "step": 2832
+    },
+    {
+      "epoch": 51.51,
+      "learning_rate": 0.00017889655172413792,
+      "loss": 0.2687,
+      "step": 2833
+    },
+    {
+      "epoch": 51.53,
+      "learning_rate": 0.0001787586206896552,
+      "loss": 0.2248,
+      "step": 2834
+    },
+    {
+      "epoch": 51.55,
+      "learning_rate": 0.00017862068965517243,
+      "loss": 0.2564,
+      "step": 2835
+    },
+    {
+      "epoch": 51.56,
+      "learning_rate": 0.00017848275862068967,
+      "loss": 0.2941,
+      "step": 2836
+    },
+    {
+      "epoch": 51.58,
+      "learning_rate": 0.0001783448275862069,
+      "loss": 0.2483,
+      "step": 2837
+    },
+    {
+      "epoch": 51.6,
+      "learning_rate": 0.00017820689655172414,
+      "loss": 0.3018,
+      "step": 2838
+    },
+    {
+      "epoch": 51.62,
+      "learning_rate": 0.00017806896551724138,
+      "loss": 0.2811,
+      "step": 2839
+    },
+    {
+      "epoch": 51.64,
+      "learning_rate": 0.00017793103448275862,
+      "loss": 0.2728,
+      "step": 2840
+    },
+    {
+      "epoch": 51.65,
+      "learning_rate": 0.00017779310344827586,
+      "loss": 0.2524,
+      "step": 2841
+    },
+    {
+      "epoch": 51.67,
+      "learning_rate": 0.00017765517241379313,
+      "loss": 0.2948,
+      "step": 2842
+    },
+    {
+      "epoch": 51.69,
+      "learning_rate": 0.00017751724137931037,
+      "loss": 0.2976,
+      "step": 2843
+    },
+    {
+      "epoch": 51.71,
+      "learning_rate": 0.0001773793103448276,
+      "loss": 0.2589,
+      "step": 2844
+    },
+    {
+      "epoch": 51.73,
+      "learning_rate": 0.00017724137931034484,
+      "loss": 0.2912,
+      "step": 2845
+    },
+    {
+      "epoch": 51.75,
+      "learning_rate": 0.00017710344827586208,
+      "loss": 0.2673,
+      "step": 2846
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 0.00017696551724137932,
+      "loss": 0.3279,
+      "step": 2847
+    },
+    {
+      "epoch": 51.78,
+      "learning_rate": 0.00017682758620689656,
+      "loss": 0.2776,
+      "step": 2848
+    },
+    {
+      "epoch": 51.8,
+      "learning_rate": 0.0001766896551724138,
+      "loss": 0.2121,
+      "step": 2849
+    },
+    {
+      "epoch": 51.82,
+      "learning_rate": 0.000176551724137931,
+      "loss": 0.2531,
+      "step": 2850
+    },
+    {
+      "epoch": 51.84,
+      "learning_rate": 0.00017641379310344828,
+      "loss": 0.2867,
+      "step": 2851
+    },
+    {
+      "epoch": 51.85,
+      "learning_rate": 0.00017627586206896552,
+      "loss": 0.2485,
+      "step": 2852
+    },
+    {
+      "epoch": 51.87,
+      "learning_rate": 0.00017613793103448276,
+      "loss": 0.2193,
+      "step": 2853
+    },
+    {
+      "epoch": 51.89,
+      "learning_rate": 0.000176,
+      "loss": 0.2668,
+      "step": 2854
+    },
+    {
+      "epoch": 51.91,
+      "learning_rate": 0.00017586206896551723,
+      "loss": 0.2117,
+      "step": 2855
+    },
+    {
+      "epoch": 51.93,
+      "learning_rate": 0.00017572413793103447,
+      "loss": 0.2962,
+      "step": 2856
+    },
+    {
+      "epoch": 51.95,
+      "learning_rate": 0.0001755862068965517,
+      "loss": 0.2647,
+      "step": 2857
+    },
+    {
+      "epoch": 51.96,
+      "learning_rate": 0.00017544827586206895,
+      "loss": 0.324,
+      "step": 2858
+    },
+    {
+      "epoch": 51.98,
+      "learning_rate": 0.00017531034482758622,
+      "loss": 0.2298,
+      "step": 2859
+    },
+    {
+      "epoch": 52.0,
+      "learning_rate": 0.00017517241379310346,
+      "loss": 0.2197,
+      "step": 2860
+    },
+    {
+      "epoch": 52.02,
+      "learning_rate": 0.0001750344827586207,
+      "loss": 0.2421,
+      "step": 2861
+    },
+    {
+      "epoch": 52.04,
+      "learning_rate": 0.00017489655172413793,
+      "loss": 0.2305,
+      "step": 2862
+    },
+    {
+      "epoch": 52.05,
+      "learning_rate": 0.00017475862068965517,
+      "loss": 0.3222,
+      "step": 2863
+    },
+    {
+      "epoch": 52.07,
+      "learning_rate": 0.0001746206896551724,
+      "loss": 0.2563,
+      "step": 2864
+    },
+    {
+      "epoch": 52.09,
+      "learning_rate": 0.00017448275862068965,
+      "loss": 0.2654,
+      "step": 2865
+    },
+    {
+      "epoch": 52.11,
+      "learning_rate": 0.0001743448275862069,
+      "loss": 0.2317,
+      "step": 2866
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 0.00017420689655172413,
+      "loss": 0.2386,
+      "step": 2867
+    },
+    {
+      "epoch": 52.15,
+      "learning_rate": 0.0001740689655172414,
+      "loss": 0.2519,
+      "step": 2868
+    },
+    {
+      "epoch": 52.16,
+      "learning_rate": 0.00017393103448275863,
+      "loss": 0.2428,
+      "step": 2869
+    },
+    {
+      "epoch": 52.18,
+      "learning_rate": 0.00017379310344827587,
+      "loss": 0.2654,
+      "step": 2870
+    },
+    {
+      "epoch": 52.2,
+      "learning_rate": 0.0001736551724137931,
+      "loss": 0.2556,
+      "step": 2871
+    },
+    {
+      "epoch": 52.22,
+      "learning_rate": 0.00017351724137931035,
+      "loss": 0.2649,
+      "step": 2872
+    },
+    {
+      "epoch": 52.24,
+      "learning_rate": 0.0001733793103448276,
+      "loss": 0.2314,
+      "step": 2873
+    },
+    {
+      "epoch": 52.25,
+      "learning_rate": 0.00017324137931034483,
+      "loss": 0.3176,
+      "step": 2874
+    },
+    {
+      "epoch": 52.27,
+      "learning_rate": 0.00017310344827586207,
+      "loss": 0.3032,
+      "step": 2875
+    },
+    {
+      "epoch": 52.29,
+      "learning_rate": 0.00017296551724137933,
+      "loss": 0.3167,
+      "step": 2876
+    },
+    {
+      "epoch": 52.31,
+      "learning_rate": 0.00017282758620689657,
+      "loss": 0.2414,
+      "step": 2877
+    },
+    {
+      "epoch": 52.33,
+      "learning_rate": 0.0001726896551724138,
+      "loss": 0.2354,
+      "step": 2878
+    },
+    {
+      "epoch": 52.35,
+      "learning_rate": 0.00017255172413793105,
+      "loss": 0.1945,
+      "step": 2879
+    },
+    {
+      "epoch": 52.36,
+      "learning_rate": 0.0001724137931034483,
+      "loss": 0.287,
+      "step": 2880
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 0.00017227586206896553,
+      "loss": 0.2497,
+      "step": 2881
+    },
+    {
+      "epoch": 52.4,
+      "learning_rate": 0.00017213793103448277,
+      "loss": 0.3121,
+      "step": 2882
+    },
+    {
+      "epoch": 52.42,
+      "learning_rate": 0.00017199999999999998,
+      "loss": 0.2128,
+      "step": 2883
+    },
+    {
+      "epoch": 52.44,
+      "learning_rate": 0.00017186206896551722,
+      "loss": 0.2895,
+      "step": 2884
+    },
+    {
+      "epoch": 52.45,
+      "learning_rate": 0.00017172413793103448,
+      "loss": 0.2693,
+      "step": 2885
+    },
+    {
+      "epoch": 52.47,
+      "learning_rate": 0.00017158620689655172,
+      "loss": 0.3368,
+      "step": 2886
+    },
+    {
+      "epoch": 52.49,
+      "learning_rate": 0.00017144827586206896,
+      "loss": 0.2482,
+      "step": 2887
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 0.0001713103448275862,
+      "loss": 0.2458,
+      "step": 2888
+    },
+    {
+      "epoch": 52.53,
+      "learning_rate": 0.00017117241379310344,
+      "loss": 0.2409,
+      "step": 2889
+    },
+    {
+      "epoch": 52.55,
+      "learning_rate": 0.00017103448275862068,
+      "loss": 0.255,
+      "step": 2890
+    },
+    {
+      "epoch": 52.56,
+      "learning_rate": 0.00017089655172413792,
+      "loss": 0.2855,
+      "step": 2891
+    },
+    {
+      "epoch": 52.58,
+      "learning_rate": 0.00017075862068965516,
+      "loss": 0.245,
+      "step": 2892
+    },
+    {
+      "epoch": 52.6,
+      "learning_rate": 0.00017062068965517242,
+      "loss": 0.2683,
+      "step": 2893
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 0.00017048275862068966,
+      "loss": 0.2924,
+      "step": 2894
+    },
+    {
+      "epoch": 52.64,
+      "learning_rate": 0.0001703448275862069,
+      "loss": 0.2749,
+      "step": 2895
+    },
+    {
+      "epoch": 52.65,
+      "learning_rate": 0.00017020689655172414,
+      "loss": 0.2363,
+      "step": 2896
+    },
+    {
+      "epoch": 52.67,
+      "learning_rate": 0.00017006896551724138,
+      "loss": 0.2708,
+      "step": 2897
+    },
+    {
+      "epoch": 52.69,
+      "learning_rate": 0.00016993103448275862,
+      "loss": 0.2625,
+      "step": 2898
+    },
+    {
+      "epoch": 52.71,
+      "learning_rate": 0.00016979310344827586,
+      "loss": 0.2797,
+      "step": 2899
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 0.0001696551724137931,
+      "loss": 0.2494,
+      "step": 2900
+    },
+    {
+      "epoch": 52.75,
+      "learning_rate": 0.00016951724137931036,
+      "loss": 0.3327,
+      "step": 2901
+    },
+    {
+      "epoch": 52.76,
+      "learning_rate": 0.0001693793103448276,
+      "loss": 0.2805,
+      "step": 2902
+    },
+    {
+      "epoch": 52.78,
+      "learning_rate": 0.00016924137931034484,
+      "loss": 0.2721,
+      "step": 2903
+    },
+    {
+      "epoch": 52.8,
+      "learning_rate": 0.00016910344827586208,
+      "loss": 0.2557,
+      "step": 2904
+    },
+    {
+      "epoch": 52.82,
+      "learning_rate": 0.00016896551724137932,
+      "loss": 0.3024,
+      "step": 2905
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 0.00016882758620689656,
+      "loss": 0.2801,
+      "step": 2906
+    },
+    {
+      "epoch": 52.85,
+      "learning_rate": 0.0001686896551724138,
+      "loss": 0.2577,
+      "step": 2907
+    },
+    {
+      "epoch": 52.87,
+      "learning_rate": 0.00016855172413793103,
+      "loss": 0.2338,
+      "step": 2908
+    },
+    {
+      "epoch": 52.89,
+      "learning_rate": 0.00016841379310344827,
+      "loss": 0.2682,
+      "step": 2909
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 0.00016827586206896554,
+      "loss": 0.2258,
+      "step": 2910
+    },
+    {
+      "epoch": 52.93,
+      "learning_rate": 0.00016813793103448278,
+      "loss": 0.2445,
+      "step": 2911
+    },
+    {
+      "epoch": 52.95,
+      "learning_rate": 0.00016800000000000002,
+      "loss": 0.2999,
+      "step": 2912
+    },
+    {
+      "epoch": 52.96,
+      "learning_rate": 0.00016786206896551726,
+      "loss": 0.2794,
+      "step": 2913
+    },
+    {
+      "epoch": 52.98,
+      "learning_rate": 0.0001677241379310345,
+      "loss": 0.2222,
+      "step": 2914
+    },
+    {
+      "epoch": 53.0,
+      "learning_rate": 0.00016758620689655173,
+      "loss": 0.2733,
+      "step": 2915
+    },
+    {
+      "epoch": 53.02,
+      "learning_rate": 0.00016744827586206897,
+      "loss": 0.2847,
+      "step": 2916
+    },
+    {
+      "epoch": 53.04,
+      "learning_rate": 0.00016731034482758619,
+      "loss": 0.2697,
+      "step": 2917
+    },
+    {
+      "epoch": 53.05,
+      "learning_rate": 0.00016717241379310345,
+      "loss": 0.2112,
+      "step": 2918
+    },
+    {
+      "epoch": 53.07,
+      "learning_rate": 0.0001670344827586207,
+      "loss": 0.2694,
+      "step": 2919
+    },
+    {
+      "epoch": 53.09,
+      "learning_rate": 0.00016689655172413793,
+      "loss": 0.2233,
+      "step": 2920
+    },
+    {
+      "epoch": 53.11,
+      "learning_rate": 0.00016675862068965517,
+      "loss": 0.22,
+      "step": 2921
+    },
+    {
+      "epoch": 53.13,
+      "learning_rate": 0.0001666206896551724,
+      "loss": 0.2428,
+      "step": 2922
+    },
+    {
+      "epoch": 53.15,
+      "learning_rate": 0.00016648275862068965,
+      "loss": 0.2222,
+      "step": 2923
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 0.00016634482758620689,
+      "loss": 0.2787,
+      "step": 2924
+    },
+    {
+      "epoch": 53.18,
+      "learning_rate": 0.00016620689655172412,
+      "loss": 0.1935,
+      "step": 2925
+    },
+    {
+      "epoch": 53.2,
+      "learning_rate": 0.0001660689655172414,
+      "loss": 0.2146,
+      "step": 2926
+    },
+    {
+      "epoch": 53.22,
+      "learning_rate": 0.00016593103448275863,
+      "loss": 0.2237,
+      "step": 2927
+    },
+    {
+      "epoch": 53.24,
+      "learning_rate": 0.00016579310344827587,
+      "loss": 0.2261,
+      "step": 2928
+    },
+    {
+      "epoch": 53.25,
+      "learning_rate": 0.0001656551724137931,
+      "loss": 0.2363,
+      "step": 2929
+    },
+    {
+      "epoch": 53.27,
+      "learning_rate": 0.00016551724137931035,
+      "loss": 0.2678,
+      "step": 2930
+    },
+    {
+      "epoch": 53.29,
+      "learning_rate": 0.00016537931034482759,
+      "loss": 0.2544,
+      "step": 2931
+    },
+    {
+      "epoch": 53.31,
+      "learning_rate": 0.00016524137931034482,
+      "loss": 0.2557,
+      "step": 2932
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 0.00016510344827586206,
+      "loss": 0.2707,
+      "step": 2933
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 0.0001649655172413793,
+      "loss": 0.1716,
+      "step": 2934
+    },
+    {
+      "epoch": 53.36,
+      "learning_rate": 0.00016482758620689657,
+      "loss": 0.2276,
+      "step": 2935
+    },
+    {
+      "epoch": 53.38,
+      "learning_rate": 0.0001646896551724138,
+      "loss": 0.2602,
+      "step": 2936
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 0.00016455172413793105,
+      "loss": 0.2437,
+      "step": 2937
+    },
+    {
+      "epoch": 53.42,
+      "learning_rate": 0.00016441379310344829,
+      "loss": 0.1994,
+      "step": 2938
+    },
+    {
+      "epoch": 53.44,
+      "learning_rate": 0.00016427586206896552,
+      "loss": 0.3578,
+      "step": 2939
+    },
+    {
+      "epoch": 53.45,
+      "learning_rate": 0.00016413793103448276,
+      "loss": 0.2268,
+      "step": 2940
+    },
+    {
+      "epoch": 53.47,
+      "learning_rate": 0.000164,
+      "loss": 0.3532,
+      "step": 2941
+    },
+    {
+      "epoch": 53.49,
+      "learning_rate": 0.00016386206896551724,
+      "loss": 0.2589,
+      "step": 2942
+    },
+    {
+      "epoch": 53.51,
+      "learning_rate": 0.0001637241379310345,
+      "loss": 0.2354,
+      "step": 2943
+    },
+    {
+      "epoch": 53.53,
+      "learning_rate": 0.00016358620689655175,
+      "loss": 0.264,
+      "step": 2944
+    },
+    {
+      "epoch": 53.55,
+      "learning_rate": 0.00016344827586206898,
+      "loss": 0.2624,
+      "step": 2945
+    },
+    {
+      "epoch": 53.56,
+      "learning_rate": 0.00016331034482758622,
+      "loss": 0.2982,
+      "step": 2946
+    },
+    {
+      "epoch": 53.58,
+      "learning_rate": 0.00016317241379310346,
+      "loss": 0.2577,
+      "step": 2947
+    },
+    {
+      "epoch": 53.6,
+      "learning_rate": 0.0001630344827586207,
+      "loss": 0.2643,
+      "step": 2948
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 0.00016289655172413794,
+      "loss": 0.307,
+      "step": 2949
+    },
+    {
+      "epoch": 53.64,
+      "learning_rate": 0.00016275862068965515,
+      "loss": 0.2826,
+      "step": 2950
+    },
+    {
+      "epoch": 53.65,
+      "learning_rate": 0.0001626206896551724,
+      "loss": 0.2571,
+      "step": 2951
+    },
+    {
+      "epoch": 53.67,
+      "learning_rate": 0.00016248275862068966,
+      "loss": 0.2675,
+      "step": 2952
+    },
+    {
+      "epoch": 53.69,
+      "learning_rate": 0.0001623448275862069,
+      "loss": 0.2117,
+      "step": 2953
+    },
+    {
+      "epoch": 53.71,
+      "learning_rate": 0.00016220689655172414,
+      "loss": 0.2491,
+      "step": 2954
+    },
+    {
+      "epoch": 53.73,
+      "learning_rate": 0.00016206896551724137,
+      "loss": 0.2737,
+      "step": 2955
+    },
+    {
+      "epoch": 53.75,
+      "learning_rate": 0.0001619310344827586,
+      "loss": 0.224,
+      "step": 2956
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 0.00016179310344827585,
+      "loss": 0.2224,
+      "step": 2957
+    },
+    {
+      "epoch": 53.78,
+      "learning_rate": 0.0001616551724137931,
+      "loss": 0.2334,
+      "step": 2958
+    },
+    {
+      "epoch": 53.8,
+      "learning_rate": 0.00016151724137931033,
+      "loss": 0.2524,
+      "step": 2959
+    },
+    {
+      "epoch": 53.82,
+      "learning_rate": 0.0001613793103448276,
+      "loss": 0.2733,
+      "step": 2960
+    },
+    {
+      "epoch": 53.84,
+      "learning_rate": 0.00016124137931034484,
+      "loss": 0.2407,
+      "step": 2961
+    },
+    {
+      "epoch": 53.85,
+      "learning_rate": 0.00016110344827586207,
+      "loss": 0.2104,
+      "step": 2962
+    },
+    {
+      "epoch": 53.87,
+      "learning_rate": 0.0001609655172413793,
+      "loss": 0.2993,
+      "step": 2963
+    },
+    {
+      "epoch": 53.89,
+      "learning_rate": 0.00016082758620689655,
+      "loss": 0.2636,
+      "step": 2964
+    },
+    {
+      "epoch": 53.91,
+      "learning_rate": 0.0001606896551724138,
+      "loss": 0.1937,
+      "step": 2965
+    },
+    {
+      "epoch": 53.93,
+      "learning_rate": 0.00016055172413793103,
+      "loss": 0.2826,
+      "step": 2966
+    },
+    {
+      "epoch": 53.95,
+      "learning_rate": 0.00016041379310344827,
+      "loss": 0.2161,
+      "step": 2967
+    },
+    {
+      "epoch": 53.96,
+      "learning_rate": 0.00016027586206896554,
+      "loss": 0.1997,
+      "step": 2968
+    },
+    {
+      "epoch": 53.98,
+      "learning_rate": 0.00016013793103448277,
+      "loss": 0.2321,
+      "step": 2969
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 0.00016,
+      "loss": 0.3031,
+      "step": 2970
+    },
+    {
+      "epoch": 54.02,
+      "learning_rate": 0.00015986206896551725,
+      "loss": 0.3106,
+      "step": 2971
+    },
+    {
+      "epoch": 54.04,
+      "learning_rate": 0.0001597241379310345,
+      "loss": 0.3072,
+      "step": 2972
+    },
+    {
+      "epoch": 54.05,
+      "learning_rate": 0.00015958620689655173,
+      "loss": 0.2255,
+      "step": 2973
+    },
+    {
+      "epoch": 54.07,
+      "learning_rate": 0.00015944827586206897,
+      "loss": 0.2574,
+      "step": 2974
+    },
+    {
+      "epoch": 54.09,
+      "learning_rate": 0.0001593103448275862,
+      "loss": 0.2664,
+      "step": 2975
+    },
+    {
+      "epoch": 54.11,
+      "learning_rate": 0.00015917241379310345,
+      "loss": 0.196,
+      "step": 2976
+    },
+    {
+      "epoch": 54.13,
+      "learning_rate": 0.0001590344827586207,
+      "loss": 0.2774,
+      "step": 2977
+    },
+    {
+      "epoch": 54.15,
+      "learning_rate": 0.00015889655172413795,
+      "loss": 0.2398,
+      "step": 2978
+    },
+    {
+      "epoch": 54.16,
+      "learning_rate": 0.0001587586206896552,
+      "loss": 0.2076,
+      "step": 2979
+    },
+    {
+      "epoch": 54.18,
+      "learning_rate": 0.00015862068965517243,
+      "loss": 0.2772,
+      "step": 2980
+    },
+    {
+      "epoch": 54.2,
+      "learning_rate": 0.00015848275862068967,
+      "loss": 0.2173,
+      "step": 2981
+    },
+    {
+      "epoch": 54.22,
+      "learning_rate": 0.0001583448275862069,
+      "loss": 0.2291,
+      "step": 2982
+    },
+    {
+      "epoch": 54.24,
+      "learning_rate": 0.00015820689655172412,
+      "loss": 0.219,
+      "step": 2983
+    },
+    {
+      "epoch": 54.25,
+      "learning_rate": 0.00015806896551724136,
+      "loss": 0.1987,
+      "step": 2984
+    },
+    {
+      "epoch": 54.27,
+      "learning_rate": 0.00015793103448275862,
+      "loss": 0.2059,
+      "step": 2985
+    },
+    {
+      "epoch": 54.29,
+      "learning_rate": 0.00015779310344827586,
+      "loss": 0.205,
+      "step": 2986
+    },
+    {
+      "epoch": 54.31,
+      "learning_rate": 0.0001576551724137931,
+      "loss": 0.3156,
+      "step": 2987
+    },
+    {
+      "epoch": 54.33,
+      "learning_rate": 0.00015751724137931034,
+      "loss": 0.2635,
+      "step": 2988
+    },
+    {
+      "epoch": 54.35,
+      "learning_rate": 0.00015737931034482758,
+      "loss": 0.263,
+      "step": 2989
+    },
+    {
+      "epoch": 54.36,
+      "learning_rate": 0.00015724137931034482,
+      "loss": 0.2545,
+      "step": 2990
+    },
+    {
+      "epoch": 54.38,
+      "learning_rate": 0.00015710344827586206,
+      "loss": 0.1952,
+      "step": 2991
+    },
+    {
+      "epoch": 54.4,
+      "learning_rate": 0.0001569655172413793,
+      "loss": 0.2896,
+      "step": 2992
+    },
+    {
+      "epoch": 54.42,
+      "learning_rate": 0.00015682758620689654,
+      "loss": 0.3388,
+      "step": 2993
+    },
+    {
+      "epoch": 54.44,
+      "learning_rate": 0.0001566896551724138,
+      "loss": 0.2965,
+      "step": 2994
+    },
+    {
+      "epoch": 54.45,
+      "learning_rate": 0.00015655172413793104,
+      "loss": 0.1949,
