diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,124035 @@
+{
+  "best_metric": 0.6265091344516582,
+  "best_model_checkpoint": "xtreme_s_xlsr_300m_fleurs_asr_western_european_nomask/checkpoint-19500",
+  "epoch": 20.0,
+  "global_step": 20600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 21.5708,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 20.8844,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-07,
+      "loss": 18.7852,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-07,
+      "loss": 19.3113,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-07,
+      "loss": 18.4036,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2e-06,
+      "loss": 17.4366,
+      "step": 6
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4999999999999998e-06,
+      "loss": 19.8087,
+      "step": 7
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8e-06,
+      "loss": 22.6603,
+      "step": 8
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1e-06,
+      "loss": 19.6383,
+      "step": 9
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4e-06,
+      "loss": 18.5671,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6999999999999996e-06,
+      "loss": 18.7439,
+      "step": 11
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9999999999999997e-06,
+      "loss": 17.8481,
+      "step": 12
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2999999999999993e-06,
+      "loss": 19.0226,
+      "step": 13
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6e-06,
+      "loss": 21.036,
+      "step": 14
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.899999999999999e-06,
+      "loss": 18.8691,
+      "step": 15
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.2e-06,
+      "loss": 19.516,
+      "step": 16
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.499999999999999e-06,
+      "loss": 18.347,
+      "step": 17
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.8e-06,
+      "loss": 17.4564,
+      "step": 18
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.1e-06,
+      "loss": 17.7358,
+      "step": 19
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.399999999999999e-06,
+      "loss": 21.3263,
+      "step": 20
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.7e-06,
+      "loss": 19.7947,
+      "step": 21
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.999999999999999e-06,
+      "loss": 18.9683,
+      "step": 22
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.3e-06,
+      "loss": 18.9511,
+      "step": 23
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.599999999999999e-06,
+      "loss": 16.7369,
+      "step": 24
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.899999999999999e-06,
+      "loss": 16.1944,
+      "step": 25
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 6.899999999999999e-06,
+      "loss": 21.5088,
+      "step": 26
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7.2e-06,
+      "loss": 20.1562,
+      "step": 27
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7.499999999999999e-06,
+      "loss": 20.1432,
+      "step": 28
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7.799999999999998e-06,
+      "loss": 18.3871,
+      "step": 29
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.099999999999999e-06,
+      "loss": 17.2953,
+      "step": 30
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.4e-06,
+      "loss": 17.8756,
+      "step": 31
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.7e-06,
+      "loss": 19.88,
+      "step": 32
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.999999999999999e-06,
+      "loss": 20.364,
+      "step": 33
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.299999999999999e-06,
+      "loss": 18.936,
+      "step": 34
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.6e-06,
+      "loss": 18.171,
+      "step": 35
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9e-06,
+      "loss": 18.195,
+      "step": 36
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.02e-05,
+      "loss": 17.0616,
+      "step": 37
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.05e-05,
+      "loss": 16.6902,
+      "step": 38
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.0799999999999998e-05,
+      "loss": 19.262,
+      "step": 39
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1099999999999999e-05,
+      "loss": 18.8784,
+      "step": 40
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.14e-05,
+      "loss": 17.5027,
+      "step": 41
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.17e-05,
+      "loss": 17.6119,
+      "step": 42
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1999999999999999e-05,
+      "loss": 15.8907,
+      "step": 43
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2299999999999999e-05,
+      "loss": 16.1014,
+      "step": 44
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.26e-05,
+      "loss": 20.8052,
+      "step": 45
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2899999999999998e-05,
+      "loss": 17.795,
+      "step": 46
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3199999999999997e-05,
+      "loss": 17.5306,
+      "step": 47
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3499999999999998e-05,
+      "loss": 15.9641,
+      "step": 48
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3799999999999998e-05,
+      "loss": 16.3233,
+      "step": 49
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.4099999999999999e-05,
+      "loss": 13.4692,
+      "step": 50
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.44e-05,
+      "loss": 16.9965,
+      "step": 51
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.47e-05,
+      "loss": 16.5989,
+      "step": 52
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 15.3964,
+      "step": 53
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.53e-05,
+      "loss": 14.8295,
+      "step": 54
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5599999999999996e-05,
+      "loss": 14.0894,
+      "step": 55
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5899999999999997e-05,
+      "loss": 12.8862,
+      "step": 56
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.6199999999999997e-05,
+      "loss": 14.8636,
+      "step": 57
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.6499999999999998e-05,
+      "loss": 14.5593,
+      "step": 58
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.68e-05,
+      "loss": 14.2551,
+      "step": 59
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.71e-05,
+      "loss": 12.7387,
+      "step": 60
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.74e-05,
+      "loss": 13.0697,
+      "step": 61
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.7699999999999997e-05,
+      "loss": 12.6469,
+      "step": 62
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.7999999999999997e-05,
+      "loss": 12.2555,
+      "step": 63
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8299999999999998e-05,
+      "loss": 13.5008,
+      "step": 64
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8599999999999998e-05,
+      "loss": 12.3855,
+      "step": 65
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.89e-05,
+      "loss": 11.6971,
+      "step": 66
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.92e-05,
+      "loss": 10.9397,
+      "step": 67
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95e-05,
+      "loss": 10.8864,
+      "step": 68
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98e-05,
+      "loss": 9.3281,
+      "step": 69
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.01e-05,
+      "loss": 11.7321,
+      "step": 70
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.04e-05,
+      "loss": 11.4566,
+      "step": 71
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.07e-05,
+      "loss": 10.7037,
+      "step": 72
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.1e-05,
+      "loss": 10.4583,
+      "step": 73
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.1299999999999996e-05,
+      "loss": 9.5085,
+      "step": 74
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.1599999999999996e-05,
+      "loss": 8.8717,
+      "step": 75
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.1899999999999997e-05,
+      "loss": 11.1006,
+      "step": 76
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.2199999999999998e-05,
+      "loss": 10.4534,
+      "step": 77
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.2499999999999998e-05,
+      "loss": 9.8379,
+      "step": 78
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.28e-05,
+      "loss": 10.0148,
+      "step": 79
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.31e-05,
+      "loss": 8.9424,
+      "step": 80
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.34e-05,
+      "loss": 8.8265,
+      "step": 81
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.3699999999999997e-05,
+      "loss": 9.7769,
+      "step": 82
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.3999999999999997e-05,
+      "loss": 9.9579,
+      "step": 83
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.4299999999999998e-05,
+      "loss": 9.5612,
+      "step": 84
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.4599999999999998e-05,
+      "loss": 8.6644,
+      "step": 85
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.49e-05,
+      "loss": 8.7807,
+      "step": 86
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.52e-05,
+      "loss": 8.6541,
+      "step": 87
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.55e-05,
+      "loss": 8.8828,
+      "step": 88
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.5799999999999997e-05,
+      "loss": 9.2106,
+      "step": 89
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.6099999999999997e-05,
+      "loss": 8.9537,
+      "step": 90
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.6399999999999995e-05,
+      "loss": 8.9203,
+      "step": 91
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.6699999999999995e-05,
+      "loss": 8.2055,
+      "step": 92
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.6999999999999996e-05,
+      "loss": 8.4133,
+      "step": 93
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.7299999999999996e-05,
+      "loss": 8.1085,
+      "step": 94
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.7599999999999997e-05,
+      "loss": 8.8725,
+      "step": 95
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.7899999999999997e-05,
+      "loss": 8.8758,
+      "step": 96
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 8.2101,
+      "step": 97
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 7.9802,
+      "step": 98
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.88e-05,
+      "loss": 7.73,
+      "step": 99
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.91e-05,
+      "loss": 7.3415,
+      "step": 100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.94e-05,
+      "loss": 9.1237,
+      "step": 101
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.97e-05,
+      "loss": 8.1518,
+      "step": 102
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 7.6631,
+      "step": 103
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0299999999999998e-05,
+      "loss": 7.7533,
+      "step": 104
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.06e-05,
+      "loss": 7.5032,
+      "step": 105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.09e-05,
+      "loss": 7.285,
+      "step": 106
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.119999999999999e-05,
+      "loss": 7.7336,
+      "step": 107
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.149999999999999e-05,
+      "loss": 8.0755,
+      "step": 108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.1799999999999994e-05,
+      "loss": 7.9322,
+      "step": 109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.2099999999999994e-05,
+      "loss": 7.5236,
+      "step": 110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.2399999999999995e-05,
+      "loss": 7.3505,
+      "step": 111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.2699999999999995e-05,
+      "loss": 7.0483,
+      "step": 112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.2999999999999996e-05,
+      "loss": 7.579,
+      "step": 113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.3299999999999996e-05,
+      "loss": 8.0822,
+      "step": 114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.36e-05,
+      "loss": 7.5678,
+      "step": 115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.39e-05,
+      "loss": 7.3537,
+      "step": 116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.42e-05,
+      "loss": 7.3048,
+      "step": 117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.45e-05,
+      "loss": 6.9404,
+      "step": 118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.48e-05,
+      "loss": 7.3091,
+      "step": 119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.51e-05,
+      "loss": 8.0984,
+      "step": 120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.539999999999999e-05,
+      "loss": 7.7869,
+      "step": 121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5699999999999994e-05,
+      "loss": 7.2823,
+      "step": 122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5999999999999994e-05,
+      "loss": 6.8487,
+      "step": 123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6299999999999995e-05,
+      "loss": 6.8981,
+      "step": 124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6599999999999995e-05,
+      "loss": 6.4681,
+      "step": 125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6899999999999996e-05,
+      "loss": 7.6376,
+      "step": 126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.7199999999999996e-05,
+      "loss": 7.1721,
+      "step": 127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.75e-05,
+      "loss": 6.6383,
+      "step": 128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.78e-05,
+      "loss": 6.8282,
+      "step": 129
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.81e-05,
+      "loss": 6.7407,
+      "step": 130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.84e-05,
+      "loss": 6.4721,
+      "step": 131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.87e-05,
+      "loss": 6.5467,
+      "step": 132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.9e-05,
+      "loss": 7.0105,
+      "step": 133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.93e-05,
+      "loss": 6.6986,
+      "step": 134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.96e-05,
+      "loss": 6.3454,
+      "step": 135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.99e-05,
+      "loss": 6.2623,
+      "step": 136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.02e-05,
+      "loss": 6.3575,
+      "step": 137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.05e-05,
+      "loss": 6.3999,
+      "step": 138
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.08e-05,
+      "loss": 6.9152,
+      "step": 139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.11e-05,
+      "loss": 6.5706,
+      "step": 140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.14e-05,
+      "loss": 6.5177,
+      "step": 141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.17e-05,
+      "loss": 6.339,
+      "step": 142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.2e-05,
+      "loss": 6.253,
+      "step": 143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.229999999999999e-05,