+      "step": 2995
+    },
+    {
+      "epoch": 54.47,
+      "learning_rate": 0.00015641379310344828,
+      "loss": 0.2219,
+      "step": 2996
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 0.00015627586206896552,
+      "loss": 0.2136,
+      "step": 2997
+    },
+    {
+      "epoch": 54.51,
+      "learning_rate": 0.00015613793103448276,
+      "loss": 0.2213,
+      "step": 2998
+    },
+    {
+      "epoch": 54.53,
+      "learning_rate": 0.000156,
+      "loss": 0.278,
+      "step": 2999
+    },
+    {
+      "epoch": 54.55,
+      "learning_rate": 0.00015586206896551724,
+      "loss": 0.2656,
+      "step": 3000
+    },
+    {
+      "epoch": 54.55,
+      "eval_cer": 0.08520367621703474,
+      "eval_loss": 0.38656094670295715,
+      "eval_runtime": 13.903,
+      "eval_samples_per_second": 118.463,
+      "eval_steps_per_second": 1.87,
+      "eval_wer": 0.34633847410887547,
+      "step": 3000
+    },
+    {
+      "epoch": 54.56,
+      "learning_rate": 0.00015572413793103448,
+      "loss": 0.2669,
+      "step": 3001
+    },
+    {
+      "epoch": 54.58,
+      "learning_rate": 0.00015558620689655174,
+      "loss": 0.2653,
+      "step": 3002
+    },
+    {
+      "epoch": 54.6,
+      "learning_rate": 0.00015544827586206898,
+      "loss": 0.2757,
+      "step": 3003
+    },
+    {
+      "epoch": 54.62,
+      "learning_rate": 0.00015531034482758622,
+      "loss": 0.2332,
+      "step": 3004
+    },
+    {
+      "epoch": 54.64,
+      "learning_rate": 0.00015517241379310346,
+      "loss": 0.2369,
+      "step": 3005
+    },
+    {
+      "epoch": 54.65,
+      "learning_rate": 0.0001550344827586207,
+      "loss": 0.2274,
+      "step": 3006
+    },
+    {
+      "epoch": 54.67,
+      "learning_rate": 0.00015489655172413794,
+      "loss": 0.2069,
+      "step": 3007
+    },
+    {
+      "epoch": 54.69,
+      "learning_rate": 0.00015475862068965517,
+      "loss": 0.2005,
+      "step": 3008
+    },
+    {
+      "epoch": 54.71,
+      "learning_rate": 0.00015462068965517241,
+      "loss": 0.2464,
+      "step": 3009
+    },
+    {
+      "epoch": 54.73,
+      "learning_rate": 0.00015448275862068968,
+      "loss": 0.2721,
+      "step": 3010
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 0.00015434482758620692,
+      "loss": 0.3104,
+      "step": 3011
+    },
+    {
+      "epoch": 54.76,
+      "learning_rate": 0.00015420689655172416,
+      "loss": 0.1952,
+      "step": 3012
+    },
+    {
+      "epoch": 54.78,
+      "learning_rate": 0.0001540689655172414,
+      "loss": 0.2726,
+      "step": 3013
+    },
+    {
+      "epoch": 54.8,
+      "learning_rate": 0.00015393103448275864,
+      "loss": 0.2891,
+      "step": 3014
+    },
+    {
+      "epoch": 54.82,
+      "learning_rate": 0.00015379310344827587,
+      "loss": 0.2521,
+      "step": 3015
+    },
+    {
+      "epoch": 54.84,
+      "learning_rate": 0.0001536551724137931,
+      "loss": 0.2658,
+      "step": 3016
+    },
+    {
+      "epoch": 54.85,
+      "learning_rate": 0.00015351724137931033,
+      "loss": 0.2631,
+      "step": 3017
+    },
+    {
+      "epoch": 54.87,
+      "learning_rate": 0.00015337931034482756,
+      "loss": 0.233,
+      "step": 3018
+    },
+    {
+      "epoch": 54.89,
+      "learning_rate": 0.00015324137931034483,
+      "loss": 0.2536,
+      "step": 3019
+    },
+    {
+      "epoch": 54.91,
+      "learning_rate": 0.00015310344827586207,
+      "loss": 0.1797,
+      "step": 3020
+    },
+    {
+      "epoch": 54.93,
+      "learning_rate": 0.0001529655172413793,
+      "loss": 0.2177,
+      "step": 3021
+    },
+    {
+      "epoch": 54.95,
+      "learning_rate": 0.00015282758620689655,
+      "loss": 0.2424,
+      "step": 3022
+    },
+    {
+      "epoch": 54.96,
+      "learning_rate": 0.0001526896551724138,
+      "loss": 0.2075,
+      "step": 3023
+    },
+    {
+      "epoch": 54.98,
+      "learning_rate": 0.00015255172413793103,
+      "loss": 0.2048,
+      "step": 3024
+    },
+    {
+      "epoch": 55.0,
+      "learning_rate": 0.00015241379310344826,
+      "loss": 0.2366,
+      "step": 3025
+    },
+    {
+      "epoch": 55.02,
+      "learning_rate": 0.0001522758620689655,
+      "loss": 0.2475,
+      "step": 3026
+    },
+    {
+      "epoch": 55.04,
+      "learning_rate": 0.00015213793103448277,
+      "loss": 0.3022,
+      "step": 3027
+    },
+    {
+      "epoch": 55.05,
+      "learning_rate": 0.000152,
+      "loss": 0.2214,
+      "step": 3028
+    },
+    {
+      "epoch": 55.07,
+      "learning_rate": 0.00015186206896551725,
+      "loss": 0.2353,
+      "step": 3029
+    },
+    {
+      "epoch": 55.09,
+      "learning_rate": 0.00015172413793103449,
+      "loss": 0.2801,
+      "step": 3030
+    },
+    {
+      "epoch": 55.11,
+      "learning_rate": 0.00015158620689655173,
+      "loss": 0.2285,
+      "step": 3031
+    },
+    {
+      "epoch": 55.13,
+      "learning_rate": 0.00015144827586206896,
+      "loss": 0.3266,
+      "step": 3032
+    },
+    {
+      "epoch": 55.15,
+      "learning_rate": 0.0001513103448275862,
+      "loss": 0.2205,
+      "step": 3033
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 0.00015117241379310344,
+      "loss": 0.2276,
+      "step": 3034
+    },
+    {
+      "epoch": 55.18,
+      "learning_rate": 0.0001510344827586207,
+      "loss": 0.2301,
+      "step": 3035
+    },
+    {
+      "epoch": 55.2,
+      "learning_rate": 0.00015089655172413795,
+      "loss": 0.1842,
+      "step": 3036
+    },
+    {
+      "epoch": 55.22,
+      "learning_rate": 0.00015075862068965519,
+      "loss": 0.2967,
+      "step": 3037
+    },
+    {
+      "epoch": 55.24,
+      "learning_rate": 0.00015062068965517243,
+      "loss": 0.2822,
+      "step": 3038
+    },
+    {
+      "epoch": 55.25,
+      "learning_rate": 0.00015048275862068966,
+      "loss": 0.261,
+      "step": 3039
+    },
+    {
+      "epoch": 55.27,
+      "learning_rate": 0.0001503448275862069,
+      "loss": 0.1849,
+      "step": 3040
+    },
+    {
+      "epoch": 55.29,
+      "learning_rate": 0.00015020689655172414,
+      "loss": 0.269,
+      "step": 3041
+    },
+    {
+      "epoch": 55.31,
+      "learning_rate": 0.00015006896551724138,
+      "loss": 0.2709,
+      "step": 3042
+    },
+    {
+      "epoch": 55.33,
+      "learning_rate": 0.00014993103448275862,
+      "loss": 0.2701,
+      "step": 3043
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 0.00014979310344827589,
+      "loss": 0.253,
+      "step": 3044
+    },
+    {
+      "epoch": 55.36,
+      "learning_rate": 0.00014965517241379312,
+      "loss": 0.2363,
+      "step": 3045
+    },
+    {
+      "epoch": 55.38,
+      "learning_rate": 0.00014951724137931036,
+      "loss": 0.2648,
+      "step": 3046
+    },
+    {
+      "epoch": 55.4,
+      "learning_rate": 0.0001493793103448276,
+      "loss": 0.2108,
+      "step": 3047
+    },
+    {
+      "epoch": 55.42,
+      "learning_rate": 0.00014924137931034484,
+      "loss": 0.2391,
+      "step": 3048
+    },
+    {
+      "epoch": 55.44,
+      "learning_rate": 0.00014910344827586208,
+      "loss": 0.2303,
+      "step": 3049
+    },
+    {
+      "epoch": 55.45,
+      "learning_rate": 0.0001489655172413793,
+      "loss": 0.2268,
+      "step": 3050
+    },
+    {
+      "epoch": 55.47,
+      "learning_rate": 0.00014882758620689653,
+      "loss": 0.2693,
+      "step": 3051
+    },
+    {
+      "epoch": 55.49,
+      "learning_rate": 0.0001486896551724138,
+      "loss": 0.2415,
+      "step": 3052
+    },
+    {
+      "epoch": 55.51,
+      "learning_rate": 0.00014855172413793104,
+      "loss": 0.221,
+      "step": 3053
+    },
+    {
+      "epoch": 55.53,
+      "learning_rate": 0.00014841379310344828,
+      "loss": 0.2663,
+      "step": 3054
+    },
+    {
+      "epoch": 55.55,
+      "learning_rate": 0.00014827586206896551,
+      "loss": 0.2553,
+      "step": 3055
+    },
+    {
+      "epoch": 55.56,
+      "learning_rate": 0.00014813793103448275,
+      "loss": 0.1906,
+      "step": 3056
+    },
+    {
+      "epoch": 55.58,
+      "learning_rate": 0.000148,
+      "loss": 0.2691,
+      "step": 3057
+    },
+    {
+      "epoch": 55.6,
+      "learning_rate": 0.00014786206896551723,
+      "loss": 0.2739,
+      "step": 3058
+    },
+    {
+      "epoch": 55.62,
+      "learning_rate": 0.00014772413793103447,
+      "loss": 0.2214,
+      "step": 3059
+    },
+    {
+      "epoch": 55.64,
+      "learning_rate": 0.0001475862068965517,
+      "loss": 0.2508,
+      "step": 3060
+    },
+    {
+      "epoch": 55.65,
+      "learning_rate": 0.00014744827586206898,
+      "loss": 0.2354,
+      "step": 3061
+    },
+    {
+      "epoch": 55.67,
+      "learning_rate": 0.00014731034482758621,
+      "loss": 0.2167,
+      "step": 3062
+    },
+    {
+      "epoch": 55.69,
+      "learning_rate": 0.00014717241379310345,
+      "loss": 0.2426,
+      "step": 3063
+    },
+    {
+      "epoch": 55.71,
+      "learning_rate": 0.0001470344827586207,
+      "loss": 0.302,
+      "step": 3064
+    },
+    {
+      "epoch": 55.73,
+      "learning_rate": 0.00014689655172413793,
+      "loss": 0.2579,
+      "step": 3065
+    },
+    {
+      "epoch": 55.75,
+      "learning_rate": 0.00014675862068965517,
+      "loss": 0.2844,
+      "step": 3066
+    },
+    {
+      "epoch": 55.76,
+      "learning_rate": 0.0001466206896551724,
+      "loss": 0.2509,
+      "step": 3067
+    },
+    {
+      "epoch": 55.78,
+      "learning_rate": 0.00014648275862068965,
+      "loss": 0.2774,
+      "step": 3068
+    },
+    {
+      "epoch": 55.8,
+      "learning_rate": 0.00014634482758620691,
+      "loss": 0.2195,
+      "step": 3069
+    },
+    {
+      "epoch": 55.82,
+      "learning_rate": 0.00014620689655172415,
+      "loss": 0.3007,
+      "step": 3070
+    },
+    {
+      "epoch": 55.84,
+      "learning_rate": 0.0001460689655172414,
+      "loss": 0.2038,
+      "step": 3071
+    },
+    {
+      "epoch": 55.85,
+      "learning_rate": 0.00014593103448275863,
+      "loss": 0.2246,
+      "step": 3072
+    },
+    {
+      "epoch": 55.87,
+      "learning_rate": 0.00014579310344827587,
+      "loss": 0.2386,
+      "step": 3073
+    },
+    {
+      "epoch": 55.89,
+      "learning_rate": 0.0001456551724137931,
+      "loss": 0.2159,
+      "step": 3074
+    },
+    {
+      "epoch": 55.91,
+      "learning_rate": 0.00014551724137931035,
+      "loss": 0.2163,
+      "step": 3075
+    },
+    {
+      "epoch": 55.93,
+      "learning_rate": 0.0001453793103448276,
+      "loss": 0.2607,
+      "step": 3076
+    },
+    {
+      "epoch": 55.95,
+      "learning_rate": 0.00014524137931034485,
+      "loss": 0.2773,
+      "step": 3077
+    },
+    {
+      "epoch": 55.96,
+      "learning_rate": 0.0001451034482758621,
+      "loss": 0.2941,
+      "step": 3078
+    },
+    {
+      "epoch": 55.98,
+      "learning_rate": 0.00014496551724137933,
+      "loss": 0.2414,
+      "step": 3079
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 0.00014482758620689657,
+      "loss": 0.203,
+      "step": 3080
+    },
+    {
+      "epoch": 56.02,
+      "learning_rate": 0.0001446896551724138,
+      "loss": 0.2874,
+      "step": 3081
+    },
+    {
+      "epoch": 56.04,
+      "learning_rate": 0.00014455172413793105,
+      "loss": 0.1982,
+      "step": 3082
+    },
+    {
+      "epoch": 56.05,
+      "learning_rate": 0.00014441379310344826,
+      "loss": 0.1882,
+      "step": 3083
+    },
+    {
+      "epoch": 56.07,
+      "learning_rate": 0.0001442758620689655,
+      "loss": 0.262,
+      "step": 3084
+    },
+    {
+      "epoch": 56.09,
+      "learning_rate": 0.00014413793103448274,
+      "loss": 0.2217,
+      "step": 3085
+    },
+    {
+      "epoch": 56.11,
+      "learning_rate": 0.000144,
+      "loss": 0.2437,
+      "step": 3086
+    },
+    {
+      "epoch": 56.13,
+      "learning_rate": 0.00014386206896551724,
+      "loss": 0.2866,
+      "step": 3087
+    },
+    {
+      "epoch": 56.15,
+      "learning_rate": 0.00014372413793103448,
+      "loss": 0.2221,
+      "step": 3088
+    },
+    {
+      "epoch": 56.16,
+      "learning_rate": 0.00014358620689655172,
+      "loss": 0.2717,
+      "step": 3089
+    },
+    {
+      "epoch": 56.18,
+      "learning_rate": 0.00014344827586206896,
+      "loss": 0.2944,
+      "step": 3090
+    },
+    {
+      "epoch": 56.2,
+      "learning_rate": 0.0001433103448275862,
+      "loss": 0.2524,
+      "step": 3091
+    },
+    {
+      "epoch": 56.22,
+      "learning_rate": 0.00014317241379310344,
+      "loss": 0.2097,
+      "step": 3092
+    },
+    {
+      "epoch": 56.24,
+      "learning_rate": 0.00014303448275862068,
+      "loss": 0.2573,
+      "step": 3093
+    },
+    {
+      "epoch": 56.25,
+      "learning_rate": 0.00014289655172413794,
+      "loss": 0.2266,
+      "step": 3094
+    },
+    {
+      "epoch": 56.27,
+      "learning_rate": 0.00014275862068965518,
+      "loss": 0.2102,
+      "step": 3095
+    },
+    {
+      "epoch": 56.29,
+      "learning_rate": 0.00014262068965517242,
+      "loss": 0.282,
+      "step": 3096
+    },
+    {
+      "epoch": 56.31,
+      "learning_rate": 0.00014248275862068966,
+      "loss": 0.2325,
+      "step": 3097
+    },
+    {
+      "epoch": 56.33,
+      "learning_rate": 0.0001423448275862069,
+      "loss": 0.2502,
+      "step": 3098
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 0.00014220689655172414,
+      "loss": 0.1979,
+      "step": 3099
+    },
+    {
+      "epoch": 56.36,
+      "learning_rate": 0.00014206896551724138,
+      "loss": 0.2528,
+      "step": 3100
+    },
+    {
+      "epoch": 56.38,
+      "learning_rate": 0.00014193103448275862,
+      "loss": 0.2106,
+      "step": 3101
+    },
+    {
+      "epoch": 56.4,
+      "learning_rate": 0.00014179310344827585,
+      "loss": 0.227,
+      "step": 3102
+    },
+    {
+      "epoch": 56.42,
+      "learning_rate": 0.00014165517241379312,
+      "loss": 0.2224,
+      "step": 3103
+    },
+    {
+      "epoch": 56.44,
+      "learning_rate": 0.00014151724137931036,
+      "loss": 0.2495,
+      "step": 3104
+    },
+    {
+      "epoch": 56.45,
+      "learning_rate": 0.0001413793103448276,
+      "loss": 0.2079,
+      "step": 3105
+    },
+    {
+      "epoch": 56.47,
+      "learning_rate": 0.00014124137931034484,
+      "loss": 0.2666,
+      "step": 3106
+    },
+    {
+      "epoch": 56.49,
+      "learning_rate": 0.00014110344827586208,
+      "loss": 0.1901,
+      "step": 3107
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 0.00014096551724137932,
+      "loss": 0.2122,
+      "step": 3108
+    },
+    {
+      "epoch": 56.53,
+      "learning_rate": 0.00014082758620689655,
+      "loss": 0.2197,
+      "step": 3109
+    },
+    {
+      "epoch": 56.55,
+      "learning_rate": 0.0001406896551724138,
+      "loss": 0.1746,
+      "step": 3110
+    },
+    {
+      "epoch": 56.56,
+      "learning_rate": 0.00014055172413793106,
+      "loss": 0.186,
+      "step": 3111
+    },
+    {
+      "epoch": 56.58,
+      "learning_rate": 0.0001404137931034483,
+      "loss": 0.2532,
+      "step": 3112
+    },
+    {
+      "epoch": 56.6,
+      "learning_rate": 0.00014027586206896554,
+      "loss": 0.2201,
+      "step": 3113
+    },
+    {
+      "epoch": 56.62,
+      "learning_rate": 0.00014013793103448278,
+      "loss": 0.2085,
+      "step": 3114
+    },
+    {
+      "epoch": 56.64,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 0.1916,
+      "step": 3115
+    },
+    {
+      "epoch": 56.65,
+      "learning_rate": 0.00013986206896551723,
+      "loss": 0.2389,
+      "step": 3116
+    },
+    {
+      "epoch": 56.67,
+      "learning_rate": 0.00013972413793103447,
+      "loss": 0.2894,
+      "step": 3117
+    },
+    {
+      "epoch": 56.69,
+      "learning_rate": 0.0001395862068965517,
+      "loss": 0.1893,
+      "step": 3118
+    },
+    {
+      "epoch": 56.71,
+      "learning_rate": 0.00013944827586206897,
+      "loss": 0.198,
+      "step": 3119
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 0.0001393103448275862,
+      "loss": 0.2253,
+      "step": 3120
+    },
+    {
+      "epoch": 56.75,
+      "learning_rate": 0.00013917241379310345,
+      "loss": 0.2851,
+      "step": 3121
+    },
+    {
+      "epoch": 56.76,
+      "learning_rate": 0.0001390344827586207,
+      "loss": 0.2682,
+      "step": 3122
+    },
+    {
+      "epoch": 56.78,
+      "learning_rate": 0.00013889655172413793,
+      "loss": 0.235,
+      "step": 3123
+    },
+    {
+      "epoch": 56.8,
+      "learning_rate": 0.00013875862068965517,
+      "loss": 0.2141,
+      "step": 3124
+    },
+    {
+      "epoch": 56.82,
+      "learning_rate": 0.0001386206896551724,
+      "loss": 0.227,
+      "step": 3125
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 0.00013848275862068964,
+      "loss": 0.2497,
+      "step": 3126
+    },
+    {
+      "epoch": 56.85,
+      "learning_rate": 0.00013834482758620688,
+      "loss": 0.1802,
+      "step": 3127
+    },
+    {
+      "epoch": 56.87,
+      "learning_rate": 0.00013820689655172415,
+      "loss": 0.2156,
+      "step": 3128
+    },
+    {
+      "epoch": 56.89,
+      "learning_rate": 0.0001380689655172414,
+      "loss": 0.2401,
+      "step": 3129
+    },
+    {
+      "epoch": 56.91,
+      "learning_rate": 0.00013793103448275863,
+      "loss": 0.1642,
+      "step": 3130
+    },
+    {
+      "epoch": 56.93,
+      "learning_rate": 0.00013779310344827587,
+      "loss": 0.2664,
+      "step": 3131
+    },
+    {
+      "epoch": 56.95,
+      "learning_rate": 0.0001376551724137931,
+      "loss": 0.2858,
+      "step": 3132
+    },
+    {
+      "epoch": 56.96,
+      "learning_rate": 0.00013751724137931034,
+      "loss": 0.2887,
+      "step": 3133
+    },
+    {
+      "epoch": 56.98,
+      "learning_rate": 0.00013737931034482758,
+      "loss": 0.2428,
+      "step": 3134
+    },
+    {
+      "epoch": 57.0,
+      "learning_rate": 0.00013724137931034482,
+      "loss": 0.2074,
+      "step": 3135
+    },
+    {
+      "epoch": 57.02,
+      "learning_rate": 0.0001371034482758621,
+      "loss": 0.2995,
+      "step": 3136
+    },
+    {
+      "epoch": 57.04,
+      "learning_rate": 0.00013696551724137933,
+      "loss": 0.2333,
+      "step": 3137
+    },
+    {
+      "epoch": 57.05,
+      "learning_rate": 0.00013682758620689657,
+      "loss": 0.2846,
+      "step": 3138
+    },
+    {
+      "epoch": 57.07,
+      "learning_rate": 0.0001366896551724138,
+      "loss": 0.2124,
+      "step": 3139
+    },
+    {
+      "epoch": 57.09,
+      "learning_rate": 0.00013655172413793104,
+      "loss": 0.2662,
+      "step": 3140
+    },
+    {
+      "epoch": 57.11,
+      "learning_rate": 0.00013641379310344828,
+      "loss": 0.2092,
+      "step": 3141
+    },
+    {
+      "epoch": 57.13,
+      "learning_rate": 0.00013627586206896552,
+      "loss": 0.2204,
+      "step": 3142
+    },
+    {
+      "epoch": 57.15,
+      "learning_rate": 0.00013613793103448276,
+      "loss": 0.2023,
+      "step": 3143
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 0.00013600000000000003,
+      "loss": 0.2417,
+      "step": 3144
+    },
+    {
+      "epoch": 57.18,
+      "learning_rate": 0.00013586206896551726,
+      "loss": 0.2666,
+      "step": 3145
+    },
+    {
+      "epoch": 57.2,
+      "learning_rate": 0.0001357241379310345,
+      "loss": 0.2299,
+      "step": 3146
+    },
+    {
+      "epoch": 57.22,
+      "learning_rate": 0.00013558620689655174,
+      "loss": 0.1603,
+      "step": 3147
+    },
+    {
+      "epoch": 57.24,
+      "learning_rate": 0.00013544827586206898,
+      "loss": 0.2744,
+      "step": 3148
+    },
+    {
+      "epoch": 57.25,
+      "learning_rate": 0.0001353103448275862,
+      "loss": 0.2701,
+      "step": 3149
+    },
+    {
+      "epoch": 57.27,
+      "learning_rate": 0.00013517241379310343,
+      "loss": 0.2706,
+      "step": 3150
+    },
+    {
+      "epoch": 57.29,
+      "learning_rate": 0.00013503448275862067,
+      "loss": 0.1794,
+      "step": 3151
+    },
+    {
+      "epoch": 57.31,
+      "learning_rate": 0.0001348965517241379,
+      "loss": 0.2525,
+      "step": 3152
+    },
+    {
+      "epoch": 57.33,
+      "learning_rate": 0.00013475862068965518,
+      "loss": 0.2744,
+      "step": 3153
+    },
+    {
+      "epoch": 57.35,
+      "learning_rate": 0.00013462068965517242,
+      "loss": 0.2052,
+      "step": 3154
+    },
+    {
+      "epoch": 57.36,
+      "learning_rate": 0.00013448275862068965,
+      "loss": 0.2275,
+      "step": 3155
+    },
+    {
+      "epoch": 57.38,
+      "learning_rate": 0.0001343448275862069,
+      "loss": 0.2157,
+      "step": 3156
+    },
+    {
+      "epoch": 57.4,
+      "learning_rate": 0.00013420689655172413,
+      "loss": 0.2101,
+      "step": 3157
+    },
+    {
+      "epoch": 57.42,
+      "learning_rate": 0.00013406896551724137,
+      "loss": 0.2746,
+      "step": 3158
+    },
+    {
+      "epoch": 57.44,
+      "learning_rate": 0.0001339310344827586,
+      "loss": 0.1846,
+      "step": 3159
+    },
+    {
+      "epoch": 57.45,
+      "learning_rate": 0.00013379310344827585,
+      "loss": 0.2845,
+      "step": 3160
+    },
+    {
+      "epoch": 57.47,
+      "learning_rate": 0.00013365517241379312,
+      "loss": 0.2673,
+      "step": 3161
+    },
+    {
+      "epoch": 57.49,
+      "learning_rate": 0.00013351724137931035,
+      "loss": 0.2494,
+      "step": 3162
+    },
+    {
+      "epoch": 57.51,
+      "learning_rate": 0.0001333793103448276,
+      "loss": 0.2058,
+      "step": 3163
+    },
+    {
+      "epoch": 57.53,
+      "learning_rate": 0.00013324137931034483,
+      "loss": 0.1957,
+      "step": 3164
+    },
+    {
+      "epoch": 57.55,
+      "learning_rate": 0.00013310344827586207,
+      "loss": 0.2092,
+      "step": 3165
+    },
+    {
+      "epoch": 57.56,
+      "learning_rate": 0.0001329655172413793,
+      "loss": 0.1845,
+      "step": 3166
+    },
+    {
+      "epoch": 57.58,
+      "learning_rate": 0.00013282758620689655,
+      "loss": 0.2466,
+      "step": 3167
+    },
+    {
+      "epoch": 57.6,
+      "learning_rate": 0.0001326896551724138,
+      "loss": 0.2152,
+      "step": 3168
+    },
+    {
+      "epoch": 57.62,
+      "learning_rate": 0.00013255172413793103,
+      "loss": 0.2169,
+      "step": 3169
+    },
+    {
+      "epoch": 57.64,
+      "learning_rate": 0.0001324137931034483,
+      "loss": 0.2395,
+      "step": 3170
+    },
+    {
+      "epoch": 57.65,
+      "learning_rate": 0.00013227586206896553,
+      "loss": 0.2855,
+      "step": 3171
+    },
+    {
+      "epoch": 57.67,
+      "learning_rate": 0.00013213793103448277,
+      "loss": 0.2614,
+      "step": 3172
+    },
+    {
+      "epoch": 57.69,
+      "learning_rate": 0.000132,
+      "loss": 0.2157,
+      "step": 3173
+    },
+    {
+      "epoch": 57.71,
+      "learning_rate": 0.00013186206896551725,
+      "loss": 0.3002,
+      "step": 3174
+    },
+    {
+      "epoch": 57.73,
+      "learning_rate": 0.0001317241379310345,
+      "loss": 0.1717,
+      "step": 3175
+    },
+    {
+      "epoch": 57.75,
+      "learning_rate": 0.00013158620689655173,
+      "loss": 0.2288,
+      "step": 3176
+    },
+    {
+      "epoch": 57.76,
+      "learning_rate": 0.00013144827586206897,
+      "loss": 0.273,
+      "step": 3177
+    },
+    {
+      "epoch": 57.78,
+      "learning_rate": 0.00013131034482758623,
+      "loss": 0.2096,
+      "step": 3178
+    },
+    {
+      "epoch": 57.8,
+      "learning_rate": 0.00013117241379310347,
+      "loss": 0.2259,
+      "step": 3179
+    },
+    {
+      "epoch": 57.82,
+      "learning_rate": 0.0001310344827586207,
+      "loss": 0.2052,
+      "step": 3180
+    },
+    {
+      "epoch": 57.84,
+      "learning_rate": 0.00013089655172413795,
+      "loss": 0.2147,
+      "step": 3181
+    },
+    {
+      "epoch": 57.85,
+      "learning_rate": 0.0001307586206896552,
+      "loss": 0.2312,
+      "step": 3182
+    },
+    {
+      "epoch": 57.87,
+      "learning_rate": 0.0001306206896551724,
+      "loss": 0.1771,
+      "step": 3183
+    },
+    {
+      "epoch": 57.89,
+      "learning_rate": 0.00013048275862068964,
+      "loss": 0.2253,
+      "step": 3184
+    },
+    {
+      "epoch": 57.91,
+      "learning_rate": 0.00013034482758620688,
+      "loss": 0.1932,
+      "step": 3185
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 0.00013020689655172412,
+      "loss": 0.3248,
+      "step": 3186
+    },
+    {
+      "epoch": 57.95,
+      "learning_rate": 0.00013006896551724138,
+      "loss": 0.2859,
+      "step": 3187
+    },
+    {
+      "epoch": 57.96,
+      "learning_rate": 0.00012993103448275862,
+      "loss": 0.2212,
+      "step": 3188
+    },
+    {
+      "epoch": 57.98,
+      "learning_rate": 0.00012979310344827586,
+      "loss": 0.2319,
+      "step": 3189
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 0.0001296551724137931,
+      "loss": 0.2228,
+      "step": 3190
+    },
+    {
+      "epoch": 58.02,
+      "learning_rate": 0.00012951724137931034,
+      "loss": 0.2411,
+      "step": 3191
+    },
+    {
+      "epoch": 58.04,
+      "learning_rate": 0.00012937931034482758,
+      "loss": 0.2477,
+      "step": 3192
+    },
+    {
+      "epoch": 58.05,
+      "learning_rate": 0.00012924137931034482,
+      "loss": 0.2145,
+      "step": 3193
+    },
+    {
+      "epoch": 58.07,
+      "learning_rate": 0.00012910344827586206,
+      "loss": 0.2449,
+      "step": 3194
+    },
+    {
+      "epoch": 58.09,
+      "learning_rate": 0.00012896551724137932,
+      "loss": 0.1985,
+      "step": 3195
+    },
+    {
+      "epoch": 58.11,
+      "learning_rate": 0.00012882758620689656,
+      "loss": 0.2658,
+      "step": 3196
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 0.0001286896551724138,
+      "loss": 0.2199,
+      "step": 3197
+    },
+    {
+      "epoch": 58.15,
+      "learning_rate": 0.00012855172413793104,
+      "loss": 0.2154,
+      "step": 3198
+    },
+    {
+      "epoch": 58.16,
+      "learning_rate": 0.00012841379310344828,
+      "loss": 0.2261,
+      "step": 3199
+    },
+    {
+      "epoch": 58.18,
+      "learning_rate": 0.00012827586206896552,
+      "loss": 0.2064,
+      "step": 3200
+    },
+    {
+      "epoch": 58.2,
+      "learning_rate": 0.00012813793103448276,
+      "loss": 0.2331,
+      "step": 3201
+    },
+    {
+      "epoch": 58.22,
+      "learning_rate": 0.000128,
+      "loss": 0.2215,
+      "step": 3202
+    },
+    {
+      "epoch": 58.24,
+      "learning_rate": 0.00012786206896551726,
+      "loss": 0.2223,
+      "step": 3203
+    },
+    {
+      "epoch": 58.25,
+      "learning_rate": 0.0001277241379310345,
+      "loss": 0.2282,
+      "step": 3204
+    },
+    {
+      "epoch": 58.27,
+      "learning_rate": 0.00012758620689655174,
+      "loss": 0.2655,
+      "step": 3205
+    },
+    {
+      "epoch": 58.29,
+      "learning_rate": 0.00012744827586206898,
+      "loss": 0.255,
+      "step": 3206
+    },
+    {
+      "epoch": 58.31,
+      "learning_rate": 0.00012731034482758622,
+      "loss": 0.2787,
+      "step": 3207
+    },
+    {
+      "epoch": 58.33,
+      "learning_rate": 0.00012717241379310346,
+      "loss": 0.2453,
+      "step": 3208
+    },
+    {
+      "epoch": 58.35,
+      "learning_rate": 0.0001270344827586207,
+      "loss": 0.2099,
+      "step": 3209
+    },
+    {
+      "epoch": 58.36,
+      "learning_rate": 0.00012689655172413793,
+      "loss": 0.2477,
+      "step": 3210
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 0.00012675862068965517,
+      "loss": 0.2205,
+      "step": 3211
+    },
+    {
+      "epoch": 58.4,
+      "learning_rate": 0.00012662068965517244,
+      "loss": 0.2259,
+      "step": 3212
+    },
+    {
+      "epoch": 58.42,
+      "learning_rate": 0.00012648275862068968,
+      "loss": 0.1965,
+      "step": 3213
+    },
+    {
+      "epoch": 58.44,
+      "learning_rate": 0.00012634482758620692,
+      "loss": 0.2119,
+      "step": 3214
+    },
+    {
+      "epoch": 58.45,
+      "learning_rate": 0.00012620689655172415,
+      "loss": 0.249,
+      "step": 3215
+    },
+    {
+      "epoch": 58.47,
+      "learning_rate": 0.00012606896551724137,
+      "loss": 0.2571,
+      "step": 3216
+    },
+    {
+      "epoch": 58.49,
+      "learning_rate": 0.0001259310344827586,
+      "loss": 0.2234,
+      "step": 3217
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 0.00012579310344827584,
+      "loss": 0.2552,
+      "step": 3218
+    },
+    {
+      "epoch": 58.53,
+      "learning_rate": 0.00012565517241379308,
+      "loss": 0.2739,
+      "step": 3219
+    },
+    {
+      "epoch": 58.55,
+      "learning_rate": 0.00012551724137931035,
+      "loss": 0.3038,
+      "step": 3220
+    },
+    {
+      "epoch": 58.56,
+      "learning_rate": 0.0001253793103448276,
+      "loss": 0.2562,
+      "step": 3221
+    },
+    {
+      "epoch": 58.58,
+      "learning_rate": 0.00012524137931034483,
+      "loss": 0.2051,
+      "step": 3222
+    },
+    {
+      "epoch": 58.6,
+      "learning_rate": 0.00012510344827586207,
+      "loss": 0.2294,
+      "step": 3223
+    },
+    {
+      "epoch": 58.62,
+      "learning_rate": 0.0001249655172413793,
+      "loss": 0.2157,
+      "step": 3224
+    },
+    {
+      "epoch": 58.64,
+      "learning_rate": 0.00012482758620689654,
+      "loss": 0.1867,
+      "step": 3225
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 0.00012468965517241378,
+      "loss": 0.2261,
+      "step": 3226
+    },
+    {
+      "epoch": 58.67,
+      "learning_rate": 0.00012455172413793102,
+      "loss": 0.271,
+      "step": 3227
+    },
+    {
+      "epoch": 58.69,
+      "learning_rate": 0.0001244137931034483,
+      "loss": 0.183,
+      "step": 3228
+    },
+    {
+      "epoch": 58.71,
+      "learning_rate": 0.00012427586206896553,
+      "loss": 0.2329,
+      "step": 3229
+    },
+    {
+      "epoch": 58.73,
+      "learning_rate": 0.00012413793103448277,
+      "loss": 0.2247,
+      "step": 3230
+    },
+    {
+      "epoch": 58.75,
+      "learning_rate": 0.000124,
+      "loss": 0.1955,
+      "step": 3231
+    },
+    {
+      "epoch": 58.76,
+      "learning_rate": 0.00012386206896551724,
+      "loss": 0.2447,
+      "step": 3232
+    },
+    {
+      "epoch": 58.78,
+      "learning_rate": 0.00012372413793103448,
+      "loss": 0.2244,
+      "step": 3233
+    },
+    {
+      "epoch": 58.8,
+      "learning_rate": 0.00012358620689655172,
+      "loss": 0.2033,
+      "step": 3234
+    },
+    {
+      "epoch": 58.82,
+      "learning_rate": 0.00012344827586206896,
+      "loss": 0.2893,
+      "step": 3235
+    },
+    {
+      "epoch": 58.84,
+      "learning_rate": 0.0001233103448275862,
+      "loss": 0.1887,
+      "step": 3236
+    },
+    {
+      "epoch": 58.85,
+      "learning_rate": 0.00012317241379310347,
+      "loss": 0.2024,
+      "step": 3237
+    },
+    {
+      "epoch": 58.87,
+      "learning_rate": 0.0001230344827586207,
+      "loss": 0.2426,
+      "step": 3238
+    },
+    {
+      "epoch": 58.89,
+      "learning_rate": 0.00012289655172413794,
+      "loss": 0.174,
+      "step": 3239
+    },
+    {
+      "epoch": 58.91,
+      "learning_rate": 0.00012275862068965518,
+      "loss": 0.2252,
+      "step": 3240
+    },
+    {
+      "epoch": 58.93,
+      "learning_rate": 0.00012262068965517242,
+      "loss": 0.2351,
+      "step": 3241
+    },
+    {
+      "epoch": 58.95,
+      "learning_rate": 0.00012248275862068966,
+      "loss": 0.2196,
+      "step": 3242
+    },
+    {
+      "epoch": 58.96,
+      "learning_rate": 0.0001223448275862069,
+      "loss": 0.2147,
+      "step": 3243
+    },
+    {
+      "epoch": 58.98,
+      "learning_rate": 0.00012220689655172414,
+      "loss": 0.1981,
+      "step": 3244
+    },
+    {
+      "epoch": 59.0,
+      "learning_rate": 0.00012206896551724138,
+      "loss": 0.2311,
+      "step": 3245
+    },
+    {
+      "epoch": 59.02,
+      "learning_rate": 0.00012193103448275862,
+      "loss": 0.2373,
+      "step": 3246
+    },
+    {
+      "epoch": 59.04,
+      "learning_rate": 0.00012179310344827587,
+      "loss": 0.2366,
+      "step": 3247
+    },
+    {
+      "epoch": 59.05,
+      "learning_rate": 0.00012165517241379311,
+      "loss": 0.2092,
+      "step": 3248
+    },
+    {
+      "epoch": 59.07,
+      "learning_rate": 0.00012151724137931035,
+      "loss": 0.2131,
+      "step": 3249
+    },
+    {
+      "epoch": 59.09,
+      "learning_rate": 0.00012137931034482759,
+      "loss": 0.2637,
+      "step": 3250
+    },
+    {
+      "epoch": 59.11,
+      "learning_rate": 0.00012124137931034483,
+      "loss": 0.2,
+      "step": 3251
+    },
+    {
+      "epoch": 59.13,
+      "learning_rate": 0.00012110344827586208,
+      "loss": 0.2233,
+      "step": 3252
+    },
+    {
+      "epoch": 59.15,
+      "learning_rate": 0.00012096551724137932,
+      "loss": 0.2185,
+      "step": 3253
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 0.00012082758620689656,
+      "loss": 0.2298,
+      "step": 3254
+    },
+    {
+      "epoch": 59.18,
+      "learning_rate": 0.0001206896551724138,
+      "loss": 0.2411,
+      "step": 3255
+    },
+    {
+      "epoch": 59.2,
+      "learning_rate": 0.00012055172413793105,
+      "loss": 0.2401,
+      "step": 3256
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 0.00012041379310344829,
+      "loss": 0.1824,
+      "step": 3257
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 0.00012027586206896551,
+      "loss": 0.2515,
+      "step": 3258
+    },
+    {
+      "epoch": 59.25,
+      "learning_rate": 0.00012013793103448275,
+      "loss": 0.2083,
+      "step": 3259
+    },
+    {
+      "epoch": 59.27,
+      "learning_rate": 0.00012,
+      "loss": 0.2233,
+      "step": 3260
+    },
+    {
+      "epoch": 59.29,
+      "learning_rate": 0.00011986206896551724,
+      "loss": 0.2417,
+      "step": 3261
+    },
+    {
+      "epoch": 59.31,
+      "learning_rate": 0.00011972413793103448,
+      "loss": 0.2383,
+      "step": 3262
+    },
+    {
+      "epoch": 59.33,
+      "learning_rate": 0.00011958620689655172,
+      "loss": 0.2471,
+      "step": 3263
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 0.00011944827586206897,
+      "loss": 0.2684,
+      "step": 3264
+    },
+    {
+      "epoch": 59.36,
+      "learning_rate": 0.00011931034482758621,
+      "loss": 0.2333,
+      "step": 3265
+    },
+    {
+      "epoch": 59.38,
+      "learning_rate": 0.00011917241379310345,
+      "loss": 0.22,
+      "step": 3266
+    },
+    {
+      "epoch": 59.4,
+      "learning_rate": 0.00011903448275862069,
+      "loss": 0.2064,
+      "step": 3267
+    },
+    {
+      "epoch": 59.42,
+      "learning_rate": 0.00011889655172413794,
+      "loss": 0.2715,
+      "step": 3268
+    },
+    {
+      "epoch": 59.44,
+      "learning_rate": 0.00011875862068965518,
+      "loss": 0.22,
+      "step": 3269
+    },
+    {
+      "epoch": 59.45,
+      "learning_rate": 0.00011862068965517242,
+      "loss": 0.1907,
+      "step": 3270
+    },
+    {
+      "epoch": 59.47,
+      "learning_rate": 0.00011848275862068966,
+      "loss": 0.2513,
+      "step": 3271
+    },
+    {
+      "epoch": 59.49,
+      "learning_rate": 0.00011834482758620691,
+      "loss": 0.1787,
+      "step": 3272
+    },
+    {
+      "epoch": 59.51,
+      "learning_rate": 0.00011820689655172415,
+      "loss": 0.2137,
+      "step": 3273
+    },
+    {
+      "epoch": 59.53,
+      "learning_rate": 0.00011806896551724138,
+      "loss": 0.2395,
+      "step": 3274
+    },
+    {
+      "epoch": 59.55,
+      "learning_rate": 0.00011793103448275861,
+      "loss": 0.2261,
+      "step": 3275
+    },
+    {
+      "epoch": 59.56,
+      "learning_rate": 0.00011779310344827585,
+      "loss": 0.2207,
+      "step": 3276
+    },
+    {
+      "epoch": 59.58,
+      "learning_rate": 0.0001176551724137931,
+      "loss": 0.2437,
+      "step": 3277
+    },
+    {
+      "epoch": 59.6,
+      "learning_rate": 0.00011751724137931035,
+      "loss": 0.2249,
+      "step": 3278
+    },
+    {
+      "epoch": 59.62,
+      "learning_rate": 0.00011737931034482758,
+      "loss": 0.2416,
+      "step": 3279
+    },
+    {
+      "epoch": 59.64,
+      "learning_rate": 0.00011724137931034482,
+      "loss": 0.2465,
+      "step": 3280
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 0.00011710344827586208,
+      "loss": 0.1846,
+      "step": 3281
+    },
+    {
+      "epoch": 59.67,
+      "learning_rate": 0.00011696551724137931,
+      "loss": 0.2231,
+      "step": 3282
+    },
+    {
+      "epoch": 59.69,
+      "learning_rate": 0.00011682758620689655,
+      "loss": 0.2252,
+      "step": 3283
+    },
+    {
+      "epoch": 59.71,
+      "learning_rate": 0.00011668965517241379,
+      "loss": 0.2213,
+      "step": 3284
+    },
+    {
+      "epoch": 59.73,
+      "learning_rate": 0.00011655172413793104,
+      "loss": 0.2825,
+      "step": 3285
+    },
+    {
+      "epoch": 59.75,
+      "learning_rate": 0.00011641379310344828,
+      "loss": 0.224,
+      "step": 3286
+    },
+    {
+      "epoch": 59.76,
+      "learning_rate": 0.00011627586206896552,
+      "loss": 0.2455,
+      "step": 3287
+    },
+    {
+      "epoch": 59.78,
+      "learning_rate": 0.00011613793103448276,
+      "loss": 0.219,
+      "step": 3288
+    },
+    {
+      "epoch": 59.8,
+      "learning_rate": 0.00011600000000000001,
+      "loss": 0.2065,
+      "step": 3289
+    },
+    {
+      "epoch": 59.82,
+      "learning_rate": 0.00011586206896551725,
+      "loss": 0.2001,
+      "step": 3290
+    },
+    {
+      "epoch": 59.84,
+      "learning_rate": 0.00011572413793103448,
+      "loss": 0.2533,
+      "step": 3291
+    },
+    {
+      "epoch": 59.85,
+      "learning_rate": 0.00011558620689655172,
+      "loss": 0.2194,
+      "step": 3292
+    },
+    {
+      "epoch": 59.87,
+      "learning_rate": 0.00011544827586206896,
+      "loss": 0.1922,
+      "step": 3293
+    },
+    {
+      "epoch": 59.89,
+      "learning_rate": 0.00011531034482758621,
+      "loss": 0.1954,
+      "step": 3294
+    },
+    {
+      "epoch": 59.91,
+      "learning_rate": 0.00011517241379310345,
+      "loss": 0.2026,
+      "step": 3295
+    },
+    {
+      "epoch": 59.93,
+      "learning_rate": 0.00011503448275862069,
+      "loss": 0.2687,
+      "step": 3296
+    },
+    {
+      "epoch": 59.95,
+      "learning_rate": 0.00011489655172413793,
+      "loss": 0.167,
+      "step": 3297
+    },
+    {
+      "epoch": 59.96,
+      "learning_rate": 0.00011475862068965518,
+      "loss": 0.2126,
+      "step": 3298
+    },
+    {
+      "epoch": 59.98,
+      "learning_rate": 0.00011462068965517242,
+      "loss": 0.2695,
+      "step": 3299
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 0.00011448275862068966,
+      "loss": 0.1646,
+      "step": 3300
+    },
+    {
+      "epoch": 60.02,
+      "learning_rate": 0.0001143448275862069,
+      "loss": 0.3128,
+      "step": 3301
+    },
+    {
+      "epoch": 60.04,
+      "learning_rate": 0.00011420689655172415,
+      "loss": 0.2274,
+      "step": 3302
+    },
+    {
+      "epoch": 60.05,
+      "learning_rate": 0.00011406896551724139,
+      "loss": 0.1842,
+      "step": 3303
+    },
+    {
+      "epoch": 60.07,
+      "learning_rate": 0.00011393103448275863,
+      "loss": 0.2078,
+      "step": 3304
+    },
+    {
+      "epoch": 60.09,
+      "learning_rate": 0.00011379310344827586,
+      "loss": 0.2388,
+      "step": 3305
+    },
+    {
+      "epoch": 60.11,
+      "learning_rate": 0.00011365517241379312,
+      "loss": 0.2227,
+      "step": 3306
+    },
+    {
+      "epoch": 60.13,
+      "learning_rate": 0.00011351724137931034,
+      "loss": 0.1904,
+      "step": 3307
+    },
+    {
+      "epoch": 60.15,
+      "learning_rate": 0.00011337931034482758,
+      "loss": 0.2251,
+      "step": 3308
+    },
+    {
+      "epoch": 60.16,
+      "learning_rate": 0.00011324137931034482,
+      "loss": 0.2177,
+      "step": 3309
+    },
+    {
+      "epoch": 60.18,
+      "learning_rate": 0.00011310344827586207,
+      "loss": 0.2043,
+      "step": 3310
+    },
+    {
+      "epoch": 60.2,
+      "learning_rate": 0.00011296551724137931,
+      "loss": 0.1806,
+      "step": 3311
+    },
+    {
+      "epoch": 60.22,
+      "learning_rate": 0.00011282758620689655,
+      "loss": 0.255,
+      "step": 3312
+    },
+    {
+      "epoch": 60.24,
+      "learning_rate": 0.00011268965517241379,
+      "loss": 0.2317,
+      "step": 3313
+    },
+    {
+      "epoch": 60.25,
+      "learning_rate": 0.00011255172413793104,
+      "loss": 0.2507,
+      "step": 3314
+    },
+    {
+      "epoch": 60.27,
+      "learning_rate": 0.00011241379310344828,
+      "loss": 0.2759,
+      "step": 3315
+    },
+    {
+      "epoch": 60.29,
+      "learning_rate": 0.00011227586206896552,
+      "loss": 0.2397,
+      "step": 3316
+    },
+    {
+      "epoch": 60.31,
+      "learning_rate": 0.00011213793103448276,
+      "loss": 0.2393,
+      "step": 3317
+    },
+    {
+      "epoch": 60.33,
+      "learning_rate": 0.000112,
+      "loss": 0.2368,
+      "step": 3318
+    },
+    {
+      "epoch": 60.35,