+      "loss": 5.7408,
+      "step": 144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.259999999999999e-05,
+      "loss": 6.9908,
+      "step": 145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.289999999999999e-05,
+      "loss": 6.2932,
+      "step": 146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.319999999999999e-05,
+      "loss": 6.1595,
+      "step": 147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3499999999999993e-05,
+      "loss": 6.1007,
+      "step": 148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3799999999999994e-05,
+      "loss": 5.8631,
+      "step": 149
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4099999999999995e-05,
+      "loss": 5.4869,
+      "step": 150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4399999999999995e-05,
+      "loss": 6.392,
+      "step": 151
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4699999999999996e-05,
+      "loss": 6.1076,
+      "step": 152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 5.9778,
+      "step": 153
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5299999999999997e-05,
+      "loss": 5.9088,
+      "step": 154
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.56e-05,
+      "loss": 5.9283,
+      "step": 155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.59e-05,
+      "loss": 5.5499,
+      "step": 156
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.62e-05,
+      "loss": 6.2441,
+      "step": 157
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.65e-05,
+      "loss": 6.1781,
+      "step": 158
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.68e-05,
+      "loss": 5.978,
+      "step": 159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.709999999999999e-05,
+      "loss": 5.816,
+      "step": 160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7399999999999993e-05,
+      "loss": 5.5698,
+      "step": 161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7699999999999994e-05,
+      "loss": 5.3348,
+      "step": 162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 5.4392,
+      "step": 163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8299999999999995e-05,
+      "loss": 5.8405,
+      "step": 164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8599999999999995e-05,
+      "loss": 5.6292,
+      "step": 165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8899999999999996e-05,
+      "loss": 5.4744,
+      "step": 166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9199999999999997e-05,
+      "loss": 5.3437,
+      "step": 167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.95e-05,
+      "loss": 5.1164,
+      "step": 168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.98e-05,
+      "loss": 5.1823,
+      "step": 169
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.01e-05,
+      "loss": 5.6929,
+      "step": 170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.04e-05,
+      "loss": 5.3673,
+      "step": 171
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.07e-05,
+      "loss": 5.2568,
+      "step": 172
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.1e-05,
+      "loss": 5.2012,
+      "step": 173
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.13e-05,
+      "loss": 5.0849,
+      "step": 174
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.1599999999999994e-05,
+      "loss": 4.9843,
+      "step": 175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.1899999999999994e-05,
+      "loss": 5.6313,
+      "step": 176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.2199999999999995e-05,
+      "loss": 5.3306,
+      "step": 177
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.2499999999999995e-05,
+      "loss": 5.1406,
+      "step": 178
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.279999999999999e-05,
+      "loss": 5.0528,
+      "step": 179
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.309999999999999e-05,
+      "loss": 4.8269,
+      "step": 180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.339999999999999e-05,
+      "loss": 5.0041,
+      "step": 181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.369999999999999e-05,
+      "loss": 5.0295,
+      "step": 182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.399999999999999e-05,
+      "loss": 5.0222,
+      "step": 183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.429999999999999e-05,
+      "loss": 5.0837,
+      "step": 184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.459999999999999e-05,
+      "loss": 4.8316,
+      "step": 185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.489999999999999e-05,
+      "loss": 4.861,
+      "step": 186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.519999999999999e-05,
+      "loss": 4.8633,
+      "step": 187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.5499999999999994e-05,
+      "loss": 4.5816,
+      "step": 188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.5799999999999994e-05,
+      "loss": 4.9932,
+      "step": 189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.6099999999999995e-05,
+      "loss": 4.763,
+      "step": 190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.6399999999999995e-05,
+      "loss": 4.7225,
+      "step": 191
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.6699999999999996e-05,
+      "loss": 4.6039,
+      "step": 192
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 4.5207,
+      "step": 193
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.73e-05,
+      "loss": 4.4418,
+      "step": 194
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.76e-05,
+      "loss": 4.8053,
+      "step": 195
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.79e-05,
+      "loss": 4.4795,
+      "step": 196
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.82e-05,
+      "loss": 4.5386,
+      "step": 197
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.85e-05,
+      "loss": 4.4503,
+      "step": 198
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.88e-05,
+      "loss": 4.5971,
+      "step": 199
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.91e-05,
+      "loss": 4.256,
+      "step": 200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 5.94e-05,
+      "loss": 4.5178,
+      "step": 201
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 5.97e-05,
+      "loss": 4.5053,
+      "step": 202
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 4.3546,
+      "step": 203
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.0299999999999995e-05,
+      "loss": 4.2786,
+      "step": 204
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.0599999999999996e-05,
+      "loss": 4.3948,
+      "step": 205
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.0899999999999996e-05,
+      "loss": 4.2465,
+      "step": 206
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.12e-05,
+      "loss": 4.2245,
+      "step": 207
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.149999999999999e-05,
+      "loss": 4.3215,
+      "step": 208
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.18e-05,
+      "loss": 4.2406,
+      "step": 209
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.209999999999999e-05,
+      "loss": 4.2692,
+      "step": 210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.239999999999999e-05,
+      "loss": 4.0468,
+      "step": 211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.269999999999999e-05,
+      "loss": 4.0571,
+      "step": 212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.299999999999999e-05,
+      "loss": 3.9491,
+      "step": 213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.33e-05,
+      "loss": 4.2397,
+      "step": 214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.359999999999999e-05,