+      "learning_rate": 0.00011186206896551725,
+      "loss": 0.1984,
+      "step": 3319
+    },
+    {
+      "epoch": 60.36,
+      "learning_rate": 0.00011172413793103449,
+      "loss": 0.2901,
+      "step": 3320
+    },
+    {
+      "epoch": 60.38,
+      "learning_rate": 0.00011158620689655173,
+      "loss": 0.2177,
+      "step": 3321
+    },
+    {
+      "epoch": 60.4,
+      "learning_rate": 0.00011144827586206897,
+      "loss": 0.2247,
+      "step": 3322
+    },
+    {
+      "epoch": 60.42,
+      "learning_rate": 0.00011131034482758622,
+      "loss": 0.2028,
+      "step": 3323
+    },
+    {
+      "epoch": 60.44,
+      "learning_rate": 0.00011117241379310345,
+      "loss": 0.2322,
+      "step": 3324
+    },
+    {
+      "epoch": 60.45,
+      "learning_rate": 0.00011103448275862068,
+      "loss": 0.1417,
+      "step": 3325
+    },
+    {
+      "epoch": 60.47,
+      "learning_rate": 0.00011089655172413792,
+      "loss": 0.2513,
+      "step": 3326
+    },
+    {
+      "epoch": 60.49,
+      "learning_rate": 0.00011075862068965518,
+      "loss": 0.2304,
+      "step": 3327
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 0.00011062068965517242,
+      "loss": 0.2307,
+      "step": 3328
+    },
+    {
+      "epoch": 60.53,
+      "learning_rate": 0.00011048275862068965,
+      "loss": 0.2036,
+      "step": 3329
+    },
+    {
+      "epoch": 60.55,
+      "learning_rate": 0.00011034482758620689,
+      "loss": 0.1828,
+      "step": 3330
+    },
+    {
+      "epoch": 60.56,
+      "learning_rate": 0.00011020689655172415,
+      "loss": 0.2149,
+      "step": 3331
+    },
+    {
+      "epoch": 60.58,
+      "learning_rate": 0.00011006896551724138,
+      "loss": 0.1689,
+      "step": 3332
+    },
+    {
+      "epoch": 60.6,
+      "learning_rate": 0.00010993103448275862,
+      "loss": 0.2795,
+      "step": 3333
+    },
+    {
+      "epoch": 60.62,
+      "learning_rate": 0.00010979310344827586,
+      "loss": 0.1986,
+      "step": 3334
+    },
+    {
+      "epoch": 60.64,
+      "learning_rate": 0.00010965517241379311,
+      "loss": 0.2422,
+      "step": 3335
+    },
+    {
+      "epoch": 60.65,
+      "learning_rate": 0.00010951724137931035,
+      "loss": 0.1859,
+      "step": 3336
+    },
+    {
+      "epoch": 60.67,
+      "learning_rate": 0.00010937931034482759,
+      "loss": 0.1928,
+      "step": 3337
+    },
+    {
+      "epoch": 60.69,
+      "learning_rate": 0.00010924137931034483,
+      "loss": 0.2082,
+      "step": 3338
+    },
+    {
+      "epoch": 60.71,
+      "learning_rate": 0.00010910344827586207,
+      "loss": 0.2107,
+      "step": 3339
+    },
+    {
+      "epoch": 60.73,
+      "learning_rate": 0.00010896551724137932,
+      "loss": 0.2179,
+      "step": 3340
+    },
+    {
+      "epoch": 60.75,
+      "learning_rate": 0.00010882758620689655,
+      "loss": 0.2174,
+      "step": 3341
+    },
+    {
+      "epoch": 60.76,
+      "learning_rate": 0.00010868965517241379,
+      "loss": 0.2086,
+      "step": 3342
+    },
+    {
+      "epoch": 60.78,
+      "learning_rate": 0.00010855172413793103,
+      "loss": 0.2039,
+      "step": 3343
+    },
+    {
+      "epoch": 60.8,
+      "learning_rate": 0.00010841379310344828,
+      "loss": 0.2275,
+      "step": 3344
+    },
+    {
+      "epoch": 60.82,
+      "learning_rate": 0.00010827586206896552,
+      "loss": 0.2356,
+      "step": 3345
+    },
+    {
+      "epoch": 60.84,
+      "learning_rate": 0.00010813793103448276,
+      "loss": 0.1889,
+      "step": 3346
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 0.000108,
+      "loss": 0.2108,
+      "step": 3347
+    },
+    {
+      "epoch": 60.87,
+      "learning_rate": 0.00010786206896551725,
+      "loss": 0.2621,
+      "step": 3348
+    },
+    {
+      "epoch": 60.89,
+      "learning_rate": 0.00010772413793103449,
+      "loss": 0.2429,
+      "step": 3349
+    },
+    {
+      "epoch": 60.91,
+      "learning_rate": 0.00010758620689655173,
+      "loss": 0.1822,
+      "step": 3350
+    },
+    {
+      "epoch": 60.93,
+      "learning_rate": 0.00010744827586206897,
+      "loss": 0.2316,
+      "step": 3351
+    },
+    {
+      "epoch": 60.95,
+      "learning_rate": 0.00010731034482758622,
+      "loss": 0.217,
+      "step": 3352
+    },
+    {
+      "epoch": 60.96,
+      "learning_rate": 0.00010717241379310346,
+      "loss": 0.2171,
+      "step": 3353
+    },
+    {
+      "epoch": 60.98,
+      "learning_rate": 0.0001070344827586207,
+      "loss": 0.1955,
+      "step": 3354
+    },
+    {
+      "epoch": 61.0,
+      "learning_rate": 0.00010689655172413793,
+      "loss": 0.2333,
+      "step": 3355
+    },
+    {
+      "epoch": 61.02,
+      "learning_rate": 0.00010675862068965519,
+      "loss": 0.2639,
+      "step": 3356
+    },
+    {
+      "epoch": 61.04,
+      "learning_rate": 0.00010662068965517241,
+      "loss": 0.2215,
+      "step": 3357
+    },
+    {
+      "epoch": 61.05,
+      "learning_rate": 0.00010648275862068965,
+      "loss": 0.2643,
+      "step": 3358
+    },
+    {
+      "epoch": 61.07,
+      "learning_rate": 0.00010634482758620689,
+      "loss": 0.2082,
+      "step": 3359
+    },
+    {
+      "epoch": 61.09,
+      "learning_rate": 0.00010620689655172413,
+      "loss": 0.2193,
+      "step": 3360
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 0.00010606896551724138,
+      "loss": 0.2204,
+      "step": 3361
+    },
+    {
+      "epoch": 61.13,
+      "learning_rate": 0.00010593103448275862,
+      "loss": 0.2538,
+      "step": 3362
+    },
+    {
+      "epoch": 61.15,
+      "learning_rate": 0.00010579310344827586,
+      "loss": 0.2486,
+      "step": 3363
+    },
+    {
+      "epoch": 61.16,
+      "learning_rate": 0.0001056551724137931,
+      "loss": 0.2253,
+      "step": 3364
+    },
+    {
+      "epoch": 61.18,
+      "learning_rate": 0.00010551724137931035,
+      "loss": 0.2262,
+      "step": 3365
+    },
+    {
+      "epoch": 61.2,
+      "learning_rate": 0.00010537931034482759,
+      "loss": 0.192,
+      "step": 3366
+    },
+    {
+      "epoch": 61.22,
+      "learning_rate": 0.00010524137931034483,
+      "loss": 0.2125,
+      "step": 3367
+    },
+    {
+      "epoch": 61.24,
+      "learning_rate": 0.00010510344827586207,
+      "loss": 0.2171,
+      "step": 3368
+    },
+    {
+      "epoch": 61.25,
+      "learning_rate": 0.00010496551724137932,
+      "loss": 0.2033,
+      "step": 3369
+    },
+    {
+      "epoch": 61.27,
+      "learning_rate": 0.00010482758620689656,
+      "loss": 0.16,
+      "step": 3370
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 0.0001046896551724138,
+      "loss": 0.2332,
+      "step": 3371
+    },
+    {
+      "epoch": 61.31,
+      "learning_rate": 0.00010455172413793104,
+      "loss": 0.2521,
+      "step": 3372
+    },
+    {
+      "epoch": 61.33,
+      "learning_rate": 0.00010441379310344829,
+      "loss": 0.1453,
+      "step": 3373
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 0.00010427586206896552,
+      "loss": 0.2088,
+      "step": 3374
+    },
+    {
+      "epoch": 61.36,
+      "learning_rate": 0.00010413793103448275,
+      "loss": 0.2223,
+      "step": 3375
+    },
+    {
+      "epoch": 61.38,
+      "learning_rate": 0.000104,
+      "loss": 0.2214,
+      "step": 3376
+    },
+    {
+      "epoch": 61.4,
+      "learning_rate": 0.00010386206896551725,
+      "loss": 0.2671,
+      "step": 3377
+    },
+    {
+      "epoch": 61.42,
+      "learning_rate": 0.00010372413793103449,
+      "loss": 0.2382,
+      "step": 3378
+    },
+    {
+      "epoch": 61.44,
+      "learning_rate": 0.00010358620689655172,
+      "loss": 0.1883,
+      "step": 3379
+    },
+    {
+      "epoch": 61.45,
+      "learning_rate": 0.00010344827586206896,
+      "loss": 0.204,
+      "step": 3380
+    },
+    {
+      "epoch": 61.47,
+      "learning_rate": 0.0001033103448275862,
+      "loss": 0.2066,
+      "step": 3381
+    },
+    {
+      "epoch": 61.49,
+      "learning_rate": 0.00010317241379310345,
+      "loss": 0.1882,
+      "step": 3382
+    },
+    {
+      "epoch": 61.51,
+      "learning_rate": 0.0001030344827586207,
+      "loss": 0.1882,
+      "step": 3383
+    },
+    {
+      "epoch": 61.53,
+      "learning_rate": 0.00010289655172413793,
+      "loss": 0.204,
+      "step": 3384
+    },
+    {
+      "epoch": 61.55,
+      "learning_rate": 0.00010275862068965517,
+      "loss": 0.1895,
+      "step": 3385
+    },
+    {
+      "epoch": 61.56,
+      "learning_rate": 0.00010262068965517242,
+      "loss": 0.279,
+      "step": 3386
+    },
+    {
+      "epoch": 61.58,
+      "learning_rate": 0.00010248275862068966,
+      "loss": 0.2076,
+      "step": 3387
+    },
+    {
+      "epoch": 61.6,
+      "learning_rate": 0.0001023448275862069,
+      "loss": 0.2428,
+      "step": 3388
+    },
+    {
+      "epoch": 61.62,
+      "learning_rate": 0.00010220689655172414,
+      "loss": 0.2153,
+      "step": 3389
+    },
+    {
+      "epoch": 61.64,
+      "learning_rate": 0.0001020689655172414,
+      "loss": 0.234,
+      "step": 3390
+    },
+    {
+      "epoch": 61.65,
+      "learning_rate": 0.00010193103448275862,
+      "loss": 0.2133,
+      "step": 3391
+    },
+    {
+      "epoch": 61.67,
+      "learning_rate": 0.00010179310344827586,
+      "loss": 0.2353,
+      "step": 3392
+    },
+    {
+      "epoch": 61.69,
+      "learning_rate": 0.0001016551724137931,
+      "loss": 0.262,
+      "step": 3393
+    },
+    {
+      "epoch": 61.71,
+      "learning_rate": 0.00010151724137931035,
+      "loss": 0.2417,
+      "step": 3394
+    },
+    {
+      "epoch": 61.73,
+      "learning_rate": 0.00010137931034482759,
+      "loss": 0.2174,
+      "step": 3395
+    },
+    {
+      "epoch": 61.75,
+      "learning_rate": 0.00010124137931034483,
+      "loss": 0.2156,
+      "step": 3396
+    },
+    {
+      "epoch": 61.76,
+      "learning_rate": 0.00010110344827586207,
+      "loss": 0.1707,
+      "step": 3397
+    },
+    {
+      "epoch": 61.78,
+      "learning_rate": 0.00010096551724137932,
+      "loss": 0.2412,
+      "step": 3398
+    },
+    {
+      "epoch": 61.8,
+      "learning_rate": 0.00010082758620689656,
+      "loss": 0.2353,
+      "step": 3399
+    },
+    {
+      "epoch": 61.82,
+      "learning_rate": 0.0001006896551724138,
+      "loss": 0.2306,
+      "step": 3400
+    },
+    {
+      "epoch": 61.84,
+      "learning_rate": 0.00010055172413793104,
+      "loss": 0.1947,
+      "step": 3401
+    },
+    {
+      "epoch": 61.85,
+      "learning_rate": 0.00010041379310344827,
+      "loss": 0.2268,
+      "step": 3402
+    },
+    {
+      "epoch": 61.87,
+      "learning_rate": 0.00010027586206896553,
+      "loss": 0.1911,
+      "step": 3403
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 0.00010013793103448277,
+      "loss": 0.2969,
+      "step": 3404
+    },
+    {
+      "epoch": 61.91,
+      "learning_rate": 0.0001,
+      "loss": 0.2185,
+      "step": 3405
+    },
+    {
+      "epoch": 61.93,
+      "learning_rate": 9.986206896551724e-05,
+      "loss": 0.2314,
+      "step": 3406
+    },
+    {
+      "epoch": 61.95,
+      "learning_rate": 9.972413793103448e-05,
+      "loss": 0.2258,
+      "step": 3407
+    },
+    {
+      "epoch": 61.96,
+      "learning_rate": 9.958620689655172e-05,
+      "loss": 0.2241,
+      "step": 3408
+    },
+    {
+      "epoch": 61.98,
+      "learning_rate": 9.944827586206896e-05,
+      "loss": 0.1658,
+      "step": 3409
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 9.93103448275862e-05,
+      "loss": 0.1927,
+      "step": 3410
+    },
+    {
+      "epoch": 62.02,
+      "learning_rate": 9.917241379310345e-05,
+      "loss": 0.2862,
+      "step": 3411
+    },
+    {
+      "epoch": 62.04,
+      "learning_rate": 9.903448275862069e-05,
+      "loss": 0.1735,
+      "step": 3412
+    },
+    {
+      "epoch": 62.05,
+      "learning_rate": 9.889655172413793e-05,
+      "loss": 0.1902,
+      "step": 3413
+    },
+    {
+      "epoch": 62.07,
+      "learning_rate": 9.875862068965517e-05,
+      "loss": 0.1961,
+      "step": 3414
+    },
+    {
+      "epoch": 62.09,
+      "learning_rate": 9.862068965517242e-05,
+      "loss": 0.1906,
+      "step": 3415
+    },
+    {
+      "epoch": 62.11,
+      "learning_rate": 9.848275862068966e-05,
+      "loss": 0.168,
+      "step": 3416
+    },
+    {
+      "epoch": 62.13,
+      "learning_rate": 9.83448275862069e-05,
+      "loss": 0.2012,
+      "step": 3417
+    },
+    {
+      "epoch": 62.15,
+      "learning_rate": 9.820689655172414e-05,
+      "loss": 0.2016,
+      "step": 3418
+    },
+    {
+      "epoch": 62.16,
+      "learning_rate": 9.806896551724139e-05,
+      "loss": 0.1997,
+      "step": 3419
+    },
+    {
+      "epoch": 62.18,
+      "learning_rate": 9.793103448275863e-05,
+      "loss": 0.2429,
+      "step": 3420
+    },
+    {
+      "epoch": 62.2,
+      "learning_rate": 9.779310344827587e-05,
+      "loss": 0.2593,
+      "step": 3421
+    },
+    {
+      "epoch": 62.22,
+      "learning_rate": 9.765517241379311e-05,
+      "loss": 0.2003,
+      "step": 3422
+    },
+    {
+      "epoch": 62.24,
+      "learning_rate": 9.751724137931036e-05,
+      "loss": 0.2142,
+      "step": 3423
+    },
+    {
+      "epoch": 62.25,
+      "learning_rate": 9.737931034482759e-05,
+      "loss": 0.216,
+      "step": 3424
+    },
+    {
+      "epoch": 62.27,
+      "learning_rate": 9.724137931034482e-05,
+      "loss": 0.2376,
+      "step": 3425
+    },
+    {
+      "epoch": 62.29,
+      "learning_rate": 9.710344827586206e-05,
+      "loss": 0.1558,
+      "step": 3426
+    },
+    {
+      "epoch": 62.31,
+      "learning_rate": 9.69655172413793e-05,
+      "loss": 0.1516,
+      "step": 3427
+    },
+    {
+      "epoch": 62.33,
+      "learning_rate": 9.682758620689656e-05,
+      "loss": 0.1899,
+      "step": 3428
+    },
+    {
+      "epoch": 62.35,
+      "learning_rate": 9.66896551724138e-05,
+      "loss": 0.2477,
+      "step": 3429
+    },
+    {
+      "epoch": 62.36,
+      "learning_rate": 9.655172413793103e-05,
+      "loss": 0.1916,
+      "step": 3430
+    },
+    {
+      "epoch": 62.38,
+      "learning_rate": 9.641379310344827e-05,
+      "loss": 0.1918,
+      "step": 3431
+    },
+    {
+      "epoch": 62.4,
+      "learning_rate": 9.627586206896552e-05,
+      "loss": 0.1846,
+      "step": 3432
+    },
+    {
+      "epoch": 62.42,
+      "learning_rate": 9.613793103448276e-05,
+      "loss": 0.2091,
+      "step": 3433
+    },
+    {
+      "epoch": 62.44,
+      "learning_rate": 9.6e-05,
+      "loss": 0.1894,
+      "step": 3434
+    },
+    {
+      "epoch": 62.45,
+      "learning_rate": 9.586206896551724e-05,
+      "loss": 0.2656,
+      "step": 3435
+    },
+    {
+      "epoch": 62.47,
+      "learning_rate": 9.57241379310345e-05,
+      "loss": 0.2016,
+      "step": 3436
+    },
+    {
+      "epoch": 62.49,
+      "learning_rate": 9.558620689655173e-05,
+      "loss": 0.2163,
+      "step": 3437
+    },
+    {
+      "epoch": 62.51,
+      "learning_rate": 9.544827586206897e-05,
+      "loss": 0.1899,
+      "step": 3438
+    },
+    {
+      "epoch": 62.53,
+      "learning_rate": 9.531034482758621e-05,
+      "loss": 0.1853,
+      "step": 3439
+    },
+    {
+      "epoch": 62.55,
+      "learning_rate": 9.517241379310345e-05,
+      "loss": 0.1754,
+      "step": 3440
+    },
+    {
+      "epoch": 62.56,
+      "learning_rate": 9.503448275862069e-05,
+      "loss": 0.2173,
+      "step": 3441
+    },
+    {
+      "epoch": 62.58,
+      "learning_rate": 9.489655172413793e-05,
+      "loss": 0.2424,
+      "step": 3442
+    },
+    {
+      "epoch": 62.6,
+      "learning_rate": 9.475862068965517e-05,
+      "loss": 0.2305,
+      "step": 3443
+    },
+    {
+      "epoch": 62.62,
+      "learning_rate": 9.46206896551724e-05,
+      "loss": 0.2123,
+      "step": 3444
+    },
+    {
+      "epoch": 62.64,
+      "learning_rate": 9.448275862068966e-05,
+      "loss": 0.2415,
+      "step": 3445
+    },
+    {
+      "epoch": 62.65,
+      "learning_rate": 9.43448275862069e-05,
+      "loss": 0.1892,
+      "step": 3446
+    },
+    {
+      "epoch": 62.67,
+      "learning_rate": 9.420689655172414e-05,
+      "loss": 0.182,
+      "step": 3447
+    },
+    {
+      "epoch": 62.69,
+      "learning_rate": 9.406896551724137e-05,
+      "loss": 0.1969,
+      "step": 3448
+    },
+    {
+      "epoch": 62.71,
+      "learning_rate": 9.393103448275863e-05,
+      "loss": 0.1885,
+      "step": 3449
+    },
+    {
+      "epoch": 62.73,
+      "learning_rate": 9.379310344827587e-05,
+      "loss": 0.19,
+      "step": 3450
+    },
+    {
+      "epoch": 62.75,
+      "learning_rate": 9.36551724137931e-05,
+      "loss": 0.2115,
+      "step": 3451
+    },
+    {
+      "epoch": 62.76,
+      "learning_rate": 9.351724137931034e-05,
+      "loss": 0.2136,
+      "step": 3452
+    },
+    {
+      "epoch": 62.78,
+      "learning_rate": 9.33793103448276e-05,
+      "loss": 0.1836,
+      "step": 3453
+    },
+    {
+      "epoch": 62.8,
+      "learning_rate": 9.324137931034484e-05,
+      "loss": 0.1873,
+      "step": 3454
+    },
+    {
+      "epoch": 62.82,
+      "learning_rate": 9.310344827586207e-05,
+      "loss": 0.2355,
+      "step": 3455
+    },
+    {
+      "epoch": 62.84,
+      "learning_rate": 9.296551724137931e-05,
+      "loss": 0.1611,
+      "step": 3456
+    },
+    {
+      "epoch": 62.85,
+      "learning_rate": 9.282758620689655e-05,
+      "loss": 0.2072,
+      "step": 3457
+    },
+    {
+      "epoch": 62.87,
+      "learning_rate": 9.268965517241379e-05,
+      "loss": 0.1546,
+      "step": 3458
+    },
+    {
+      "epoch": 62.89,
+      "learning_rate": 9.255172413793103e-05,
+      "loss": 0.212,
+      "step": 3459
+    },
+    {
+      "epoch": 62.91,
+      "learning_rate": 9.241379310344827e-05,
+      "loss": 0.128,
+      "step": 3460
+    },
+    {
+      "epoch": 62.93,
+      "learning_rate": 9.227586206896552e-05,
+      "loss": 0.2303,
+      "step": 3461
+    },
+    {
+      "epoch": 62.95,
+      "learning_rate": 9.213793103448276e-05,
+      "loss": 0.2113,
+      "step": 3462
+    },
+    {
+      "epoch": 62.96,
+      "learning_rate": 9.2e-05,
+      "loss": 0.2531,
+      "step": 3463
+    },
+    {
+      "epoch": 62.98,
+      "learning_rate": 9.186206896551724e-05,
+      "loss": 0.2028,
+      "step": 3464
+    },
+    {
+      "epoch": 63.0,
+      "learning_rate": 9.172413793103449e-05,
+      "loss": 0.1078,
+      "step": 3465
+    },
+    {
+      "epoch": 63.02,
+      "learning_rate": 9.158620689655173e-05,
+      "loss": 0.2248,
+      "step": 3466
+    },
+    {
+      "epoch": 63.04,
+      "learning_rate": 9.144827586206897e-05,
+      "loss": 0.1918,
+      "step": 3467
+    },
+    {
+      "epoch": 63.05,
+      "learning_rate": 9.131034482758621e-05,
+      "loss": 0.1674,
+      "step": 3468
+    },
+    {
+      "epoch": 63.07,
+      "learning_rate": 9.117241379310345e-05,
+      "loss": 0.1909,
+      "step": 3469
+    },
+    {
+      "epoch": 63.09,
+      "learning_rate": 9.10344827586207e-05,
+      "loss": 0.1866,
+      "step": 3470
+    },
+    {
+      "epoch": 63.11,
+      "learning_rate": 9.089655172413794e-05,
+      "loss": 0.1716,
+      "step": 3471
+    },
+    {
+      "epoch": 63.13,
+      "learning_rate": 9.075862068965518e-05,
+      "loss": 0.1853,
+      "step": 3472
+    },
+    {
+      "epoch": 63.15,
+      "learning_rate": 9.062068965517242e-05,
+      "loss": 0.1788,
+      "step": 3473
+    },
+    {
+      "epoch": 63.16,
+      "learning_rate": 9.048275862068966e-05,
+      "loss": 0.2037,
+      "step": 3474
+    },
+    {
+      "epoch": 63.18,
+      "learning_rate": 9.03448275862069e-05,
+      "loss": 0.2298,
+      "step": 3475
+    },
+    {
+      "epoch": 63.2,
+      "learning_rate": 9.020689655172413e-05,
+      "loss": 0.2071,
+      "step": 3476
+    },
+    {
+      "epoch": 63.22,
+      "learning_rate": 9.006896551724137e-05,
+      "loss": 0.1582,
+      "step": 3477
+    },
+    {
+      "epoch": 63.24,
+      "learning_rate": 8.993103448275863e-05,
+      "loss": 0.2003,
+      "step": 3478
+    },
+    {
+      "epoch": 63.25,
+      "learning_rate": 8.979310344827586e-05,
+      "loss": 0.2291,
+      "step": 3479
+    },
+    {
+      "epoch": 63.27,
+      "learning_rate": 8.96551724137931e-05,
+      "loss": 0.2057,
+      "step": 3480
+    },
+    {
+      "epoch": 63.29,
+      "learning_rate": 8.951724137931034e-05,
+      "loss": 0.2531,
+      "step": 3481
+    },
+    {
+      "epoch": 63.31,
+      "learning_rate": 8.93793103448276e-05,