+      "loss": 4.0732,
+      "step": 215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.39e-05,
+      "loss": 4.0468,
+      "step": 216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.419999999999999e-05,
+      "loss": 4.0571,
+      "step": 217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.45e-05,
+      "loss": 3.9212,
+      "step": 218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.479999999999999e-05,
+      "loss": 3.7819,
+      "step": 219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.51e-05,
+      "loss": 4.0382,
+      "step": 220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.539999999999999e-05,
+      "loss": 3.9625,
+      "step": 221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.57e-05,
+      "loss": 3.8712,
+      "step": 222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.599999999999999e-05,
+      "loss": 3.887,
+      "step": 223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.63e-05,
+      "loss": 3.8663,
+      "step": 224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.659999999999999e-05,
+      "loss": 3.6364,
+      "step": 225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.69e-05,
+      "loss": 3.908,
+      "step": 226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.72e-05,
+      "loss": 3.9241,
+      "step": 227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.75e-05,
+      "loss": 3.7936,
+      "step": 228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.78e-05,
+      "loss": 3.8212,
+      "step": 229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.81e-05,
+      "loss": 3.627,
+      "step": 230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.84e-05,
+      "loss": 3.6948,
+      "step": 231
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 6.87e-05,
+      "loss": 3.5747,
+      "step": 232
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 6.9e-05,
+      "loss": 3.7271,
+      "step": 233
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 6.93e-05,
+      "loss": 3.7019,
+      "step": 234
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 6.96e-05,
+      "loss": 3.7043,
+      "step": 235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 6.989999999999999e-05,
+      "loss": 3.6992,
+      "step": 236
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.02e-05,
+      "loss": 3.6387,
+      "step": 237
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.049999999999999e-05,
+      "loss": 3.4792,
+      "step": 238
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.079999999999999e-05,
+      "loss": 3.6625,
+      "step": 239
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.11e-05,
+      "loss": 3.6068,
+      "step": 240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.139999999999999e-05,
+      "loss": 3.5104,
+      "step": 241
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.17e-05,
+      "loss": 3.6004,
+      "step": 242
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.199999999999999e-05,
+      "loss": 3.6128,
+      "step": 243
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.23e-05,
+      "loss": 3.5804,
+      "step": 244
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.259999999999999e-05,
+      "loss": 3.693,
+      "step": 245
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.29e-05,
+      "loss": 3.5225,
+      "step": 246
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.319999999999999e-05,
+      "loss": 3.5164,
+      "step": 247
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.35e-05,
+      "loss": 3.5148,
+      "step": 248
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.379999999999999e-05,
+      "loss": 3.4295,
+      "step": 249
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.41e-05,
+      "loss": 3.0131,
+      "step": 250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.439999999999999e-05,
+      "loss": 3.6089,
+      "step": 251
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.47e-05,
+      "loss": 3.47,
+      "step": 252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.5e-05,
+      "loss": 3.4282,
+      "step": 253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.529999999999999e-05,
+      "loss": 3.487,
+      "step": 254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.56e-05,
+      "loss": 3.3932,
+      "step": 255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.589999999999999e-05,
+      "loss": 3.4193,
+      "step": 256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.62e-05,
+      "loss": 3.4082,
+      "step": 257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.649999999999999e-05,
+      "loss": 3.499,
+      "step": 258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.68e-05,
+      "loss": 3.391,
+      "step": 259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.709999999999999e-05,
+      "loss": 3.5363,
+      "step": 260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.74e-05,
+      "loss": 3.4239,
+      "step": 261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.769999999999999e-05,
+      "loss": 3.4229,
+      "step": 262
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.8e-05,
+      "loss": 3.3603,
+      "step": 263
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.829999999999999e-05,
+      "loss": 3.4676,
+      "step": 264
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.86e-05,
+      "loss": 3.3532,
+      "step": 265
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.89e-05,
+      "loss": 3.4107,
+      "step": 266
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.92e-05,
+      "loss": 3.3679,
+      "step": 267
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.95e-05,
+      "loss": 3.4845,
+      "step": 268
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.98e-05,
+      "loss": 3.1881,
+      "step": 269
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.01e-05,
+      "loss": 3.3872,
+      "step": 270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.04e-05,
+      "loss": 3.2703,
+      "step": 271
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.07e-05,
+      "loss": 3.3354,
+      "step": 272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.1e-05,
+      "loss": 3.2932,
+      "step": 273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.13e-05,
+      "loss": 3.4137,
+      "step": 274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.16e-05,
+      "loss": 3.1952,
+      "step": 275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.19e-05,
+      "loss": 3.3932,
+      "step": 276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.22e-05,
+      "loss": 3.3847,
+      "step": 277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.25e-05,
+      "loss": 3.3521,
+      "step": 278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.28e-05,
+      "loss": 3.3553,
+      "step": 279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.31e-05,
+      "loss": 3.3966,
+      "step": 280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.34e-05,
+      "loss": 3.2386,
+      "step": 281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.37e-05,
+      "loss": 3.1789,
+      "step": 282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.4e-05,
+      "loss": 3.3294,
+      "step": 283
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43e-05,
+      "loss": 3.2167,
+      "step": 284
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.459999999999998e-05,
+      "loss": 3.1864,
+      "step": 285
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489999999999999e-05,
+      "loss": 3.3375,
+      "step": 286
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.519999999999998e-05,
+      "loss": 3.2776,
+      "step": 287
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.549999999999999e-05,