+      "loss": 0.242,
+      "step": 3482
+    },
+    {
+      "epoch": 63.33,
+      "learning_rate": 8.924137931034483e-05,
+      "loss": 0.1642,
+      "step": 3483
+    },
+    {
+      "epoch": 63.35,
+      "learning_rate": 8.910344827586207e-05,
+      "loss": 0.1518,
+      "step": 3484
+    },
+    {
+      "epoch": 63.36,
+      "learning_rate": 8.896551724137931e-05,
+      "loss": 0.2091,
+      "step": 3485
+    },
+    {
+      "epoch": 63.38,
+      "learning_rate": 8.882758620689656e-05,
+      "loss": 0.2024,
+      "step": 3486
+    },
+    {
+      "epoch": 63.4,
+      "learning_rate": 8.86896551724138e-05,
+      "loss": 0.1825,
+      "step": 3487
+    },
+    {
+      "epoch": 63.42,
+      "learning_rate": 8.855172413793104e-05,
+      "loss": 0.1876,
+      "step": 3488
+    },
+    {
+      "epoch": 63.44,
+      "learning_rate": 8.841379310344828e-05,
+      "loss": 0.2688,
+      "step": 3489
+    },
+    {
+      "epoch": 63.45,
+      "learning_rate": 8.82758620689655e-05,
+      "loss": 0.128,
+      "step": 3490
+    },
+    {
+      "epoch": 63.47,
+      "learning_rate": 8.813793103448276e-05,
+      "loss": 0.2102,
+      "step": 3491
+    },
+    {
+      "epoch": 63.49,
+      "learning_rate": 8.8e-05,
+      "loss": 0.2216,
+      "step": 3492
+    },
+    {
+      "epoch": 63.51,
+      "learning_rate": 8.786206896551724e-05,
+      "loss": 0.1832,
+      "step": 3493
+    },
+    {
+      "epoch": 63.53,
+      "learning_rate": 8.772413793103448e-05,
+      "loss": 0.167,
+      "step": 3494
+    },
+    {
+      "epoch": 63.55,
+      "learning_rate": 8.758620689655173e-05,
+      "loss": 0.1774,
+      "step": 3495
+    },
+    {
+      "epoch": 63.56,
+      "learning_rate": 8.744827586206897e-05,
+      "loss": 0.2007,
+      "step": 3496
+    },
+    {
+      "epoch": 63.58,
+      "learning_rate": 8.73103448275862e-05,
+      "loss": 0.2232,
+      "step": 3497
+    },
+    {
+      "epoch": 63.6,
+      "learning_rate": 8.717241379310345e-05,
+      "loss": 0.2322,
+      "step": 3498
+    },
+    {
+      "epoch": 63.62,
+      "learning_rate": 8.70344827586207e-05,
+      "loss": 0.2039,
+      "step": 3499
+    },
+    {
+      "epoch": 63.64,
+      "learning_rate": 8.689655172413794e-05,
+      "loss": 0.2586,
+      "step": 3500
+    },
+    {
+      "epoch": 63.64,
+      "eval_cer": 0.08037601763144875,
+      "eval_loss": 0.3778879642486572,
+      "eval_runtime": 14.1231,
+      "eval_samples_per_second": 116.618,
+      "eval_steps_per_second": 1.841,
+      "eval_wer": 0.3296905321213359,
+      "step": 3500
+    },
+    {
+      "epoch": 63.65,
+      "learning_rate": 8.675862068965518e-05,
+      "loss": 0.1846,
+      "step": 3501
+    },
+    {
+      "epoch": 63.67,
+      "learning_rate": 8.662068965517241e-05,
+      "loss": 0.1943,
+      "step": 3502
+    },
+    {
+      "epoch": 63.69,
+      "learning_rate": 8.648275862068967e-05,
+      "loss": 0.19,
+      "step": 3503
+    },
+    {
+      "epoch": 63.71,
+      "learning_rate": 8.63448275862069e-05,
+      "loss": 0.1971,
+      "step": 3504
+    },
+    {
+      "epoch": 63.73,
+      "learning_rate": 8.620689655172414e-05,
+      "loss": 0.2012,
+      "step": 3505
+    },
+    {
+      "epoch": 63.75,
+      "learning_rate": 8.606896551724138e-05,
+      "loss": 0.1777,
+      "step": 3506
+    },
+    {
+      "epoch": 63.76,
+      "learning_rate": 8.593103448275861e-05,
+      "loss": 0.1909,
+      "step": 3507
+    },
+    {
+      "epoch": 63.78,
+      "learning_rate": 8.579310344827586e-05,
+      "loss": 0.1931,
+      "step": 3508
+    },
+    {
+      "epoch": 63.8,
+      "learning_rate": 8.56551724137931e-05,
+      "loss": 0.1797,
+      "step": 3509
+    },
+    {
+      "epoch": 63.82,
+      "learning_rate": 8.551724137931034e-05,
+      "loss": 0.2398,
+      "step": 3510
+    },
+    {
+      "epoch": 63.84,
+      "learning_rate": 8.537931034482758e-05,
+      "loss": 0.1741,
+      "step": 3511
+    },
+    {
+      "epoch": 63.85,
+      "learning_rate": 8.524137931034483e-05,
+      "loss": 0.218,
+      "step": 3512
+    },
+    {
+      "epoch": 63.87,
+      "learning_rate": 8.510344827586207e-05,
+      "loss": 0.1791,
+      "step": 3513
+    },
+    {
+      "epoch": 63.89,
+      "learning_rate": 8.496551724137931e-05,
+      "loss": 0.2013,
+      "step": 3514
+    },
+    {
+      "epoch": 63.91,
+      "learning_rate": 8.482758620689655e-05,
+      "loss": 0.1785,
+      "step": 3515
+    },
+    {
+      "epoch": 63.93,
+      "learning_rate": 8.46896551724138e-05,
+      "loss": 0.2523,
+      "step": 3516
+    },
+    {
+      "epoch": 63.95,
+      "learning_rate": 8.455172413793104e-05,
+      "loss": 0.1891,
+      "step": 3517
+    },
+    {
+      "epoch": 63.96,
+      "learning_rate": 8.441379310344828e-05,
+      "loss": 0.2348,
+      "step": 3518
+    },
+    {
+      "epoch": 63.98,
+      "learning_rate": 8.427586206896552e-05,
+      "loss": 0.1896,
+      "step": 3519
+    },
+    {
+      "epoch": 64.0,
+      "learning_rate": 8.413793103448277e-05,
+      "loss": 0.1427,
+      "step": 3520
+    },
+    {
+      "epoch": 64.02,
+      "learning_rate": 8.400000000000001e-05,
+      "loss": 0.2089,
+      "step": 3521
+    },
+    {
+      "epoch": 64.04,
+      "learning_rate": 8.386206896551725e-05,
+      "loss": 0.2092,
+      "step": 3522
+    },
+    {
+      "epoch": 64.05,
+      "learning_rate": 8.372413793103449e-05,
+      "loss": 0.2382,
+      "step": 3523
+    },
+    {
+      "epoch": 64.07,
+      "learning_rate": 8.358620689655173e-05,
+      "loss": 0.1767,
+      "step": 3524
+    },
+    {
+      "epoch": 64.09,
+      "learning_rate": 8.344827586206896e-05,
+      "loss": 0.1981,
+      "step": 3525
+    },
+    {
+      "epoch": 64.11,
+      "learning_rate": 8.33103448275862e-05,
+      "loss": 0.1987,
+      "step": 3526
+    },
+    {
+      "epoch": 64.13,
+      "learning_rate": 8.317241379310344e-05,
+      "loss": 0.2339,
+      "step": 3527
+    },
+    {
+      "epoch": 64.15,
+      "learning_rate": 8.30344827586207e-05,
+      "loss": 0.1812,
+      "step": 3528
+    },
+    {
+      "epoch": 64.16,
+      "learning_rate": 8.289655172413793e-05,
+      "loss": 0.1889,
+      "step": 3529
+    },
+    {
+      "epoch": 64.18,
+      "learning_rate": 8.275862068965517e-05,
+      "loss": 0.1731,
+      "step": 3530
+    },
+    {
+      "epoch": 64.2,
+      "learning_rate": 8.262068965517241e-05,
+      "loss": 0.2092,
+      "step": 3531
+    },
+    {
+      "epoch": 64.22,
+      "learning_rate": 8.248275862068965e-05,
+      "loss": 0.2223,
+      "step": 3532
+    },
+    {
+      "epoch": 64.24,
+      "learning_rate": 8.23448275862069e-05,
+      "loss": 0.1968,
+      "step": 3533
+    },
+    {
+      "epoch": 64.25,
+      "learning_rate": 8.220689655172414e-05,
+      "loss": 0.1892,
+      "step": 3534
+    },
+    {
+      "epoch": 64.27,
+      "learning_rate": 8.206896551724138e-05,
+      "loss": 0.1968,
+      "step": 3535
+    },
+    {
+      "epoch": 64.29,
+      "learning_rate": 8.193103448275862e-05,
+      "loss": 0.1575,
+      "step": 3536
+    },
+    {
+      "epoch": 64.31,
+      "learning_rate": 8.179310344827587e-05,
+      "loss": 0.1956,
+      "step": 3537
+    },
+    {
+      "epoch": 64.33,
+      "learning_rate": 8.165517241379311e-05,
+      "loss": 0.1972,
+      "step": 3538
+    },
+    {
+      "epoch": 64.35,
+      "learning_rate": 8.151724137931035e-05,
+      "loss": 0.2363,
+      "step": 3539
+    },
+    {
+      "epoch": 64.36,
+      "learning_rate": 8.137931034482758e-05,
+      "loss": 0.1901,
+      "step": 3540
+    },
+    {
+      "epoch": 64.38,
+      "learning_rate": 8.124137931034483e-05,
+      "loss": 0.1698,
+      "step": 3541
+    },
+    {
+      "epoch": 64.4,
+      "learning_rate": 8.110344827586207e-05,
+      "loss": 0.2295,
+      "step": 3542
+    },
+    {
+      "epoch": 64.42,
+      "learning_rate": 8.09655172413793e-05,
+      "loss": 0.2187,
+      "step": 3543
+    },
+    {
+      "epoch": 64.44,
+      "learning_rate": 8.082758620689655e-05,
+      "loss": 0.1555,
+      "step": 3544
+    },
+    {
+      "epoch": 64.45,
+      "learning_rate": 8.06896551724138e-05,
+      "loss": 0.1642,
+      "step": 3545
+    },
+    {
+      "epoch": 64.47,
+      "learning_rate": 8.055172413793104e-05,
+      "loss": 0.245,
+      "step": 3546
+    },
+    {
+      "epoch": 64.49,
+      "learning_rate": 8.041379310344828e-05,
+      "loss": 0.1908,
+      "step": 3547
+    },
+    {
+      "epoch": 64.51,
+      "learning_rate": 8.027586206896552e-05,
+      "loss": 0.2091,
+      "step": 3548
+    },
+    {
+      "epoch": 64.53,
+      "learning_rate": 8.013793103448277e-05,
+      "loss": 0.1826,
+      "step": 3549
+    },
+    {
+      "epoch": 64.55,
+      "learning_rate": 8e-05,
+      "loss": 0.1546,
+      "step": 3550
+    },
+    {
+      "epoch": 64.56,
+      "learning_rate": 7.986206896551725e-05,
+      "loss": 0.2015,
+      "step": 3551
+    },
+    {
+      "epoch": 64.58,
+      "learning_rate": 7.972413793103448e-05,
+      "loss": 0.1996,
+      "step": 3552
+    },
+    {
+      "epoch": 64.6,
+      "learning_rate": 7.958620689655172e-05,
+      "loss": 0.2074,
+      "step": 3553
+    },
+    {
+      "epoch": 64.62,
+      "learning_rate": 7.944827586206898e-05,
+      "loss": 0.1777,
+      "step": 3554
+    },
+    {
+      "epoch": 64.64,
+      "learning_rate": 7.931034482758621e-05,
+      "loss": 0.1987,
+      "step": 3555
+    },
+    {
+      "epoch": 64.65,
+      "learning_rate": 7.917241379310345e-05,
+      "loss": 0.1811,
+      "step": 3556
+    },
+    {
+      "epoch": 64.67,
+      "learning_rate": 7.903448275862068e-05,
+      "loss": 0.1428,
+      "step": 3557
+    },
+    {
+      "epoch": 64.69,
+      "learning_rate": 7.889655172413793e-05,
+      "loss": 0.2082,
+      "step": 3558
+    },
+    {
+      "epoch": 64.71,
+      "learning_rate": 7.875862068965517e-05,
+      "loss": 0.1692,
+      "step": 3559
+    },
+    {
+      "epoch": 64.73,
+      "learning_rate": 7.862068965517241e-05,
+      "loss": 0.1877,
+      "step": 3560
+    },
+    {
+      "epoch": 64.75,
+      "learning_rate": 7.848275862068965e-05,
+      "loss": 0.1981,
+      "step": 3561
+    },
+    {
+      "epoch": 64.76,
+      "learning_rate": 7.83448275862069e-05,
+      "loss": 0.2104,
+      "step": 3562
+    },
+    {
+      "epoch": 64.78,
+      "learning_rate": 7.820689655172414e-05,
+      "loss": 0.2503,
+      "step": 3563
+    },
+    {
+      "epoch": 64.8,
+      "learning_rate": 7.806896551724138e-05,
+      "loss": 0.1959,
+      "step": 3564
+    },
+    {
+      "epoch": 64.82,
+      "learning_rate": 7.793103448275862e-05,
+      "loss": 0.2177,
+      "step": 3565
+    },
+    {
+      "epoch": 64.84,
+      "learning_rate": 7.779310344827587e-05,
+      "loss": 0.2003,
+      "step": 3566
+    },
+    {
+      "epoch": 64.85,
+      "learning_rate": 7.765517241379311e-05,
+      "loss": 0.2054,
+      "step": 3567
+    },
+    {
+      "epoch": 64.87,
+      "learning_rate": 7.751724137931035e-05,
+      "loss": 0.1661,
+      "step": 3568
+    },
+    {
+      "epoch": 64.89,
+      "learning_rate": 7.737931034482759e-05,
+      "loss": 0.1661,
+      "step": 3569
+    },
+    {
+      "epoch": 64.91,
+      "learning_rate": 7.724137931034484e-05,
+      "loss": 0.193,
+      "step": 3570
+    },
+    {
+      "epoch": 64.93,
+      "learning_rate": 7.710344827586208e-05,
+      "loss": 0.2714,
+      "step": 3571
+    },
+    {
+      "epoch": 64.95,
+      "learning_rate": 7.696551724137932e-05,
+      "loss": 0.1868,
+      "step": 3572
+    },
+    {
+      "epoch": 64.96,
+      "learning_rate": 7.682758620689654e-05,
+      "loss": 0.2128,
+      "step": 3573
+    },
+    {
+      "epoch": 64.98,
+      "learning_rate": 7.668965517241378e-05,
+      "loss": 0.2751,
+      "step": 3574
+    },
+    {
+      "epoch": 65.0,
+      "learning_rate": 7.655172413793103e-05,
+      "loss": 0.1744,
+      "step": 3575
+    },
+    {
+      "epoch": 65.02,
+      "learning_rate": 7.641379310344827e-05,
+      "loss": 0.2465,
+      "step": 3576
+    },
+    {
+      "epoch": 65.04,
+      "learning_rate": 7.627586206896551e-05,
+      "loss": 0.1487,
+      "step": 3577
+    },
+    {
+      "epoch": 65.05,
+      "learning_rate": 7.613793103448275e-05,
+      "loss": 0.2039,
+      "step": 3578
+    },
+    {
+      "epoch": 65.07,
+      "learning_rate": 7.6e-05,
+      "loss": 0.2506,
+      "step": 3579
+    },
+    {
+      "epoch": 65.09,
+      "learning_rate": 7.586206896551724e-05,
+      "loss": 0.2288,
+      "step": 3580
+    },
+    {
+      "epoch": 65.11,
+      "learning_rate": 7.572413793103448e-05,
+      "loss": 0.2024,
+      "step": 3581
+    },
+    {
+      "epoch": 65.13,
+      "learning_rate": 7.558620689655172e-05,
+      "loss": 0.2156,
+      "step": 3582
+    },
+    {
+      "epoch": 65.15,
+      "learning_rate": 7.544827586206897e-05,
+      "loss": 0.2292,
+      "step": 3583
+    },
+    {
+      "epoch": 65.16,
+      "learning_rate": 7.531034482758621e-05,
+      "loss": 0.2036,
+      "step": 3584
+    },
+    {
+      "epoch": 65.18,
+      "learning_rate": 7.517241379310345e-05,
+      "loss": 0.2123,
+      "step": 3585
+    },
+    {
+      "epoch": 65.2,
+      "learning_rate": 7.503448275862069e-05,
+      "loss": 0.2463,
+      "step": 3586
+    },
+    {
+      "epoch": 65.22,
+      "learning_rate": 7.489655172413794e-05,
+      "loss": 0.1583,
+      "step": 3587
+    },
+    {
+      "epoch": 65.24,
+      "learning_rate": 7.475862068965518e-05,
+      "loss": 0.1948,
+      "step": 3588
+    },
+    {
+      "epoch": 65.25,
+      "learning_rate": 7.462068965517242e-05,
+      "loss": 0.2202,
+      "step": 3589
+    },
+    {
+      "epoch": 65.27,
+      "learning_rate": 7.448275862068965e-05,
+      "loss": 0.2451,
+      "step": 3590
+    },
+    {
+      "epoch": 65.29,
+      "learning_rate": 7.43448275862069e-05,
+      "loss": 0.1866,
+      "step": 3591
+    },
+    {
+      "epoch": 65.31,
+      "learning_rate": 7.420689655172414e-05,
+      "loss": 0.1792,
+      "step": 3592
+    },
+    {
+      "epoch": 65.33,
+      "learning_rate": 7.406896551724138e-05,
+      "loss": 0.1512,
+      "step": 3593
+    },
+    {
+      "epoch": 65.35,
+      "learning_rate": 7.393103448275862e-05,
+      "loss": 0.1928,
+      "step": 3594
+    },
+    {
+      "epoch": 65.36,
+      "learning_rate": 7.379310344827585e-05,
+      "loss": 0.2347,
+      "step": 3595
+    },
+    {
+      "epoch": 65.38,
+      "learning_rate": 7.365517241379311e-05,
+      "loss": 0.2047,
+      "step": 3596
+    },
+    {
+      "epoch": 65.4,
+      "learning_rate": 7.351724137931035e-05,
+      "loss": 0.1812,
+      "step": 3597
+    },
+    {
+      "epoch": 65.42,
+      "learning_rate": 7.337931034482759e-05,
+      "loss": 0.1692,
+      "step": 3598
+    },
+    {
+      "epoch": 65.44,
+      "learning_rate": 7.324137931034482e-05,
+      "loss": 0.2217,
+      "step": 3599
+    },
+    {
+      "epoch": 65.45,
+      "learning_rate": 7.310344827586208e-05,
+      "loss": 0.2055,
+      "step": 3600
+    },
+    {
+      "epoch": 65.47,
+      "learning_rate": 7.296551724137932e-05,
+      "loss": 0.2224,
+      "step": 3601
+    },
+    {
+      "epoch": 65.49,
+      "learning_rate": 7.282758620689655e-05,
+      "loss": 0.2172,
+      "step": 3602
+    },
+    {
+      "epoch": 65.51,
+      "learning_rate": 7.26896551724138e-05,
+      "loss": 0.1995,
+      "step": 3603
+    },
+    {
+      "epoch": 65.53,
+      "learning_rate": 7.255172413793105e-05,
+      "loss": 0.1612,
+      "step": 3604
+    },
+    {
+      "epoch": 65.55,
+      "learning_rate": 7.241379310344828e-05,
+      "loss": 0.2107,
+      "step": 3605
+    },
+    {
+      "epoch": 65.56,
+      "learning_rate": 7.227586206896552e-05,
+      "loss": 0.1705,
+      "step": 3606
+    },
+    {
+      "epoch": 65.58,
+      "learning_rate": 7.213793103448275e-05,
+      "loss": 0.2049,
+      "step": 3607
+    },
+    {
+      "epoch": 65.6,
+      "learning_rate": 7.2e-05,
+      "loss": 0.2387,
+      "step": 3608
+    },
+    {
+      "epoch": 65.62,
+      "learning_rate": 7.186206896551724e-05,
+      "loss": 0.1419,
+      "step": 3609
+    },
+    {
+      "epoch": 65.64,
+      "learning_rate": 7.172413793103448e-05,
+      "loss": 0.1728,
+      "step": 3610
+    },
+    {
+      "epoch": 65.65,
+      "learning_rate": 7.158620689655172e-05,
+      "loss": 0.2635,
+      "step": 3611
+    },
+    {
+      "epoch": 65.67,
+      "learning_rate": 7.144827586206897e-05,
+      "loss": 0.1504,
+      "step": 3612
+    },
+    {
+      "epoch": 65.69,
+      "learning_rate": 7.131034482758621e-05,
+      "loss": 0.2772,
+      "step": 3613
+    },
+    {
+      "epoch": 65.71,
+      "learning_rate": 7.117241379310345e-05,
+      "loss": 0.2162,
+      "step": 3614
+    },
+    {
+      "epoch": 65.73,
+      "learning_rate": 7.103448275862069e-05,
+      "loss": 0.1954,
+      "step": 3615
+    },
+    {
+      "epoch": 65.75,
+      "learning_rate": 7.089655172413793e-05,
+      "loss": 0.1831,
+      "step": 3616
+    },
+    {
+      "epoch": 65.76,
+      "learning_rate": 7.075862068965518e-05,
+      "loss": 0.2703,
+      "step": 3617
+    },
+    {
+      "epoch": 65.78,
+      "learning_rate": 7.062068965517242e-05,
+      "loss": 0.2195,
+      "step": 3618
+    },
+    {
+      "epoch": 65.8,
+      "learning_rate": 7.048275862068966e-05,
+      "loss": 0.2109,
+      "step": 3619
+    },
+    {
+      "epoch": 65.82,
+      "learning_rate": 7.03448275862069e-05,
+      "loss": 0.2458,
+      "step": 3620
+    },
+    {
+      "epoch": 65.84,
+      "learning_rate": 7.020689655172415e-05,
+      "loss": 0.1965,
+      "step": 3621
+    },
+    {
+      "epoch": 65.85,
+      "learning_rate": 7.006896551724139e-05,
+      "loss": 0.2379,
+      "step": 3622
+    },
+    {
+      "epoch": 65.87,
+      "learning_rate": 6.993103448275861e-05,
+      "loss": 0.1686,
+      "step": 3623
+    },
+    {
+      "epoch": 65.89,
+      "learning_rate": 6.979310344827585e-05,
+      "loss": 0.1661,
+      "step": 3624
+    },
+    {
+      "epoch": 65.91,
+      "learning_rate": 6.96551724137931e-05,
+      "loss": 0.1549,
+      "step": 3625
+    },
+    {
+      "epoch": 65.93,
+      "learning_rate": 6.951724137931034e-05,
+      "loss": 0.1771,
+      "step": 3626
+    },
+    {
+      "epoch": 65.95,
+      "learning_rate": 6.937931034482758e-05,
+      "loss": 0.1829,
+      "step": 3627
+    },
+    {
+      "epoch": 65.96,
+      "learning_rate": 6.924137931034482e-05,
+      "loss": 0.1598,
+      "step": 3628
+    },
+    {
+      "epoch": 65.98,
+      "learning_rate": 6.910344827586207e-05,
+      "loss": 0.2119,
+      "step": 3629
+    },
+    {
+      "epoch": 66.0,
+      "learning_rate": 6.896551724137931e-05,
+      "loss": 0.1028,
+      "step": 3630
+    },
+    {
+      "epoch": 66.02,
+      "learning_rate": 6.882758620689655e-05,
+      "loss": 0.2338,
+      "step": 3631
+    },
+    {
+      "epoch": 66.04,
+      "learning_rate": 6.868965517241379e-05,
+      "loss": 0.1948,
+      "step": 3632
+    },
+    {
+      "epoch": 66.05,
+      "learning_rate": 6.855172413793104e-05,
+      "loss": 0.2175,
+      "step": 3633
+    },
+    {
+      "epoch": 66.07,
+      "learning_rate": 6.841379310344828e-05,
+      "loss": 0.1704,
+      "step": 3634
+    },
+    {
+      "epoch": 66.09,
+      "learning_rate": 6.827586206896552e-05,
+      "loss": 0.1816,
+      "step": 3635
+    },
+    {
+      "epoch": 66.11,
+      "learning_rate": 6.813793103448276e-05,
+      "loss": 0.2176,
+      "step": 3636
+    },
+    {
+      "epoch": 66.13,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.2745,
+      "step": 3637
+    },
+    {
+      "epoch": 66.15,
+      "learning_rate": 6.786206896551725e-05,
+      "loss": 0.2043,
+      "step": 3638
+    },
+    {
+      "epoch": 66.16,
+      "learning_rate": 6.772413793103449e-05,
+      "loss": 0.2047,
+      "step": 3639
+    },
+    {
+      "epoch": 66.18,
+      "learning_rate": 6.758620689655172e-05,
+      "loss": 0.2104,
+      "step": 3640
+    },
+    {
+      "epoch": 66.2,
+      "learning_rate": 6.744827586206896e-05,
+      "loss": 0.1798,
+      "step": 3641
+    },
+    {
+      "epoch": 66.22,
+      "learning_rate": 6.731034482758621e-05,
+      "loss": 0.2024,
+      "step": 3642
+    },
+    {
+      "epoch": 66.24,
+      "learning_rate": 6.717241379310345e-05,