+      "loss": 3.0096,
+      "step": 288
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.579999999999998e-05,
+      "loss": 3.2637,
+      "step": 289
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.609999999999999e-05,
+      "loss": 3.2664,
+      "step": 290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.639999999999999e-05,
+      "loss": 3.4265,
+      "step": 291
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.669999999999998e-05,
+      "loss": 3.4416,
+      "step": 292
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.699999999999999e-05,
+      "loss": 3.179,
+      "step": 293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.729999999999998e-05,
+      "loss": 3.1948,
+      "step": 294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.759999999999999e-05,
+      "loss": 3.2879,
+      "step": 295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.789999999999998e-05,
+      "loss": 3.3267,
+      "step": 296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.819999999999999e-05,
+      "loss": 3.266,
+      "step": 297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.849999999999998e-05,
+      "loss": 3.2779,
+      "step": 298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.879999999999999e-05,
+      "loss": 3.2002,
+      "step": 299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.909999999999998e-05,
+      "loss": 3.4092,
+      "step": 300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.939999999999999e-05,
+      "loss": 3.2419,
+      "step": 301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.969999999999998e-05,
+      "loss": 3.2481,
+      "step": 302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 3.2577,
+      "step": 303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.029999999999999e-05,
+      "loss": 3.3427,
+      "step": 304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.059999999999999e-05,
+      "loss": 3.2278,
+      "step": 305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.089999999999999e-05,
+      "loss": 3.1598,
+      "step": 306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.12e-05,
+      "loss": 3.1953,
+      "step": 307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.149999999999999e-05,
+      "loss": 3.2609,
+      "step": 308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.18e-05,
+      "loss": 3.2328,
+      "step": 309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.209999999999999e-05,
+      "loss": 3.3111,
+      "step": 310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.24e-05,
+      "loss": 3.3145,
+      "step": 311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.269999999999999e-05,
+      "loss": 3.2803,
+      "step": 312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.3e-05,
+      "loss": 2.8937,
+      "step": 313
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.329999999999999e-05,
+      "loss": 3.2609,
+      "step": 314
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.36e-05,
+      "loss": 3.3124,
+      "step": 315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.389999999999999e-05,
+      "loss": 3.2017,
+      "step": 316
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.419999999999999e-05,
+      "loss": 3.1251,
+      "step": 317
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.449999999999999e-05,
+      "loss": 3.4832,
+      "step": 318
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.479999999999999e-05,
+      "loss": 3.1529,
+      "step": 319
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.51e-05,
+      "loss": 3.1757,
+      "step": 320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.539999999999999e-05,
+      "loss": 3.222,
+      "step": 321
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.57e-05,
+      "loss": 3.1902,
+      "step": 322
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.599999999999999e-05,
+      "loss": 3.2565,
+      "step": 323
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.63e-05,
+      "loss": 3.3118,
+      "step": 324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.659999999999999e-05,
+      "loss": 3.2403,
+      "step": 325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.69e-05,
+      "loss": 3.2393,
+      "step": 326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.719999999999999e-05,
+      "loss": 3.2776,
+      "step": 327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.75e-05,
+      "loss": 3.253,
+      "step": 328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.779999999999999e-05,
+      "loss": 3.354,
+      "step": 329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.81e-05,
+      "loss": 3.2909,
+      "step": 330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.839999999999999e-05,
+      "loss": 3.229,
+      "step": 331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.87e-05,
+      "loss": 3.3418,
+      "step": 332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.9e-05,
+      "loss": 3.2035,
+      "step": 333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.93e-05,
+      "loss": 3.1851,
+      "step": 334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.96e-05,
+      "loss": 3.2177,
+      "step": 335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.99e-05,
+      "loss": 3.2328,
+      "step": 336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001002,
+      "loss": 3.3206,
+      "step": 337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001005,
+      "loss": 3.0517,
+      "step": 338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001008,
+      "loss": 3.1781,
+      "step": 339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001011,
+      "loss": 3.2372,
+      "step": 340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001014,
+      "loss": 3.2307,
+      "step": 341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00010169999999999999,
+      "loss": 3.1572,
+      "step": 342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000102,
+      "loss": 3.1702,
+      "step": 343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00010229999999999999,
+      "loss": 3.2945,
+      "step": 344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001026,
+      "loss": 3.2838,
+      "step": 345
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001029,
+      "loss": 3.2225,
+      "step": 346
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010319999999999999,
+      "loss": 3.318,
+      "step": 347
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010349999999999998,
+      "loss": 3.1662,
+      "step": 348
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010379999999999999,
+      "loss": 3.2137,
+      "step": 349
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010409999999999998,
+      "loss": 3.0761,
+      "step": 350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010439999999999999,
+      "loss": 3.2429,
+      "step": 351
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010469999999999998,
+      "loss": 3.2071,
+      "step": 352
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010499999999999999,
+      "loss": 3.3073,
+      "step": 353
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010529999999999998,
+      "loss": 3.0862,
+      "step": 354
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00010559999999999998,
+      "loss": 3.3085,
+      "step": 355
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010589999999999999,
+      "loss": 3.1823,
+      "step": 356
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010619999999999998,
+      "loss": 3.0491,
+      "step": 357
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010649999999999999,
+      "loss": 3.2795,
+      "step": 358
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010679999999999998,
+      "loss": 3.2322,
+      "step": 359
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010709999999999999,
+      "loss": 3.2165,
+      "step": 360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010739999999999998,
+      "loss": 3.1511,
+      "step": 361
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010769999999999999,
+      "loss": 3.2248,
+      "step": 362
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010799999999999998,
+      "loss": 3.1603,
+      "step": 363
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010829999999999999,
+      "loss": 3.131,
+      "step": 364
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00010859999999999998,
+      "loss": 3.3292,
+      "step": 365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00010889999999999999,
+      "loss": 3.2581,
+      "step": 366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00010919999999999998,
+      "loss": 3.4997,
+      "step": 367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00010949999999999999,
+      "loss": 3.2606,
+      "step": 368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00010979999999999999,
+      "loss": 3.2057,
+      "step": 369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00011009999999999999,
+      "loss": 3.1574,
+      "step": 370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00011039999999999999,
+      "loss": 3.2116,
+      "step": 371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001107,
+      "loss": 3.1117,
+      "step": 372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00011099999999999999,
+      "loss": 3.2235,
+      "step": 373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001113,
+      "loss": 3.2043,
+      "step": 374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00011159999999999999,
+      "loss": 3.1431,
+      "step": 375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001119,
+      "loss": 3.2359,
+      "step": 376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00011219999999999999,
+      "loss": 3.2853,
+      "step": 377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001125,
+      "loss": 3.3084,
+      "step": 378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00011279999999999999,
+      "loss": 3.1246,
+      "step": 379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00011309999999999998,
+      "loss": 3.3661,
+      "step": 380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00011339999999999999,
+      "loss": 3.2595,
+      "step": 381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00011369999999999999,
+      "loss": 3.0363,
+      "step": 382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 3.1946,
+      "step": 383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00011429999999999999,
+      "loss": 3.2214,
+      "step": 384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001146,
+      "loss": 3.3189,
+      "step": 385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00011489999999999999,
+      "loss": 3.2432,
+      "step": 386
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001152,
+      "loss": 3.2441,
+      "step": 387
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00011549999999999999,
+      "loss": 2.9075,
+      "step": 388
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001158,
+      "loss": 3.0961,
+      "step": 389
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00011609999999999999,
+      "loss": 3.1905,
+      "step": 390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001164,
+      "loss": 3.1204,
+      "step": 391
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00011669999999999999,
+      "loss": 3.082,
+      "step": 392
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000117,
+      "loss": 3.3085,
+      "step": 393
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00011729999999999999,
+      "loss": 3.1925,
+      "step": 394
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001176,
+      "loss": 3.197,
+      "step": 395
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00011789999999999999,
+      "loss": 3.2427,
+      "step": 396
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001182,
+      "loss": 3.2022,
+      "step": 397
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001185,
+      "loss": 3.1768,
+      "step": 398
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001188,
+      "loss": 3.3828,
+      "step": 399
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001191,
+      "loss": 3.0107,
+      "step": 400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001194,
+      "loss": 3.1996,
+      "step": 401
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001197,
+      "loss": 3.3547,
+      "step": 402
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 3.081,
+      "step": 403
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001203,
+      "loss": 3.1805,
+      "step": 404
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00012059999999999999,
+      "loss": 3.2415,
+      "step": 405
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001209,
+      "loss": 3.1783,
+      "step": 406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00012119999999999999,
+      "loss": 2.9772,
+      "step": 407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001215,
+      "loss": 3.2495,
+      "step": 408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00012179999999999999,
+      "loss": 3.2373,
+      "step": 409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00012209999999999999,
+      "loss": 3.2222,
+      "step": 410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001224,
+      "loss": 3.1272,
+      "step": 411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00012269999999999997,
+      "loss": 3.2722,
+      "step": 412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00012299999999999998,
+      "loss": 3.0961,
+      "step": 413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001233,
+      "loss": 3.2415,
+      "step": 414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001236,
+      "loss": 3.2185,
+      "step": 415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00012389999999999998,
+      "loss": 3.1714,
+      "step": 416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00012419999999999998,
+      "loss": 3.1469,
+      "step": 417
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001245,
+      "loss": 3.2171,
+      "step": 418
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00012479999999999997,
+      "loss": 3.0702,
+      "step": 419
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00012509999999999998,
+      "loss": 3.1932,
+      "step": 420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00012539999999999999,
+      "loss": 3.1859,
+      "step": 421
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001257,
+      "loss": 3.0934,
+      "step": 422
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00012599999999999997,
+      "loss": 3.2063,
+      "step": 423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00012629999999999998,
+      "loss": 3.1777,
+      "step": 424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001266,
+      "loss": 3.0123,
+      "step": 425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001269,
+      "loss": 3.2899,
+      "step": 426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00012719999999999997,
+      "loss": 3.2012,
+      "step": 427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012749999999999998,
+      "loss": 3.2724,
+      "step": 428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001278,
+      "loss": 3.3238,
+      "step": 429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001281,
+      "loss": 3.3304,