+      "loss": 0.2189,
+      "step": 3643
+    },
+    {
+      "epoch": 66.25,
+      "learning_rate": 6.703448275862069e-05,
+      "loss": 0.1563,
+      "step": 3644
+    },
+    {
+      "epoch": 66.27,
+      "learning_rate": 6.689655172413792e-05,
+      "loss": 0.2405,
+      "step": 3645
+    },
+    {
+      "epoch": 66.29,
+      "learning_rate": 6.675862068965518e-05,
+      "loss": 0.2196,
+      "step": 3646
+    },
+    {
+      "epoch": 66.31,
+      "learning_rate": 6.662068965517242e-05,
+      "loss": 0.2207,
+      "step": 3647
+    },
+    {
+      "epoch": 66.33,
+      "learning_rate": 6.648275862068966e-05,
+      "loss": 0.1741,
+      "step": 3648
+    },
+    {
+      "epoch": 66.35,
+      "learning_rate": 6.63448275862069e-05,
+      "loss": 0.1903,
+      "step": 3649
+    },
+    {
+      "epoch": 66.36,
+      "learning_rate": 6.620689655172415e-05,
+      "loss": 0.2006,
+      "step": 3650
+    },
+    {
+      "epoch": 66.38,
+      "learning_rate": 6.606896551724139e-05,
+      "loss": 0.1968,
+      "step": 3651
+    },
+    {
+      "epoch": 66.4,
+      "learning_rate": 6.593103448275862e-05,
+      "loss": 0.2293,
+      "step": 3652
+    },
+    {
+      "epoch": 66.42,
+      "learning_rate": 6.579310344827586e-05,
+      "loss": 0.1697,
+      "step": 3653
+    },
+    {
+      "epoch": 66.44,
+      "learning_rate": 6.565517241379312e-05,
+      "loss": 0.1983,
+      "step": 3654
+    },
+    {
+      "epoch": 66.45,
+      "learning_rate": 6.551724137931035e-05,
+      "loss": 0.1526,
+      "step": 3655
+    },
+    {
+      "epoch": 66.47,
+      "learning_rate": 6.53793103448276e-05,
+      "loss": 0.2072,
+      "step": 3656
+    },
+    {
+      "epoch": 66.49,
+      "learning_rate": 6.524137931034482e-05,
+      "loss": 0.1982,
+      "step": 3657
+    },
+    {
+      "epoch": 66.51,
+      "learning_rate": 6.510344827586206e-05,
+      "loss": 0.2034,
+      "step": 3658
+    },
+    {
+      "epoch": 66.53,
+      "learning_rate": 6.496551724137931e-05,
+      "loss": 0.1951,
+      "step": 3659
+    },
+    {
+      "epoch": 66.55,
+      "learning_rate": 6.482758620689655e-05,
+      "loss": 0.2108,
+      "step": 3660
+    },
+    {
+      "epoch": 66.56,
+      "learning_rate": 6.468965517241379e-05,
+      "loss": 0.2102,
+      "step": 3661
+    },
+    {
+      "epoch": 66.58,
+      "learning_rate": 6.455172413793103e-05,
+      "loss": 0.2142,
+      "step": 3662
+    },
+    {
+      "epoch": 66.6,
+      "learning_rate": 6.441379310344828e-05,
+      "loss": 0.2209,
+      "step": 3663
+    },
+    {
+      "epoch": 66.62,
+      "learning_rate": 6.427586206896552e-05,
+      "loss": 0.1396,
+      "step": 3664
+    },
+    {
+      "epoch": 66.64,
+      "learning_rate": 6.413793103448276e-05,
+      "loss": 0.2381,
+      "step": 3665
+    },
+    {
+      "epoch": 66.65,
+      "learning_rate": 6.4e-05,
+      "loss": 0.2031,
+      "step": 3666
+    },
+    {
+      "epoch": 66.67,
+      "learning_rate": 6.386206896551725e-05,
+      "loss": 0.1667,
+      "step": 3667
+    },
+    {
+      "epoch": 66.69,
+      "learning_rate": 6.372413793103449e-05,
+      "loss": 0.1637,
+      "step": 3668
+    },
+    {
+      "epoch": 66.71,
+      "learning_rate": 6.358620689655173e-05,
+      "loss": 0.1957,
+      "step": 3669
+    },
+    {
+      "epoch": 66.73,
+      "learning_rate": 6.344827586206897e-05,
+      "loss": 0.1997,
+      "step": 3670
+    },
+    {
+      "epoch": 66.75,
+      "learning_rate": 6.331034482758622e-05,
+      "loss": 0.1491,
+      "step": 3671
+    },
+    {
+      "epoch": 66.76,
+      "learning_rate": 6.317241379310346e-05,
+      "loss": 0.1389,
+      "step": 3672
+    },
+    {
+      "epoch": 66.78,
+      "learning_rate": 6.303448275862068e-05,
+      "loss": 0.1652,
+      "step": 3673
+    },
+    {
+      "epoch": 66.8,
+      "learning_rate": 6.289655172413792e-05,
+      "loss": 0.1781,
+      "step": 3674
+    },
+    {
+      "epoch": 66.82,
+      "learning_rate": 6.275862068965517e-05,
+      "loss": 0.2092,
+      "step": 3675
+    },
+    {
+      "epoch": 66.84,
+      "learning_rate": 6.262068965517241e-05,
+      "loss": 0.1884,
+      "step": 3676
+    },
+    {
+      "epoch": 66.85,
+      "learning_rate": 6.248275862068965e-05,
+      "loss": 0.2014,
+      "step": 3677
+    },
+    {
+      "epoch": 66.87,
+      "learning_rate": 6.234482758620689e-05,
+      "loss": 0.1853,
+      "step": 3678
+    },
+    {
+      "epoch": 66.89,
+      "learning_rate": 6.220689655172414e-05,
+      "loss": 0.172,
+      "step": 3679
+    },
+    {
+      "epoch": 66.91,
+      "learning_rate": 6.206896551724138e-05,
+      "loss": 0.147,
+      "step": 3680
+    },
+    {
+      "epoch": 66.93,
+      "learning_rate": 6.193103448275862e-05,
+      "loss": 0.2306,
+      "step": 3681
+    },
+    {
+      "epoch": 66.95,
+      "learning_rate": 6.179310344827586e-05,
+      "loss": 0.1694,
+      "step": 3682
+    },
+    {
+      "epoch": 66.96,
+      "learning_rate": 6.16551724137931e-05,
+      "loss": 0.1997,
+      "step": 3683
+    },
+    {
+      "epoch": 66.98,
+      "learning_rate": 6.151724137931035e-05,
+      "loss": 0.1327,
+      "step": 3684
+    },
+    {
+      "epoch": 67.0,
+      "learning_rate": 6.137931034482759e-05,
+      "loss": 0.1044,
+      "step": 3685
+    },
+    {
+      "epoch": 67.02,
+      "learning_rate": 6.124137931034483e-05,
+      "loss": 0.2062,
+      "step": 3686
+    },
+    {
+      "epoch": 67.04,
+      "learning_rate": 6.110344827586207e-05,
+      "loss": 0.2021,
+      "step": 3687
+    },
+    {
+      "epoch": 67.05,
+      "learning_rate": 6.096551724137931e-05,
+      "loss": 0.1679,
+      "step": 3688
+    },
+    {
+      "epoch": 67.07,
+      "learning_rate": 6.0827586206896554e-05,
+      "loss": 0.1993,
+      "step": 3689
+    },
+    {
+      "epoch": 67.09,
+      "learning_rate": 6.068965517241379e-05,
+      "loss": 0.1659,
+      "step": 3690
+    },
+    {
+      "epoch": 67.11,
+      "learning_rate": 6.055172413793104e-05,
+      "loss": 0.1885,
+      "step": 3691
+    },
+    {
+      "epoch": 67.13,
+      "learning_rate": 6.041379310344828e-05,
+      "loss": 0.1703,
+      "step": 3692
+    },
+    {
+      "epoch": 67.15,
+      "learning_rate": 6.0275862068965524e-05,
+      "loss": 0.1854,
+      "step": 3693
+    },
+    {
+      "epoch": 67.16,
+      "learning_rate": 6.0137931034482756e-05,
+      "loss": 0.1813,
+      "step": 3694
+    },
+    {
+      "epoch": 67.18,
+      "learning_rate": 6e-05,
+      "loss": 0.1992,
+      "step": 3695
+    },
+    {
+      "epoch": 67.2,
+      "learning_rate": 5.986206896551724e-05,
+      "loss": 0.2167,
+      "step": 3696
+    },
+    {
+      "epoch": 67.22,
+      "learning_rate": 5.9724137931034486e-05,
+      "loss": 0.2129,
+      "step": 3697
+    },
+    {
+      "epoch": 67.24,
+      "learning_rate": 5.9586206896551725e-05,
+      "loss": 0.1853,
+      "step": 3698
+    },
+    {
+      "epoch": 67.25,
+      "learning_rate": 5.944827586206897e-05,
+      "loss": 0.2031,
+      "step": 3699
+    },
+    {
+      "epoch": 67.27,
+      "learning_rate": 5.931034482758621e-05,
+      "loss": 0.2198,
+      "step": 3700
+    },
+    {
+      "epoch": 67.29,
+      "learning_rate": 5.9172413793103456e-05,
+      "loss": 0.1618,
+      "step": 3701
+    },
+    {
+      "epoch": 67.31,
+      "learning_rate": 5.903448275862069e-05,
+      "loss": 0.2261,
+      "step": 3702
+    },
+    {
+      "epoch": 67.33,
+      "learning_rate": 5.889655172413793e-05,
+      "loss": 0.2049,
+      "step": 3703
+    },
+    {
+      "epoch": 67.35,
+      "learning_rate": 5.875862068965517e-05,
+      "loss": 0.1784,
+      "step": 3704
+    },
+    {
+      "epoch": 67.36,
+      "learning_rate": 5.862068965517241e-05,
+      "loss": 0.1829,
+      "step": 3705
+    },
+    {
+      "epoch": 67.38,
+      "learning_rate": 5.848275862068966e-05,
+      "loss": 0.1623,
+      "step": 3706
+    },
+    {
+      "epoch": 67.4,
+      "learning_rate": 5.8344827586206896e-05,
+      "loss": 0.1917,
+      "step": 3707
+    },
+    {
+      "epoch": 67.42,
+      "learning_rate": 5.820689655172414e-05,
+      "loss": 0.2228,
+      "step": 3708
+    },
+    {
+      "epoch": 67.44,
+      "learning_rate": 5.806896551724138e-05,
+      "loss": 0.1906,
+      "step": 3709
+    },
+    {
+      "epoch": 67.45,
+      "learning_rate": 5.7931034482758627e-05,
+      "loss": 0.1841,
+      "step": 3710
+    },
+    {
+      "epoch": 67.47,
+      "learning_rate": 5.779310344827586e-05,
+      "loss": 0.212,
+      "step": 3711
+    },
+    {
+      "epoch": 67.49,
+      "learning_rate": 5.7655172413793105e-05,
+      "loss": 0.193,
+      "step": 3712
+    },
+    {
+      "epoch": 67.51,
+      "learning_rate": 5.7517241379310343e-05,
+      "loss": 0.1518,
+      "step": 3713
+    },
+    {
+      "epoch": 67.53,
+      "learning_rate": 5.737931034482759e-05,
+      "loss": 0.1957,
+      "step": 3714
+    },
+    {
+      "epoch": 67.55,
+      "learning_rate": 5.724137931034483e-05,
+      "loss": 0.1596,
+      "step": 3715
+    },
+    {
+      "epoch": 67.56,
+      "learning_rate": 5.7103448275862074e-05,
+      "loss": 0.1448,
+      "step": 3716
+    },
+    {
+      "epoch": 67.58,
+      "learning_rate": 5.696551724137931e-05,
+      "loss": 0.2191,
+      "step": 3717
+    },
+    {
+      "epoch": 67.6,
+      "learning_rate": 5.682758620689656e-05,
+      "loss": 0.179,
+      "step": 3718
+    },
+    {
+      "epoch": 67.62,
+      "learning_rate": 5.668965517241379e-05,
+      "loss": 0.2017,
+      "step": 3719
+    },
+    {
+      "epoch": 67.64,
+      "learning_rate": 5.6551724137931037e-05,
+      "loss": 0.16,
+      "step": 3720
+    },
+    {
+      "epoch": 67.65,
+      "learning_rate": 5.6413793103448275e-05,
+      "loss": 0.1841,
+      "step": 3721
+    },
+    {
+      "epoch": 67.67,
+      "learning_rate": 5.627586206896552e-05,
+      "loss": 0.1453,
+      "step": 3722
+    },
+    {
+      "epoch": 67.69,
+      "learning_rate": 5.613793103448276e-05,
+      "loss": 0.1976,
+      "step": 3723
+    },
+    {
+      "epoch": 67.71,
+      "learning_rate": 5.6e-05,
+      "loss": 0.2016,
+      "step": 3724
+    },
+    {
+      "epoch": 67.73,
+      "learning_rate": 5.5862068965517245e-05,
+      "loss": 0.1951,
+      "step": 3725
+    },
+    {
+      "epoch": 67.75,
+      "learning_rate": 5.5724137931034484e-05,
+      "loss": 0.1767,
+      "step": 3726
+    },
+    {
+      "epoch": 67.76,
+      "learning_rate": 5.558620689655172e-05,
+      "loss": 0.205,
+      "step": 3727
+    },
+    {
+      "epoch": 67.78,
+      "learning_rate": 5.544827586206896e-05,
+      "loss": 0.2003,
+      "step": 3728
+    },
+    {
+      "epoch": 67.8,
+      "learning_rate": 5.531034482758621e-05,
+      "loss": 0.1706,
+      "step": 3729
+    },
+    {
+      "epoch": 67.82,
+      "learning_rate": 5.5172413793103446e-05,
+      "loss": 0.1853,
+      "step": 3730
+    },
+    {
+      "epoch": 67.84,
+      "learning_rate": 5.503448275862069e-05,
+      "loss": 0.1804,
+      "step": 3731
+    },
+    {
+      "epoch": 67.85,
+      "learning_rate": 5.489655172413793e-05,
+      "loss": 0.162,
+      "step": 3732
+    },
+    {
+      "epoch": 67.87,
+      "learning_rate": 5.475862068965518e-05,
+      "loss": 0.2027,
+      "step": 3733
+    },
+    {
+      "epoch": 67.89,
+      "learning_rate": 5.4620689655172416e-05,
+      "loss": 0.1643,
+      "step": 3734
+    },
+    {
+      "epoch": 67.91,
+      "learning_rate": 5.448275862068966e-05,
+      "loss": 0.1751,
+      "step": 3735
+    },
+    {
+      "epoch": 67.93,
+      "learning_rate": 5.4344827586206894e-05,
+      "loss": 0.1978,
+      "step": 3736
+    },
+    {
+      "epoch": 67.95,
+      "learning_rate": 5.420689655172414e-05,
+      "loss": 0.1575,
+      "step": 3737
+    },
+    {
+      "epoch": 67.96,
+      "learning_rate": 5.406896551724138e-05,
+      "loss": 0.187,
+      "step": 3738
+    },
+    {
+      "epoch": 67.98,
+      "learning_rate": 5.3931034482758624e-05,
+      "loss": 0.2067,
+      "step": 3739
+    },
+    {
+      "epoch": 68.0,
+      "learning_rate": 5.379310344827586e-05,
+      "loss": 0.2059,
+      "step": 3740
+    },
+    {
+      "epoch": 68.02,
+      "learning_rate": 5.365517241379311e-05,
+      "loss": 0.2226,
+      "step": 3741
+    },
+    {
+      "epoch": 68.04,
+      "learning_rate": 5.351724137931035e-05,
+      "loss": 0.1564,
+      "step": 3742
+    },
+    {
+      "epoch": 68.05,
+      "learning_rate": 5.3379310344827594e-05,
+      "loss": 0.2303,
+      "step": 3743
+    },
+    {
+      "epoch": 68.07,
+      "learning_rate": 5.3241379310344826e-05,
+      "loss": 0.1542,
+      "step": 3744
+    },
+    {
+      "epoch": 68.09,
+      "learning_rate": 5.3103448275862065e-05,
+      "loss": 0.157,
+      "step": 3745
+    },
+    {
+      "epoch": 68.11,
+      "learning_rate": 5.296551724137931e-05,
+      "loss": 0.2179,
+      "step": 3746
+    },
+    {
+      "epoch": 68.13,
+      "learning_rate": 5.282758620689655e-05,
+      "loss": 0.2516,
+      "step": 3747
+    },
+    {
+      "epoch": 68.15,
+      "learning_rate": 5.2689655172413795e-05,
+      "loss": 0.1697,
+      "step": 3748
+    },
+    {
+      "epoch": 68.16,
+      "learning_rate": 5.2551724137931034e-05,
+      "loss": 0.1696,
+      "step": 3749
+    },
+    {
+      "epoch": 68.18,
+      "learning_rate": 5.241379310344828e-05,
+      "loss": 0.2047,
+      "step": 3750
+    },
+    {
+      "epoch": 68.2,
+      "learning_rate": 5.227586206896552e-05,
+      "loss": 0.1493,
+      "step": 3751
+    },
+    {
+      "epoch": 68.22,
+      "learning_rate": 5.213793103448276e-05,
+      "loss": 0.174,
+      "step": 3752
+    },
+    {
+      "epoch": 68.24,
+      "learning_rate": 5.2e-05,
+      "loss": 0.1395,
+      "step": 3753
+    },
+    {
+      "epoch": 68.25,
+      "learning_rate": 5.186206896551724e-05,
+      "loss": 0.1993,
+      "step": 3754
+    },
+    {
+      "epoch": 68.27,
+      "learning_rate": 5.172413793103448e-05,
+      "loss": 0.1791,
+      "step": 3755
+    },
+    {
+      "epoch": 68.29,
+      "learning_rate": 5.158620689655173e-05,
+      "loss": 0.1355,
+      "step": 3756
+    },
+    {
+      "epoch": 68.31,
+      "learning_rate": 5.1448275862068966e-05,
+      "loss": 0.1261,
+      "step": 3757
+    },
+    {
+      "epoch": 68.33,
+      "learning_rate": 5.131034482758621e-05,
+      "loss": 0.2273,
+      "step": 3758
+    },
+    {
+      "epoch": 68.35,
+      "learning_rate": 5.117241379310345e-05,
+      "loss": 0.1519,
+      "step": 3759
+    },
+    {
+      "epoch": 68.36,
+      "learning_rate": 5.10344827586207e-05,
+      "loss": 0.2077,
+      "step": 3760
+    },
+    {
+      "epoch": 68.38,
+      "learning_rate": 5.089655172413793e-05,
+      "loss": 0.1685,
+      "step": 3761
+    },
+    {
+      "epoch": 68.4,
+      "learning_rate": 5.0758620689655175e-05,
+      "loss": 0.1406,
+      "step": 3762
+    },
+    {
+      "epoch": 68.42,
+      "learning_rate": 5.0620689655172414e-05,
+      "loss": 0.1867,
+      "step": 3763
+    },
+    {
+      "epoch": 68.44,
+      "learning_rate": 5.048275862068966e-05,
+      "loss": 0.1543,
+      "step": 3764
+    },
+    {
+      "epoch": 68.45,
+      "learning_rate": 5.03448275862069e-05,
+      "loss": 0.1445,
+      "step": 3765
+    },
+    {
+      "epoch": 68.47,
+      "learning_rate": 5.020689655172414e-05,
+      "loss": 0.2419,
+      "step": 3766
+    },
+    {
+      "epoch": 68.49,
+      "learning_rate": 5.006896551724138e-05,
+      "loss": 0.2123,
+      "step": 3767
+    },
+    {
+      "epoch": 68.51,
+      "learning_rate": 4.993103448275862e-05,
+      "loss": 0.1903,
+      "step": 3768
+    },
+    {
+      "epoch": 68.53,
+      "learning_rate": 4.979310344827586e-05,
+      "loss": 0.1733,
+      "step": 3769
+    },
+    {
+      "epoch": 68.55,
+      "learning_rate": 4.96551724137931e-05,
+      "loss": 0.1747,
+      "step": 3770
+    },
+    {
+      "epoch": 68.56,
+      "learning_rate": 4.9517241379310346e-05,
+      "loss": 0.2397,
+      "step": 3771
+    },
+    {
+      "epoch": 68.58,
+      "learning_rate": 4.9379310344827584e-05,
+      "loss": 0.196,
+      "step": 3772
+    },
+    {
+      "epoch": 68.6,
+      "learning_rate": 4.924137931034483e-05,
+      "loss": 0.2041,
+      "step": 3773
+    },
+    {
+      "epoch": 68.62,
+      "learning_rate": 4.910344827586207e-05,
+      "loss": 0.1844,
+      "step": 3774
+    },
+    {
+      "epoch": 68.64,
+      "learning_rate": 4.8965517241379315e-05,
+      "loss": 0.2108,
+      "step": 3775
+    },
+    {
+      "epoch": 68.65,
+      "learning_rate": 4.8827586206896554e-05,
+      "loss": 0.1622,
+      "step": 3776
+    },
+    {
+      "epoch": 68.67,
+      "learning_rate": 4.868965517241379e-05,
+      "loss": 0.1894,
+      "step": 3777
+    },
+    {
+      "epoch": 68.69,
+      "learning_rate": 4.855172413793103e-05,
+      "loss": 0.1362,
+      "step": 3778
+    },
+    {
+      "epoch": 68.71,
+      "learning_rate": 4.841379310344828e-05,
+      "loss": 0.1896,
+      "step": 3779
+    },
+    {
+      "epoch": 68.73,
+      "learning_rate": 4.8275862068965517e-05,
+      "loss": 0.2302,
+      "step": 3780
+    },
+    {
+      "epoch": 68.75,
+      "learning_rate": 4.813793103448276e-05,
+      "loss": 0.1524,
+      "step": 3781
+    },
+    {
+      "epoch": 68.76,
+      "learning_rate": 4.8e-05,
+      "loss": 0.2052,
+      "step": 3782
+    },
+    {
+      "epoch": 68.78,
+      "learning_rate": 4.786206896551725e-05,
+      "loss": 0.1618,
+      "step": 3783
+    },
+    {
+      "epoch": 68.8,
+      "learning_rate": 4.7724137931034486e-05,
+      "loss": 0.159,
+      "step": 3784
+    },
+    {
+      "epoch": 68.82,
+      "learning_rate": 4.7586206896551725e-05,
+      "loss": 0.1863,
+      "step": 3785
+    },
+    {
+      "epoch": 68.84,
+      "learning_rate": 4.7448275862068964e-05,
+      "loss": 0.2292,
+      "step": 3786
+    },
+    {
+      "epoch": 68.85,
+      "learning_rate": 4.73103448275862e-05,
+      "loss": 0.1735,
+      "step": 3787
+    },
+    {
+      "epoch": 68.87,
+      "learning_rate": 4.717241379310345e-05,
+      "loss": 0.1871,
+      "step": 3788
+    },
+    {
+      "epoch": 68.89,
+      "learning_rate": 4.703448275862069e-05,
+      "loss": 0.1635,
+      "step": 3789
+    },
+    {
+      "epoch": 68.91,
+      "learning_rate": 4.689655172413793e-05,
+      "loss": 0.1129,
+      "step": 3790
+    },
+    {
+      "epoch": 68.93,
+      "learning_rate": 4.675862068965517e-05,
+      "loss": 0.1637,
+      "step": 3791
+    },
+    {
+      "epoch": 68.95,
+      "learning_rate": 4.662068965517242e-05,
+      "loss": 0.1793,
+      "step": 3792
+    },
+    {
+      "epoch": 68.96,
+      "learning_rate": 4.648275862068966e-05,
+      "loss": 0.1744,
+      "step": 3793
+    },
+    {
+      "epoch": 68.98,
+      "learning_rate": 4.6344827586206896e-05,
+      "loss": 0.1382,
+      "step": 3794
+    },
+    {
+      "epoch": 69.0,
+      "learning_rate": 4.6206896551724135e-05,
+      "loss": 0.1925,
+      "step": 3795
+    },
+    {
+      "epoch": 69.02,
+      "learning_rate": 4.606896551724138e-05,
+      "loss": 0.1815,
+      "step": 3796
+    },
+    {
+      "epoch": 69.04,
+      "learning_rate": 4.593103448275862e-05,
+      "loss": 0.181,
+      "step": 3797
+    },
+    {
+      "epoch": 69.05,
+      "learning_rate": 4.5793103448275865e-05,
+      "loss": 0.1799,
+      "step": 3798
+    },
+    {
+      "epoch": 69.07,
+      "learning_rate": 4.5655172413793104e-05,
+      "loss": 0.2041,
+      "step": 3799
+    },
+    {
+      "epoch": 69.09,
+      "learning_rate": 4.551724137931035e-05,
+      "loss": 0.2119,
+      "step": 3800
+    },
+    {
+      "epoch": 69.11,
+      "learning_rate": 4.537931034482759e-05,
+      "loss": 0.1529,
+      "step": 3801
+    },
+    {
+      "epoch": 69.13,
+      "learning_rate": 4.524137931034483e-05,
+      "loss": 0.2269,
+      "step": 3802
+    },
+    {
+      "epoch": 69.15,
+      "learning_rate": 4.510344827586207e-05,
+      "loss": 0.1596,
+      "step": 3803
+    },
+    {
+      "epoch": 69.16,
+      "learning_rate": 4.496551724137931e-05,
+      "loss": 0.2122,
+      "step": 3804
+    },
+    {
+      "epoch": 69.18,
+      "learning_rate": 4.482758620689655e-05,
+      "loss": 0.1931,
+      "step": 3805
+    },
+    {
+      "epoch": 69.2,
+      "learning_rate": 4.46896551724138e-05,