+      "step": 430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012839999999999998,
+      "loss": 3.203,
+      "step": 431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012869999999999998,
+      "loss": 3.0034,
+      "step": 432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000129,
+      "loss": 3.2289,
+      "step": 433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001293,
+      "loss": 3.3934,
+      "step": 434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012959999999999998,
+      "loss": 3.1664,
+      "step": 435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012989999999999999,
+      "loss": 3.1843,
+      "step": 436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001302,
+      "loss": 3.2146,
+      "step": 437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001305,
+      "loss": 3.1865,
+      "step": 438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013079999999999998,
+      "loss": 3.242,
+      "step": 439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001311,
+      "loss": 3.1515,
+      "step": 440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001314,
+      "loss": 3.1999,
+      "step": 441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013169999999999998,
+      "loss": 3.3009,
+      "step": 442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013199999999999998,
+      "loss": 3.2711,
+      "step": 443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001323,
+      "loss": 3.0543,
+      "step": 444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001326,
+      "loss": 3.1127,
+      "step": 445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013289999999999998,
+      "loss": 3.2251,
+      "step": 446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013319999999999999,
+      "loss": 3.1265,
+      "step": 447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001335,
+      "loss": 3.2516,
+      "step": 448
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001338,
+      "loss": 3.2357,
+      "step": 449
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013409999999999998,
+      "loss": 2.8892,
+      "step": 450
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001344,
+      "loss": 3.1593,
+      "step": 451
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001347,
+      "loss": 3.2581,
+      "step": 452
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000135,
+      "loss": 3.122,
+      "step": 453
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013529999999999998,
+      "loss": 3.2091,
+      "step": 454
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001356,
+      "loss": 3.219,
+      "step": 455
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001359,
+      "loss": 3.3029,
+      "step": 456
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001362,
+      "loss": 3.0923,
+      "step": 457
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013649999999999998,
+      "loss": 3.202,
+      "step": 458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001368,
+      "loss": 3.1816,
+      "step": 459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001371,
+      "loss": 3.1523,
+      "step": 460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001374,
+      "loss": 3.1652,
+      "step": 461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00013769999999999999,
+      "loss": 3.1823,
+      "step": 462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000138,
+      "loss": 3.0631,
+      "step": 463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001383,
+      "loss": 3.1574,
+      "step": 464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001386,
+      "loss": 3.2535,
+      "step": 465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001389,
+      "loss": 3.2028,
+      "step": 466
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001392,
+      "loss": 3.1971,
+      "step": 467
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001395,
+      "loss": 3.1708,
+      "step": 468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00013979999999999998,
+      "loss": 2.9569,
+      "step": 469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001401,
+      "loss": 3.279,
+      "step": 470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001404,
+      "loss": 3.1915,
+      "step": 471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00014069999999999998,
+      "loss": 3.2021,
+      "step": 472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 3.0978,
+      "step": 473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001413,
+      "loss": 3.2507,
+      "step": 474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00014159999999999997,
+      "loss": 3.0255,
+      "step": 475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00014189999999999998,
+      "loss": 3.1982,
+      "step": 476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001422,
+      "loss": 3.2753,
+      "step": 477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001425,
+      "loss": 3.2267,
+      "step": 478
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00014279999999999997,
+      "loss": 3.2532,
+      "step": 479
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00014309999999999998,
+      "loss": 3.0974,
+      "step": 480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001434,
+      "loss": 3.1696,
+      "step": 481
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00014369999999999997,
+      "loss": 3.0584,
+      "step": 482
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00014399999999999998,
+      "loss": 3.1414,
+      "step": 483
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00014429999999999998,
+      "loss": 3.1692,
+      "step": 484
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001446,
+      "loss": 3.2211,
+      "step": 485
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00014489999999999997,
+      "loss": 3.0825,
+      "step": 486
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00014519999999999998,
+      "loss": 3.224,
+      "step": 487
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00014549999999999999,
+      "loss": 3.031,
+      "step": 488
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001458,
+      "loss": 3.1818,
+      "step": 489
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00014609999999999997,
+      "loss": 3.1698,
+      "step": 490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00014639999999999998,
+      "loss": 3.1092,
+      "step": 491
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001467,
+      "loss": 3.1211,
+      "step": 492
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000147,
+      "loss": 3.1192,
+      "step": 493
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00014729999999999998,
+      "loss": 3.2278,
+      "step": 494
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00014759999999999998,
+      "loss": 3.1595,
+      "step": 495
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001479,
+      "loss": 3.1492,
+      "step": 496
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001482,
+      "loss": 3.0913,
+      "step": 497
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00014849999999999998,
+      "loss": 3.0448,
+      "step": 498
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00014879999999999998,
+      "loss": 3.1383,
+      "step": 499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001491,
+      "loss": 3.1411,
+      "step": 500
+    },
+    {
+      "epoch": 0.49,
+      "eval_cer": 1.0,
+      "eval_loss": 3.1673061847686768,
+      "eval_runtime": 225.9357,
+      "eval_samples_per_second": 39.011,
+      "eval_steps_per_second": 4.877,
+      "eval_wer": 1.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001494,