+      "loss": 0.1818,
+      "step": 3806
+    },
+    {
+      "epoch": 69.22,
+      "learning_rate": 4.4551724137931036e-05,
+      "loss": 0.1602,
+      "step": 3807
+    },
+    {
+      "epoch": 69.24,
+      "learning_rate": 4.441379310344828e-05,
+      "loss": 0.1955,
+      "step": 3808
+    },
+    {
+      "epoch": 69.25,
+      "learning_rate": 4.427586206896552e-05,
+      "loss": 0.1936,
+      "step": 3809
+    },
+    {
+      "epoch": 69.27,
+      "learning_rate": 4.413793103448275e-05,
+      "loss": 0.1897,
+      "step": 3810
+    },
+    {
+      "epoch": 69.29,
+      "learning_rate": 4.4e-05,
+      "loss": 0.1468,
+      "step": 3811
+    },
+    {
+      "epoch": 69.31,
+      "learning_rate": 4.386206896551724e-05,
+      "loss": 0.1955,
+      "step": 3812
+    },
+    {
+      "epoch": 69.33,
+      "learning_rate": 4.3724137931034484e-05,
+      "loss": 0.119,
+      "step": 3813
+    },
+    {
+      "epoch": 69.35,
+      "learning_rate": 4.358620689655172e-05,
+      "loss": 0.1575,
+      "step": 3814
+    },
+    {
+      "epoch": 69.36,
+      "learning_rate": 4.344827586206897e-05,
+      "loss": 0.2415,
+      "step": 3815
+    },
+    {
+      "epoch": 69.38,
+      "learning_rate": 4.331034482758621e-05,
+      "loss": 0.1882,
+      "step": 3816
+    },
+    {
+      "epoch": 69.4,
+      "learning_rate": 4.317241379310345e-05,
+      "loss": 0.2023,
+      "step": 3817
+    },
+    {
+      "epoch": 69.42,
+      "learning_rate": 4.303448275862069e-05,
+      "loss": 0.2021,
+      "step": 3818
+    },
+    {
+      "epoch": 69.44,
+      "learning_rate": 4.289655172413793e-05,
+      "loss": 0.1793,
+      "step": 3819
+    },
+    {
+      "epoch": 69.45,
+      "learning_rate": 4.275862068965517e-05,
+      "loss": 0.1727,
+      "step": 3820
+    },
+    {
+      "epoch": 69.47,
+      "learning_rate": 4.2620689655172416e-05,
+      "loss": 0.2026,
+      "step": 3821
+    },
+    {
+      "epoch": 69.49,
+      "learning_rate": 4.2482758620689655e-05,
+      "loss": 0.2417,
+      "step": 3822
+    },
+    {
+      "epoch": 69.51,
+      "learning_rate": 4.23448275862069e-05,
+      "loss": 0.2379,
+      "step": 3823
+    },
+    {
+      "epoch": 69.53,
+      "learning_rate": 4.220689655172414e-05,
+      "loss": 0.1531,
+      "step": 3824
+    },
+    {
+      "epoch": 69.55,
+      "learning_rate": 4.2068965517241385e-05,
+      "loss": 0.1656,
+      "step": 3825
+    },
+    {
+      "epoch": 69.56,
+      "learning_rate": 4.1931034482758624e-05,
+      "loss": 0.1611,
+      "step": 3826
+    },
+    {
+      "epoch": 69.58,
+      "learning_rate": 4.179310344827586e-05,
+      "loss": 0.1823,
+      "step": 3827
+    },
+    {
+      "epoch": 69.6,
+      "learning_rate": 4.16551724137931e-05,
+      "loss": 0.1583,
+      "step": 3828
+    },
+    {
+      "epoch": 69.62,
+      "learning_rate": 4.151724137931035e-05,
+      "loss": 0.1417,
+      "step": 3829
+    },
+    {
+      "epoch": 69.64,
+      "learning_rate": 4.1379310344827587e-05,
+      "loss": 0.1724,
+      "step": 3830
+    },
+    {
+      "epoch": 69.65,
+      "learning_rate": 4.1241379310344826e-05,
+      "loss": 0.1787,
+      "step": 3831
+    },
+    {
+      "epoch": 69.67,
+      "learning_rate": 4.110344827586207e-05,
+      "loss": 0.1285,
+      "step": 3832
+    },
+    {
+      "epoch": 69.69,
+      "learning_rate": 4.096551724137931e-05,
+      "loss": 0.1669,
+      "step": 3833
+    },
+    {
+      "epoch": 69.71,
+      "learning_rate": 4.0827586206896556e-05,
+      "loss": 0.2401,
+      "step": 3834
+    },
+    {
+      "epoch": 69.73,
+      "learning_rate": 4.068965517241379e-05,
+      "loss": 0.1677,
+      "step": 3835
+    },
+    {
+      "epoch": 69.75,
+      "learning_rate": 4.0551724137931034e-05,
+      "loss": 0.1304,
+      "step": 3836
+    },
+    {
+      "epoch": 69.76,
+      "learning_rate": 4.041379310344827e-05,
+      "loss": 0.1866,
+      "step": 3837
+    },
+    {
+      "epoch": 69.78,
+      "learning_rate": 4.027586206896552e-05,
+      "loss": 0.1357,
+      "step": 3838
+    },
+    {
+      "epoch": 69.8,
+      "learning_rate": 4.013793103448276e-05,
+      "loss": 0.1818,
+      "step": 3839
+    },
+    {
+      "epoch": 69.82,
+      "learning_rate": 4e-05,
+      "loss": 0.2384,
+      "step": 3840
+    },
+    {
+      "epoch": 69.84,
+      "learning_rate": 3.986206896551724e-05,
+      "loss": 0.1452,
+      "step": 3841
+    },
+    {
+      "epoch": 69.85,
+      "learning_rate": 3.972413793103449e-05,
+      "loss": 0.1834,
+      "step": 3842
+    },
+    {
+      "epoch": 69.87,
+      "learning_rate": 3.958620689655173e-05,
+      "loss": 0.1523,
+      "step": 3843
+    },
+    {
+      "epoch": 69.89,
+      "learning_rate": 3.9448275862068966e-05,
+      "loss": 0.1672,
+      "step": 3844
+    },
+    {
+      "epoch": 69.91,
+      "learning_rate": 3.9310344827586205e-05,
+      "loss": 0.1799,
+      "step": 3845
+    },
+    {
+      "epoch": 69.93,
+      "learning_rate": 3.917241379310345e-05,
+      "loss": 0.1736,
+      "step": 3846
+    },
+    {
+      "epoch": 69.95,
+      "learning_rate": 3.903448275862069e-05,
+      "loss": 0.1472,
+      "step": 3847
+    },
+    {
+      "epoch": 69.96,
+      "learning_rate": 3.8896551724137935e-05,
+      "loss": 0.1753,
+      "step": 3848
+    },
+    {
+      "epoch": 69.98,
+      "learning_rate": 3.8758620689655174e-05,
+      "loss": 0.165,
+      "step": 3849
+    },
+    {
+      "epoch": 70.0,
+      "learning_rate": 3.862068965517242e-05,
+      "loss": 0.1173,
+      "step": 3850
+    },
+    {
+      "epoch": 70.02,
+      "learning_rate": 3.848275862068966e-05,
+      "loss": 0.2227,
+      "step": 3851
+    },
+    {
+      "epoch": 70.04,
+      "learning_rate": 3.834482758620689e-05,
+      "loss": 0.1652,
+      "step": 3852
+    },
+    {
+      "epoch": 70.05,
+      "learning_rate": 3.820689655172414e-05,
+      "loss": 0.1697,
+      "step": 3853
+    },
+    {
+      "epoch": 70.07,
+      "learning_rate": 3.8068965517241376e-05,
+      "loss": 0.1891,
+      "step": 3854
+    },
+    {
+      "epoch": 70.09,
+      "learning_rate": 3.793103448275862e-05,
+      "loss": 0.1655,
+      "step": 3855
+    },
+    {
+      "epoch": 70.11,
+      "learning_rate": 3.779310344827586e-05,
+      "loss": 0.1942,
+      "step": 3856
+    },
+    {
+      "epoch": 70.13,
+      "learning_rate": 3.7655172413793106e-05,
+      "loss": 0.194,
+      "step": 3857
+    },
+    {
+      "epoch": 70.15,
+      "learning_rate": 3.7517241379310345e-05,
+      "loss": 0.1851,
+      "step": 3858
+    },
+    {
+      "epoch": 70.16,
+      "learning_rate": 3.737931034482759e-05,
+      "loss": 0.1687,
+      "step": 3859
+    },
+    {
+      "epoch": 70.18,
+      "learning_rate": 3.724137931034482e-05,
+      "loss": 0.1374,
+      "step": 3860
+    },
+    {
+      "epoch": 70.2,
+      "learning_rate": 3.710344827586207e-05,
+      "loss": 0.2039,
+      "step": 3861
+    },
+    {
+      "epoch": 70.22,
+      "learning_rate": 3.696551724137931e-05,
+      "loss": 0.1536,
+      "step": 3862
+    },
+    {
+      "epoch": 70.24,
+      "learning_rate": 3.6827586206896554e-05,
+      "loss": 0.212,
+      "step": 3863
+    },
+    {
+      "epoch": 70.25,
+      "learning_rate": 3.668965517241379e-05,
+      "loss": 0.1843,
+      "step": 3864
+    },
+    {
+      "epoch": 70.27,
+      "learning_rate": 3.655172413793104e-05,
+      "loss": 0.2156,
+      "step": 3865
+    },
+    {
+      "epoch": 70.29,
+      "learning_rate": 3.641379310344828e-05,
+      "loss": 0.1923,
+      "step": 3866
+    },
+    {
+      "epoch": 70.31,
+      "learning_rate": 3.627586206896552e-05,
+      "loss": 0.1486,
+      "step": 3867
+    },
+    {
+      "epoch": 70.33,
+      "learning_rate": 3.613793103448276e-05,
+      "loss": 0.1755,
+      "step": 3868
+    },
+    {
+      "epoch": 70.35,
+      "learning_rate": 3.6e-05,
+      "loss": 0.1795,
+      "step": 3869
+    },
+    {
+      "epoch": 70.36,
+      "learning_rate": 3.586206896551724e-05,
+      "loss": 0.2035,
+      "step": 3870
+    },
+    {
+      "epoch": 70.38,
+      "learning_rate": 3.5724137931034486e-05,
+      "loss": 0.1778,
+      "step": 3871
+    },
+    {
+      "epoch": 70.4,
+      "learning_rate": 3.5586206896551725e-05,
+      "loss": 0.2259,
+      "step": 3872
+    },
+    {
+      "epoch": 70.42,
+      "learning_rate": 3.5448275862068964e-05,
+      "loss": 0.1831,
+      "step": 3873
+    },
+    {
+      "epoch": 70.44,
+      "learning_rate": 3.531034482758621e-05,
+      "loss": 0.1304,
+      "step": 3874
+    },
+    {
+      "epoch": 70.45,
+      "learning_rate": 3.517241379310345e-05,
+      "loss": 0.1738,
+      "step": 3875
+    },
+    {
+      "epoch": 70.47,
+      "learning_rate": 3.5034482758620694e-05,
+      "loss": 0.2492,
+      "step": 3876
+    },
+    {
+      "epoch": 70.49,
+      "learning_rate": 3.4896551724137926e-05,
+      "loss": 0.2187,
+      "step": 3877
+    },
+    {
+      "epoch": 70.51,
+      "learning_rate": 3.475862068965517e-05,
+      "loss": 0.1774,
+      "step": 3878
+    },
+    {
+      "epoch": 70.53,
+      "learning_rate": 3.462068965517241e-05,
+      "loss": 0.155,
+      "step": 3879
+    },
+    {
+      "epoch": 70.55,
+      "learning_rate": 3.4482758620689657e-05,
+      "loss": 0.1853,
+      "step": 3880
+    },
+    {
+      "epoch": 70.56,
+      "learning_rate": 3.4344827586206896e-05,
+      "loss": 0.1257,
+      "step": 3881
+    },
+    {
+      "epoch": 70.58,
+      "learning_rate": 3.420689655172414e-05,
+      "loss": 0.1888,
+      "step": 3882
+    },
+    {
+      "epoch": 70.6,
+      "learning_rate": 3.406896551724138e-05,
+      "loss": 0.1705,
+      "step": 3883
+    },
+    {
+      "epoch": 70.62,
+      "learning_rate": 3.3931034482758626e-05,
+      "loss": 0.1937,
+      "step": 3884
+    },
+    {
+      "epoch": 70.64,
+      "learning_rate": 3.379310344827586e-05,
+      "loss": 0.1525,
+      "step": 3885
+    },
+    {
+      "epoch": 70.65,
+      "learning_rate": 3.3655172413793104e-05,
+      "loss": 0.1536,
+      "step": 3886
+    },
+    {
+      "epoch": 70.67,
+      "learning_rate": 3.351724137931034e-05,
+      "loss": 0.1741,
+      "step": 3887
+    },
+    {
+      "epoch": 70.69,
+      "learning_rate": 3.337931034482759e-05,
+      "loss": 0.1372,
+      "step": 3888
+    },
+    {
+      "epoch": 70.71,
+      "learning_rate": 3.324137931034483e-05,
+      "loss": 0.171,
+      "step": 3889
+    },
+    {
+      "epoch": 70.73,
+      "learning_rate": 3.310344827586207e-05,
+      "loss": 0.1593,
+      "step": 3890
+    },
+    {
+      "epoch": 70.75,
+      "learning_rate": 3.296551724137931e-05,
+      "loss": 0.1426,
+      "step": 3891
+    },
+    {
+      "epoch": 70.76,
+      "learning_rate": 3.282758620689656e-05,
+      "loss": 0.2056,
+      "step": 3892
+    },
+    {
+      "epoch": 70.78,
+      "learning_rate": 3.26896551724138e-05,
+      "loss": 0.2209,
+      "step": 3893
+    },
+    {
+      "epoch": 70.8,
+      "learning_rate": 3.255172413793103e-05,
+      "loss": 0.1783,
+      "step": 3894
+    },
+    {
+      "epoch": 70.82,
+      "learning_rate": 3.2413793103448275e-05,
+      "loss": 0.1843,
+      "step": 3895
+    },
+    {
+      "epoch": 70.84,
+      "learning_rate": 3.2275862068965514e-05,
+      "loss": 0.2195,
+      "step": 3896
+    },
+    {
+      "epoch": 70.85,
+      "learning_rate": 3.213793103448276e-05,
+      "loss": 0.1468,
+      "step": 3897
+    },
+    {
+      "epoch": 70.87,
+      "learning_rate": 3.2e-05,
+      "loss": 0.1677,
+      "step": 3898
+    },
+    {
+      "epoch": 70.89,
+      "learning_rate": 3.1862068965517244e-05,
+      "loss": 0.1773,
+      "step": 3899
+    },
+    {
+      "epoch": 70.91,
+      "learning_rate": 3.172413793103448e-05,
+      "loss": 0.175,
+      "step": 3900
+    },
+    {
+      "epoch": 70.93,
+      "learning_rate": 3.158620689655173e-05,
+      "loss": 0.1799,
+      "step": 3901
+    },
+    {
+      "epoch": 70.95,
+      "learning_rate": 3.144827586206896e-05,
+      "loss": 0.1739,
+      "step": 3902
+    },
+    {
+      "epoch": 70.96,
+      "learning_rate": 3.131034482758621e-05,
+      "loss": 0.1492,
+      "step": 3903
+    },
+    {
+      "epoch": 70.98,
+      "learning_rate": 3.1172413793103446e-05,
+      "loss": 0.1228,
+      "step": 3904
+    },
+    {
+      "epoch": 71.0,
+      "learning_rate": 3.103448275862069e-05,
+      "loss": 0.1043,
+      "step": 3905
+    },
+    {
+      "epoch": 71.02,
+      "learning_rate": 3.089655172413793e-05,
+      "loss": 0.1918,
+      "step": 3906
+    },
+    {
+      "epoch": 71.04,
+      "learning_rate": 3.0758620689655176e-05,
+      "loss": 0.1499,
+      "step": 3907
+    },
+    {
+      "epoch": 71.05,
+      "learning_rate": 3.0620689655172415e-05,
+      "loss": 0.2009,
+      "step": 3908
+    },
+    {
+      "epoch": 71.07,
+      "learning_rate": 3.0482758620689654e-05,
+      "loss": 0.1657,
+      "step": 3909
+    },
+    {
+      "epoch": 71.09,
+      "learning_rate": 3.0344827586206897e-05,
+      "loss": 0.2425,
+      "step": 3910
+    },
+    {
+      "epoch": 71.11,
+      "learning_rate": 3.020689655172414e-05,
+      "loss": 0.159,
+      "step": 3911
+    },
+    {
+      "epoch": 71.13,
+      "learning_rate": 3.0068965517241378e-05,
+      "loss": 0.2151,
+      "step": 3912
+    },
+    {
+      "epoch": 71.15,
+      "learning_rate": 2.993103448275862e-05,
+      "loss": 0.1735,
+      "step": 3913
+    },
+    {
+      "epoch": 71.16,
+      "learning_rate": 2.9793103448275863e-05,
+      "loss": 0.1979,
+      "step": 3914
+    },
+    {
+      "epoch": 71.18,
+      "learning_rate": 2.9655172413793105e-05,
+      "loss": 0.1833,
+      "step": 3915
+    },
+    {
+      "epoch": 71.2,
+      "learning_rate": 2.9517241379310344e-05,
+      "loss": 0.2176,
+      "step": 3916
+    },
+    {
+      "epoch": 71.22,
+      "learning_rate": 2.9379310344827586e-05,
+      "loss": 0.1106,
+      "step": 3917
+    },
+    {
+      "epoch": 71.24,
+      "learning_rate": 2.924137931034483e-05,
+      "loss": 0.1988,
+      "step": 3918
+    },
+    {
+      "epoch": 71.25,
+      "learning_rate": 2.910344827586207e-05,
+      "loss": 0.2073,
+      "step": 3919
+    },
+    {
+      "epoch": 71.27,
+      "learning_rate": 2.8965517241379313e-05,
+      "loss": 0.1227,
+      "step": 3920
+    },
+    {
+      "epoch": 71.29,
+      "learning_rate": 2.8827586206896552e-05,
+      "loss": 0.2163,
+      "step": 3921
+    },
+    {
+      "epoch": 71.31,
+      "learning_rate": 2.8689655172413795e-05,
+      "loss": 0.1659,
+      "step": 3922
+    },
+    {
+      "epoch": 71.33,
+      "learning_rate": 2.8551724137931037e-05,
+      "loss": 0.1845,
+      "step": 3923
+    },
+    {
+      "epoch": 71.35,
+      "learning_rate": 2.841379310344828e-05,
+      "loss": 0.1382,
+      "step": 3924
+    },
+    {
+      "epoch": 71.36,
+      "learning_rate": 2.8275862068965518e-05,
+      "loss": 0.1623,
+      "step": 3925
+    },
+    {
+      "epoch": 71.38,
+      "learning_rate": 2.813793103448276e-05,
+      "loss": 0.1756,
+      "step": 3926
+    },
+    {
+      "epoch": 71.4,
+      "learning_rate": 2.8e-05,
+      "loss": 0.1717,
+      "step": 3927
+    },
+    {
+      "epoch": 71.42,
+      "learning_rate": 2.7862068965517242e-05,
+      "loss": 0.223,
+      "step": 3928
+    },
+    {
+      "epoch": 71.44,
+      "learning_rate": 2.772413793103448e-05,
+      "loss": 0.1961,
+      "step": 3929
+    },
+    {
+      "epoch": 71.45,
+      "learning_rate": 2.7586206896551723e-05,
+      "loss": 0.1562,
+      "step": 3930
+    },
+    {
+      "epoch": 71.47,
+      "learning_rate": 2.7448275862068966e-05,
+      "loss": 0.2419,
+      "step": 3931
+    },
+    {
+      "epoch": 71.49,
+      "learning_rate": 2.7310344827586208e-05,
+      "loss": 0.2261,
+      "step": 3932
+    },
+    {
+      "epoch": 71.51,
+      "learning_rate": 2.7172413793103447e-05,
+      "loss": 0.1884,
+      "step": 3933
+    },
+    {
+      "epoch": 71.53,
+      "learning_rate": 2.703448275862069e-05,
+      "loss": 0.1634,
+      "step": 3934
+    },
+    {
+      "epoch": 71.55,
+      "learning_rate": 2.689655172413793e-05,
+      "loss": 0.1867,
+      "step": 3935
+    },
+    {
+      "epoch": 71.56,
+      "learning_rate": 2.6758620689655174e-05,
+      "loss": 0.1494,
+      "step": 3936
+    },
+    {
+      "epoch": 71.58,
+      "learning_rate": 2.6620689655172413e-05,
+      "loss": 0.1846,
+      "step": 3937
+    },
+    {
+      "epoch": 71.6,
+      "learning_rate": 2.6482758620689655e-05,
+      "loss": 0.193,
+      "step": 3938
+    },
+    {
+      "epoch": 71.62,
+      "learning_rate": 2.6344827586206898e-05,
+      "loss": 0.1333,
+      "step": 3939
+    },
+    {
+      "epoch": 71.64,
+      "learning_rate": 2.620689655172414e-05,
+      "loss": 0.1791,
+      "step": 3940
+    },
+    {
+      "epoch": 71.65,
+      "learning_rate": 2.606896551724138e-05,
+      "loss": 0.1255,
+      "step": 3941
+    },
+    {
+      "epoch": 71.67,
+      "learning_rate": 2.593103448275862e-05,
+      "loss": 0.1246,
+      "step": 3942
+    },
+    {
+      "epoch": 71.69,
+      "learning_rate": 2.5793103448275864e-05,
+      "loss": 0.1953,
+      "step": 3943
+    },
+    {
+      "epoch": 71.71,
+      "learning_rate": 2.5655172413793106e-05,
+      "loss": 0.1796,
+      "step": 3944
+    },
+    {
+      "epoch": 71.73,
+      "learning_rate": 2.551724137931035e-05,
+      "loss": 0.155,
+      "step": 3945
+    },
+    {
+      "epoch": 71.75,
+      "learning_rate": 2.5379310344827587e-05,
+      "loss": 0.1601,
+      "step": 3946
+    },
+    {
+      "epoch": 71.76,
+      "learning_rate": 2.524137931034483e-05,
+      "loss": 0.1668,
+      "step": 3947
+    },
+    {
+      "epoch": 71.78,
+      "learning_rate": 2.510344827586207e-05,
+      "loss": 0.2056,
+      "step": 3948
+    },
+    {
+      "epoch": 71.8,
+      "learning_rate": 2.496551724137931e-05,
+      "loss": 0.1264,
+      "step": 3949
+    },
+    {
+      "epoch": 71.82,
+      "learning_rate": 2.482758620689655e-05,
+      "loss": 0.2001,
+      "step": 3950
+    },
+    {
+      "epoch": 71.84,
+      "learning_rate": 2.4689655172413792e-05,
+      "loss": 0.2439,
+      "step": 3951
+    },
+    {
+      "epoch": 71.85,
+      "learning_rate": 2.4551724137931035e-05,
+      "loss": 0.1906,
+      "step": 3952
+    },
+    {
+      "epoch": 71.87,
+      "learning_rate": 2.4413793103448277e-05,
+      "loss": 0.1584,
+      "step": 3953
+    },
+    {
+      "epoch": 71.89,
+      "learning_rate": 2.4275862068965516e-05,
+      "loss": 0.1608,
+      "step": 3954
+    },
+    {
+      "epoch": 71.91,
+      "learning_rate": 2.4137931034482758e-05,
+      "loss": 0.1884,
+      "step": 3955
+    },
+    {
+      "epoch": 71.93,
+      "learning_rate": 2.4e-05,
+      "loss": 0.197,
+      "step": 3956
+    },
+    {
+      "epoch": 71.95,
+      "learning_rate": 2.3862068965517243e-05,
+      "loss": 0.1721,
+      "step": 3957
+    },
+    {
+      "epoch": 71.96,
+      "learning_rate": 2.3724137931034482e-05,
+      "loss": 0.2049,
+      "step": 3958
+    },
+    {
+      "epoch": 71.98,
+      "learning_rate": 2.3586206896551724e-05,
+      "loss": 0.1916,
+      "step": 3959
+    },
+    {
+      "epoch": 72.0,
+      "learning_rate": 2.3448275862068967e-05,
+      "loss": 0.2269,
+      "step": 3960
+    },
+    {
+      "epoch": 72.02,
+      "learning_rate": 2.331034482758621e-05,
+      "loss": 0.1919,
+      "step": 3961
+    },
+    {
+      "epoch": 72.04,
+      "learning_rate": 2.3172413793103448e-05,
+      "loss": 0.2066,
+      "step": 3962
+    },
+    {
+      "epoch": 72.05,
+      "learning_rate": 2.303448275862069e-05,
+      "loss": 0.1602,
+      "step": 3963
+    },
+    {
+      "epoch": 72.07,
+      "learning_rate": 2.2896551724137933e-05,
+      "loss": 0.1697,
+      "step": 3964
+    },
+    {
+      "epoch": 72.09,
+      "learning_rate": 2.2758620689655175e-05,
+      "loss": 0.2154,
+      "step": 3965
+    },
+    {
+      "epoch": 72.11,
+      "learning_rate": 2.2620689655172414e-05,
+      "loss": 0.1889,
+      "step": 3966
+    },
+    {
+      "epoch": 72.13,
+      "learning_rate": 2.2482758620689656e-05,
+      "loss": 0.1565,
+      "step": 3967
+    },
+    {
+      "epoch": 72.15,
+      "learning_rate": 2.23448275862069e-05,
+      "loss": 0.1572,
+      "step": 3968
+    },
+    {
+      "epoch": 72.16,
+      "learning_rate": 2.220689655172414e-05,
+      "loss": 0.1756,
+      "step": 3969
+    },
+    {
+      "epoch": 72.18,
+      "learning_rate": 2.2068965517241377e-05,
+      "loss": 0.1459,
+      "step": 3970
+    },
+    {
+      "epoch": 72.2,
+      "learning_rate": 2.193103448275862e-05,
+      "loss": 0.1671,
+      "step": 3971
+    },
+    {
+      "epoch": 72.22,
+      "learning_rate": 2.179310344827586e-05,
+      "loss": 0.166,
+      "step": 3972
+    },
+    {
+      "epoch": 72.24,
+      "learning_rate": 2.1655172413793104e-05,
+      "loss": 0.2289,
+      "step": 3973
+    },
+    {
+      "epoch": 72.25,
+      "learning_rate": 2.1517241379310346e-05,
+      "loss": 0.1781,
+      "step": 3974
+    },
+    {
+      "epoch": 72.27,
+      "learning_rate": 2.1379310344827585e-05,
+      "loss": 0.1633,
+      "step": 3975
+    },
+    {
+      "epoch": 72.29,
+      "learning_rate": 2.1241379310344827e-05,
+      "loss": 0.1462,
+      "step": 3976
+    },
+    {
+      "epoch": 72.31,
+      "learning_rate": 2.110344827586207e-05,
+      "loss": 0.131,
+      "step": 3977
+    },
+    {
+      "epoch": 72.33,
+      "learning_rate": 2.0965517241379312e-05,
+      "loss": 0.1773,
+      "step": 3978
+    },
+    {
+      "epoch": 72.35,
+      "learning_rate": 2.082758620689655e-05,
+      "loss": 0.1499,
+      "step": 3979
+    },
+    {
+      "epoch": 72.36,
+      "learning_rate": 2.0689655172413793e-05,
+      "loss": 0.1911,
+      "step": 3980
+    },
+    {
+      "epoch": 72.38,
+      "learning_rate": 2.0551724137931036e-05,
+      "loss": 0.2211,
+      "step": 3981
+    },
+    {
+      "epoch": 72.4,
+      "learning_rate": 2.0413793103448278e-05,
+      "loss": 0.1891,
+      "step": 3982
+    },
+    {
+      "epoch": 72.42,
+      "learning_rate": 2.0275862068965517e-05,
+      "loss": 0.2047,
+      "step": 3983
+    },
+    {
+      "epoch": 72.44,
+      "learning_rate": 2.013793103448276e-05,
+      "loss": 0.1351,
+      "step": 3984
+    },
+    {
+      "epoch": 72.45,
+      "learning_rate": 2e-05,
+      "loss": 0.1481,
+      "step": 3985
+    },
+    {
+      "epoch": 72.47,
+      "learning_rate": 1.9862068965517244e-05,
+      "loss": 0.1833,
+      "step": 3986
+    },
+    {
+      "epoch": 72.49,
+      "learning_rate": 1.9724137931034483e-05,
+      "loss": 0.1826,
+      "step": 3987
+    },
+    {
+      "epoch": 72.51,
+      "learning_rate": 1.9586206896551725e-05,
+      "loss": 0.1998,
+      "step": 3988
+    },
+    {
+      "epoch": 72.53,
+      "learning_rate": 1.9448275862068968e-05,
+      "loss": 0.1466,
+      "step": 3989
+    },
+    {
+      "epoch": 72.55,
+      "learning_rate": 1.931034482758621e-05,
+      "loss": 0.136,
+      "step": 3990
+    },
+    {
+      "epoch": 72.56,
+      "learning_rate": 1.9172413793103446e-05,
+      "loss": 0.1555,
+      "step": 3991
+    },
+    {
+      "epoch": 72.58,
+      "learning_rate": 1.9034482758620688e-05,
+      "loss": 0.1622,
+      "step": 3992
+    },
+    {
+      "epoch": 72.6,
+      "learning_rate": 1.889655172413793e-05,
+      "loss": 0.2148,
+      "step": 3993
+    },
+    {
+      "epoch": 72.62,
+      "learning_rate": 1.8758620689655173e-05,
+      "loss": 0.1654,
+      "step": 3994
+    },
+    {
+      "epoch": 72.64,
+      "learning_rate": 1.862068965517241e-05,
+      "loss": 0.1725,
+      "step": 3995
+    },
+    {
+      "epoch": 72.65,
+      "learning_rate": 1.8482758620689654e-05,
+      "loss": 0.1682,
+      "step": 3996
+    },
+    {
+      "epoch": 72.67,
+      "learning_rate": 1.8344827586206896e-05,
+      "loss": 0.1346,
+      "step": 3997
+    },
+    {
+      "epoch": 72.69,
+      "learning_rate": 1.820689655172414e-05,
+      "loss": 0.1302,
+      "step": 3998
+    },
+    {
+      "epoch": 72.71,
+      "learning_rate": 1.806896551724138e-05,
+      "loss": 0.1572,
+      "step": 3999
+    },
+    {
+      "epoch": 72.73,
+      "learning_rate": 1.793103448275862e-05,
+      "loss": 0.2041,
+      "step": 4000
+    },
+    {
+      "epoch": 72.73,
+      "eval_cer": 0.0776023628540158,
+      "eval_loss": 0.3853955566883087,
+      "eval_runtime": 14.1398,
+      "eval_samples_per_second": 116.48,
+      "eval_steps_per_second": 1.839,
+      "eval_wer": 0.32335818608926564,
+      "step": 4000
+    },
+    {
+      "epoch": 72.75,
+      "learning_rate": 1.7793103448275862e-05,
+      "loss": 0.1399,
+      "step": 4001
+    },
+    {
+      "epoch": 72.76,
+      "learning_rate": 1.7655172413793105e-05,
+      "loss": 0.1633,
+      "step": 4002
+    },
+    {
+      "epoch": 72.78,
+      "learning_rate": 1.7517241379310347e-05,
+      "loss": 0.1807,
+      "step": 4003
+    },
+    {
+      "epoch": 72.8,
+      "learning_rate": 1.7379310344827586e-05,
+      "loss": 0.1683,
+      "step": 4004
+    },
+    {
+      "epoch": 72.82,
+      "learning_rate": 1.7241379310344828e-05,
+      "loss": 0.1856,
+      "step": 4005
+    },
+    {
+      "epoch": 72.84,
+      "learning_rate": 1.710344827586207e-05,
+      "loss": 0.1631,
+      "step": 4006
+    },
+    {
+      "epoch": 72.85,
+      "learning_rate": 1.6965517241379313e-05,
+      "loss": 0.1816,
+      "step": 4007
+    },
+    {
+      "epoch": 72.87,
+      "learning_rate": 1.6827586206896552e-05,
+      "loss": 0.1687,
+      "step": 4008
+    },
+    {
+      "epoch": 72.89,
+      "learning_rate": 1.6689655172413794e-05,
+      "loss": 0.1818,
+      "step": 4009
+    },
+    {
+      "epoch": 72.91,
+      "learning_rate": 1.6551724137931037e-05,
+      "loss": 0.1641,
+      "step": 4010
+    },
+    {
+      "epoch": 72.93,
+      "learning_rate": 1.641379310344828e-05,
+      "loss": 0.1948,
+      "step": 4011
+    },
+    {
+      "epoch": 72.95,
+      "learning_rate": 1.6275862068965515e-05,
+      "loss": 0.1887,
+      "step": 4012
+    },
+    {
+      "epoch": 72.96,
+      "learning_rate": 1.6137931034482757e-05,
+      "loss": 0.1821,
+      "step": 4013
+    },
+    {
+      "epoch": 72.98,
+      "learning_rate": 1.6e-05,
+      "loss": 0.1342,
+      "step": 4014
+    },
+    {
+      "epoch": 73.0,
+      "learning_rate": 1.586206896551724e-05,
+      "loss": 0.1972,
+      "step": 4015
+    },
+    {
+      "epoch": 73.02,
+      "learning_rate": 1.572413793103448e-05,
+      "loss": 0.1601,
+      "step": 4016
+    },
+    {
+      "epoch": 73.04,
+      "learning_rate": 1.5586206896551723e-05,
+      "loss": 0.1482,
+      "step": 4017
+    },
+    {
+      "epoch": 73.05,
+      "learning_rate": 1.5448275862068965e-05,
+      "loss": 0.1575,
+      "step": 4018
+    },
+    {
+      "epoch": 73.07,
+      "learning_rate": 1.5310344827586208e-05,
+      "loss": 0.2248,
+      "step": 4019
+    },
+    {
+      "epoch": 73.09,
+      "learning_rate": 1.5172413793103448e-05,
+      "loss": 0.1503,
+      "step": 4020
+    },
+    {
+      "epoch": 73.11,
+      "learning_rate": 1.5034482758620689e-05,
+      "loss": 0.192,
+      "step": 4021
+    },
+    {
+      "epoch": 73.13,
+      "learning_rate": 1.4896551724137931e-05,
+      "loss": 0.2242,
+      "step": 4022
+    },
+    {
+      "epoch": 73.15,
+      "learning_rate": 1.4758620689655172e-05,
+      "loss": 0.1587,
+      "step": 4023
+    },
+    {
+      "epoch": 73.16,
+      "learning_rate": 1.4620689655172414e-05,
+      "loss": 0.1643,
+      "step": 4024
+    },
+    {
+      "epoch": 73.18,
+      "learning_rate": 1.4482758620689657e-05,
+      "loss": 0.1743,
+      "step": 4025
+    },
+    {
+      "epoch": 73.2,
+      "learning_rate": 1.4344827586206897e-05,
+      "loss": 0.1553,
+      "step": 4026
+    },
+    {
+      "epoch": 73.22,
+      "learning_rate": 1.420689655172414e-05,
+      "loss": 0.2206,
+      "step": 4027
+    },
+    {
+      "epoch": 73.24,
+      "learning_rate": 1.406896551724138e-05,
+      "loss": 0.1631,
+      "step": 4028
+    },
+    {
+      "epoch": 73.25,
+      "learning_rate": 1.3931034482758621e-05,
+      "loss": 0.1537,
+      "step": 4029
+    },
+    {
+      "epoch": 73.27,
+      "learning_rate": 1.3793103448275862e-05,
+      "loss": 0.1609,
+      "step": 4030
+    },
+    {
+      "epoch": 73.29,
+      "learning_rate": 1.3655172413793104e-05,
+      "loss": 0.2013,
+      "step": 4031
+    },
+    {
+      "epoch": 73.31,
+      "learning_rate": 1.3517241379310345e-05,
+      "loss": 0.15,
+      "step": 4032
+    },
+    {
+      "epoch": 73.33,
+      "learning_rate": 1.3379310344827587e-05,
+      "loss": 0.1926,
+      "step": 4033
+    },
+    {
+      "epoch": 73.35,
+      "learning_rate": 1.3241379310344828e-05,
+      "loss": 0.1658,
+      "step": 4034
+    },
+    {
+      "epoch": 73.36,
+      "learning_rate": 1.310344827586207e-05,
+      "loss": 0.1728,
+      "step": 4035
+    },
+    {
+      "epoch": 73.38,
+      "learning_rate": 1.296551724137931e-05,
+      "loss": 0.1773,
+      "step": 4036
+    },
+    {
+      "epoch": 73.4,
+      "learning_rate": 1.2827586206896553e-05,
+      "loss": 0.1645,
+      "step": 4037
+    },
+    {
+      "epoch": 73.42,
+      "learning_rate": 1.2689655172413794e-05,
+      "loss": 0.201,
+      "step": 4038
+    },
+    {
+      "epoch": 73.44,
+      "learning_rate": 1.2551724137931034e-05,
+      "loss": 0.1275,
+      "step": 4039
+    },
+    {
+      "epoch": 73.45,
+      "learning_rate": 1.2413793103448275e-05,
+      "loss": 0.1641,
+      "step": 4040
+    },
+    {
+      "epoch": 73.47,
+      "learning_rate": 1.2275862068965517e-05,
+      "loss": 0.2088,
+      "step": 4041
+    },
+    {
+      "epoch": 73.49,
+      "learning_rate": 1.2137931034482758e-05,
+      "loss": 0.1944,
+      "step": 4042
+    },
+    {
+      "epoch": 73.51,
+      "learning_rate": 1.2e-05,
+      "loss": 0.1868,
+      "step": 4043
+    },
+    {
+      "epoch": 73.53,
+      "learning_rate": 1.1862068965517241e-05,
+      "loss": 0.1894,
+      "step": 4044
+    },
+    {
+      "epoch": 73.55,
+      "learning_rate": 1.1724137931034483e-05,
+      "loss": 0.1769,
+      "step": 4045
+    },
+    {
+      "epoch": 73.56,
+      "learning_rate": 1.1586206896551724e-05,
+      "loss": 0.1764,
+      "step": 4046
+    },
+    {
+      "epoch": 73.58,
+      "learning_rate": 1.1448275862068966e-05,
+      "loss": 0.2064,
+      "step": 4047
+    },
+    {
+      "epoch": 73.6,
+      "learning_rate": 1.1310344827586207e-05,
+      "loss": 0.1742,
+      "step": 4048
+    },
+    {
+      "epoch": 73.62,
+      "learning_rate": 1.117241379310345e-05,
+      "loss": 0.1706,
+      "step": 4049
+    },
+    {
+      "epoch": 73.64,
+      "learning_rate": 1.1034482758620688e-05,
+      "loss": 0.1947,
+      "step": 4050
+    },
+    {
+      "epoch": 73.65,
+      "learning_rate": 1.089655172413793e-05,
+      "loss": 0.1364,
+      "step": 4051
+    },
+    {
+      "epoch": 73.67,
+      "learning_rate": 1.0758620689655173e-05,
+      "loss": 0.1311,
+      "step": 4052
+    },
+    {
+      "epoch": 73.69,
+      "learning_rate": 1.0620689655172414e-05,
+      "loss": 0.2123,
+      "step": 4053
+    },
+    {
+      "epoch": 73.71,
+      "learning_rate": 1.0482758620689656e-05,
+      "loss": 0.1789,
+      "step": 4054
+    },
+    {
+      "epoch": 73.73,
+      "learning_rate": 1.0344827586206897e-05,
+      "loss": 0.2293,
+      "step": 4055
+    },
+    {
+      "epoch": 73.75,
+      "learning_rate": 1.0206896551724139e-05,
+      "loss": 0.1596,
+      "step": 4056
+    },
+    {
+      "epoch": 73.76,
+      "learning_rate": 1.006896551724138e-05,
+      "loss": 0.1375,
+      "step": 4057
+    },
+    {
+      "epoch": 73.78,
+      "learning_rate": 9.931034482758622e-06,
+      "loss": 0.1698,
+      "step": 4058
+    },
+    {
+      "epoch": 73.8,
+      "learning_rate": 9.793103448275863e-06,
+      "loss": 0.1726,
+      "step": 4059
+    },
+    {
+      "epoch": 73.82,
+      "learning_rate": 9.655172413793105e-06,
+      "loss": 0.1863,
+      "step": 4060
+    },
+    {
+      "epoch": 73.84,
+      "learning_rate": 9.517241379310344e-06,
+      "loss": 0.176,
+      "step": 4061
+    },
+    {
+      "epoch": 73.85,
+      "learning_rate": 9.379310344827586e-06,
+      "loss": 0.1351,
+      "step": 4062
+    },
+    {
+      "epoch": 73.87,
+      "learning_rate": 9.241379310344827e-06,
+      "loss": 0.1787,
+      "step": 4063
+    },
+    {
+      "epoch": 73.89,
+      "learning_rate": 9.10344827586207e-06,
+      "loss": 0.1541,
+      "step": 4064
+    },
+    {
+      "epoch": 73.91,
+      "learning_rate": 8.96551724137931e-06,
+      "loss": 0.101,
+      "step": 4065
+    },
+    {
+      "epoch": 73.93,
+      "learning_rate": 8.827586206896552e-06,
+      "loss": 0.1631,
+      "step": 4066
+    },
+    {
+      "epoch": 73.95,
+      "learning_rate": 8.689655172413793e-06,
+      "loss": 0.1542,
+      "step": 4067
+    },
+    {
+      "epoch": 73.96,
+      "learning_rate": 8.551724137931035e-06,
+      "loss": 0.2086,
+      "step": 4068
+    },
+    {
+      "epoch": 73.98,
+      "learning_rate": 8.413793103448276e-06,
+      "loss": 0.1497,
+      "step": 4069
+    },
+    {
+      "epoch": 74.0,
+      "learning_rate": 8.275862068965518e-06,
+      "loss": 0.116,
+      "step": 4070
+    },
+    {
+      "epoch": 74.02,
+      "learning_rate": 8.137931034482757e-06,
+      "loss": 0.1992,
+      "step": 4071
+    },
+    {
+      "epoch": 74.04,
+      "learning_rate": 8e-06,
+      "loss": 0.1728,
+      "step": 4072
+    },
+    {
+      "epoch": 74.05,
+      "learning_rate": 7.86206896551724e-06,
+      "loss": 0.2187,
+      "step": 4073
+    },
+    {
+      "epoch": 74.07,
+      "learning_rate": 7.724137931034483e-06,
+      "loss": 0.1181,
+      "step": 4074
+    },
+    {
+      "epoch": 74.09,
+      "learning_rate": 7.586206896551724e-06,
+      "loss": 0.2273,
+      "step": 4075
+    },
+    {
+      "epoch": 74.11,
+      "learning_rate": 7.448275862068966e-06,
+      "loss": 0.1964,
+      "step": 4076
+    },
+    {
+      "epoch": 74.13,
+      "learning_rate": 7.310344827586207e-06,
+      "loss": 0.1683,
+      "step": 4077
+    },
+    {
+      "epoch": 74.15,
+      "learning_rate": 7.172413793103449e-06,
+      "loss": 0.1731,
+      "step": 4078
+    },
+    {
+      "epoch": 74.16,
+      "learning_rate": 7.03448275862069e-06,
+      "loss": 0.1735,
+      "step": 4079
+    },
+    {
+      "epoch": 74.18,
+      "learning_rate": 6.896551724137931e-06,
+      "loss": 0.1613,
+      "step": 4080
+    },
+    {
+      "epoch": 74.2,
+      "learning_rate": 6.758620689655172e-06,
+      "loss": 0.166,
+      "step": 4081
+    },
+    {
+      "epoch": 74.22,
+      "learning_rate": 6.620689655172414e-06,
+      "loss": 0.1457,
+      "step": 4082
+    },
+    {
+      "epoch": 74.24,
+      "learning_rate": 6.482758620689655e-06,
+      "loss": 0.1408,
+      "step": 4083
+    },
+    {
+      "epoch": 74.25,
+      "learning_rate": 6.344827586206897e-06,
+      "loss": 0.1918,
+      "step": 4084
+    },
+    {
+      "epoch": 74.27,
+      "learning_rate": 6.2068965517241375e-06,
+      "loss": 0.2259,
+      "step": 4085
+    },
+    {
+      "epoch": 74.29,
+      "learning_rate": 6.068965517241379e-06,
+      "loss": 0.1618,
+      "step": 4086
+    },
+    {
+      "epoch": 74.31,
+      "learning_rate": 5.9310344827586205e-06,
+      "loss": 0.1894,
+      "step": 4087
+    },
+    {
+      "epoch": 74.33,
+      "learning_rate": 5.793103448275862e-06,
+      "loss": 0.1578,
+      "step": 4088
+    },
+    {
+      "epoch": 74.35,
+      "learning_rate": 5.6551724137931035e-06,
+      "loss": 0.1955,
+      "step": 4089
+    },
+    {
+      "epoch": 74.36,
+      "learning_rate": 5.517241379310344e-06,
+      "loss": 0.1746,
+      "step": 4090
+    },
+    {
+      "epoch": 74.38,
+      "learning_rate": 5.3793103448275865e-06,
+      "loss": 0.1933,
+      "step": 4091
+    },
+    {
+      "epoch": 74.4,
+      "learning_rate": 5.241379310344828e-06,
+      "loss": 0.1962,
+      "step": 4092
+    },
+    {
+      "epoch": 74.42,
+      "learning_rate": 5.1034482758620695e-06,
+      "loss": 0.19,
+      "step": 4093
+    },
+    {
+      "epoch": 74.44,
+      "learning_rate": 4.965517241379311e-06,
+      "loss": 0.2024,
+      "step": 4094
+    },
+    {
+      "epoch": 74.45,
+      "learning_rate": 4.8275862068965525e-06,
+      "loss": 0.1238,
+      "step": 4095
+    },
+    {
+      "epoch": 74.47,
+      "learning_rate": 4.689655172413793e-06,
+      "loss": 0.1947,
+      "step": 4096
+    },
+    {
+      "epoch": 74.49,
+      "learning_rate": 4.551724137931035e-06,
+      "loss": 0.1965,
+      "step": 4097
+    },
+    {
+      "epoch": 74.51,
+      "learning_rate": 4.413793103448276e-06,
+      "loss": 0.1662,
+      "step": 4098
+    },
+    {
+      "epoch": 74.53,
+      "learning_rate": 4.275862068965518e-06,
+      "loss": 0.1612,
+      "step": 4099
+    },
+    {
+      "epoch": 74.55,
+      "learning_rate": 4.137931034482759e-06,
+      "loss": 0.1909,
+      "step": 4100
+    },
+    {
+      "epoch": 74.56,
+      "learning_rate": 4e-06,
+      "loss": 0.1778,
+      "step": 4101
+    },
+    {
+      "epoch": 74.58,
+      "learning_rate": 3.862068965517241e-06,
+      "loss": 0.1615,
+      "step": 4102
+    },
+    {
+      "epoch": 74.6,
+      "learning_rate": 3.724137931034483e-06,
+      "loss": 0.1565,
+      "step": 4103
+    },
+    {
+      "epoch": 74.62,
+      "learning_rate": 3.5862068965517243e-06,
+      "loss": 0.1415,
+      "step": 4104
+    },
+    {
+      "epoch": 74.64,
+      "learning_rate": 3.4482758620689654e-06,
+      "loss": 0.1738,
+      "step": 4105
+    },
+    {
+      "epoch": 74.65,
+      "learning_rate": 3.310344827586207e-06,
+      "loss": 0.1582,
+      "step": 4106
+    },
+    {
+      "epoch": 74.67,
+      "learning_rate": 3.1724137931034484e-06,
+      "loss": 0.2142,
+      "step": 4107
+    },
+    {
+      "epoch": 74.69,
+      "learning_rate": 3.0344827586206895e-06,
+      "loss": 0.1648,
+      "step": 4108
+    },
+    {
+      "epoch": 74.71,
+      "learning_rate": 2.896551724137931e-06,
+      "loss": 0.2031,
+      "step": 4109
+    },
+    {
+      "epoch": 74.73,
+      "learning_rate": 2.758620689655172e-06,
+      "loss": 0.1839,
+      "step": 4110
+    },
+    {
+      "epoch": 74.75,
+      "learning_rate": 2.620689655172414e-06,
+      "loss": 0.1642,
+      "step": 4111
+    },
+    {
+      "epoch": 74.76,
+      "learning_rate": 2.4827586206896555e-06,
+      "loss": 0.1829,
+      "step": 4112
+    },
+    {
+      "epoch": 74.78,
+      "learning_rate": 2.3448275862068966e-06,
+      "loss": 0.1235,
+      "step": 4113
+    },
+    {
+      "epoch": 74.8,
+      "learning_rate": 2.206896551724138e-06,
+      "loss": 0.1747,
+      "step": 4114
+    },
+    {
+      "epoch": 74.82,
+      "learning_rate": 2.0689655172413796e-06,
+      "loss": 0.1866,
+      "step": 4115
+    },
+    {
+      "epoch": 74.84,
+      "learning_rate": 1.9310344827586207e-06,
+      "loss": 0.2032,
+      "step": 4116
+    },
+    {
+      "epoch": 74.85,
+      "learning_rate": 1.7931034482758622e-06,
+      "loss": 0.2196,
+      "step": 4117
+    },
+    {
+      "epoch": 74.87,
+      "learning_rate": 1.6551724137931035e-06,
+      "loss": 0.184,
+      "step": 4118
+    },
+    {
+      "epoch": 74.89,
+      "learning_rate": 1.5172413793103447e-06,
+      "loss": 0.1783,
+      "step": 4119
+    },
+    {
+      "epoch": 74.91,
+      "learning_rate": 1.379310344827586e-06,
+      "loss": 0.1289,
+      "step": 4120
+    },
+    {
+      "epoch": 74.93,
+      "learning_rate": 1.2413793103448277e-06,
+      "loss": 0.237,
+      "step": 4121
+    },
+    {
+      "epoch": 74.95,
+      "learning_rate": 1.103448275862069e-06,
+      "loss": 0.1485,
+      "step": 4122
+    },
+    {
+      "epoch": 74.96,
+      "learning_rate": 9.655172413793103e-07,
+      "loss": 0.1542,
+      "step": 4123
+    },
+    {
+      "epoch": 74.98,
+      "learning_rate": 8.275862068965517e-07,
+      "loss": 0.2093,
+      "step": 4124
+    },
+    {
+      "epoch": 75.0,
+      "learning_rate": 6.89655172413793e-07,
+      "loss": 0.1497,
+      "step": 4125
+    },
+    {
+      "epoch": 75.0,
+      "step": 4125,
+      "total_flos": 3.6285883659271012e+19,
+      "train_loss": 0.6802920068195372,
+      "train_runtime": 3123.2662,
+      "train_samples_per_second": 83.518,
+      "train_steps_per_second": 1.321
     }
   ],
-  "max_steps": 2750,
-  "num_train_epochs": 50,
-  "total_flos": 2.4190366418747785e+19,
+  "max_steps": 4125,
+  "num_train_epochs": 75,
+  "total_flos": 3.6285883659271012e+19,
   "trial_name": null,
   "trial_params": null
 }