+      "loss": 3.1003,
+      "step": 501
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00014969999999999998,
+      "loss": 3.162,
+      "step": 502
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015,
+      "loss": 3.1332,
+      "step": 503
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001503,
+      "loss": 3.1213,
+      "step": 504
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015059999999999997,
+      "loss": 3.0938,
+      "step": 505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015089999999999998,
+      "loss": 3.1503,
+      "step": 506
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001512,
+      "loss": 3.0889,
+      "step": 507
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001515,
+      "loss": 3.1476,
+      "step": 508
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015179999999999998,
+      "loss": 3.2199,
+      "step": 509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015209999999999998,
+      "loss": 3.1255,
+      "step": 510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001524,
+      "loss": 3.2314,
+      "step": 511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001527,
+      "loss": 3.1023,
+      "step": 512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015299999999999998,
+      "loss": 2.9617,
+      "step": 513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015329999999999999,
+      "loss": 3.2176,
+      "step": 514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001536,
+      "loss": 3.0719,
+      "step": 515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001539,
+      "loss": 3.1304,
+      "step": 516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015419999999999998,
+      "loss": 3.1171,
+      "step": 517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001545,
+      "loss": 3.1312,
+      "step": 518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001548,
+      "loss": 3.1008,
+      "step": 519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001551,
+      "loss": 3.15,
+      "step": 520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015539999999999998,
+      "loss": 3.1321,
+      "step": 521
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001557,
+      "loss": 3.1089,
+      "step": 522
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000156,
+      "loss": 3.1479,
+      "step": 523
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001563,
+      "loss": 3.2236,
+      "step": 524
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015659999999999998,
+      "loss": 2.8275,
+      "step": 525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001569,
+      "loss": 3.0829,
+      "step": 526
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001572,
+      "loss": 3.105,
+      "step": 527
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015749999999999998,
+      "loss": 3.1749,
+      "step": 528
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001578,
+      "loss": 3.1341,
+      "step": 529
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001581,
+      "loss": 3.1704,
+      "step": 530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001584,
+      "loss": 3.1081,
+      "step": 531
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015869999999999998,
+      "loss": 2.9146,
+      "step": 532
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000159,
+      "loss": 3.1034,
+      "step": 533
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001593,
+      "loss": 3.2314,
+      "step": 534
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001596,
+      "loss": 3.117,
+      "step": 535
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015989999999999998,
+      "loss": 3.1529,
+      "step": 536
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001602,
+      "loss": 3.1863,
+      "step": 537
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001605,
+      "loss": 3.0017,
+      "step": 538
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001608,
+      "loss": 3.1325,
+      "step": 539
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016109999999999999,
+      "loss": 3.1952,
+      "step": 540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001614,
+      "loss": 3.203,
+      "step": 541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001617,
+      "loss": 3.1069,
+      "step": 542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000162,
+      "loss": 3.1265,
+      "step": 543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001623,
+      "loss": 3.0699,
+      "step": 544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001626,
+      "loss": 3.0817,
+      "step": 545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001629,
+      "loss": 3.0686,
+      "step": 546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001632,
+      "loss": 3.0809,
+      "step": 547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001635,
+      "loss": 3.1436,
+      "step": 548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001638,
+      "loss": 3.1343,
+      "step": 549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001641,
+      "loss": 2.9484,
+      "step": 550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001644,
+      "loss": 3.0933,
+      "step": 551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001647,
+      "loss": 3.0559,
+      "step": 552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000165,
+      "loss": 3.1352,
+      "step": 553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001653,
+      "loss": 3.1666,
+      "step": 554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001656,
+      "loss": 3.1104,
+      "step": 555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001659,
+      "loss": 3.1451,
+      "step": 556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001662,
+      "loss": 3.0521,
+      "step": 557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001665,
+      "loss": 3.1007,
+      "step": 558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001668,
+      "loss": 3.1638,
+      "step": 559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001671,
+      "loss": 3.1375,
+      "step": 560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001674,
+      "loss": 3.1051,
+      "step": 561
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001677,
+      "loss": 3.0681,
+      "step": 562
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000168,
+      "loss": 2.9591,
+      "step": 563
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001683,
+      "loss": 3.059,
+      "step": 564
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001686,
+      "loss": 3.0796,
+      "step": 565
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016889999999999996,
+      "loss": 3.1021,
+      "step": 566
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016919999999999997,
+      "loss": 3.051,
+      "step": 567
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016949999999999997,
+      "loss": 3.1241,
+      "step": 568
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016979999999999998,
+      "loss": 2.9081,
+      "step": 569
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017009999999999996,
+      "loss": 3.069,
+      "step": 570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017039999999999997,
+      "loss": 3.0923,
+      "step": 571
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017069999999999998,
+      "loss": 3.0633,
+      "step": 572
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017099999999999998,
+      "loss": 3.1309,
+      "step": 573
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017129999999999996,
+      "loss": 3.126,
+      "step": 574
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017159999999999997,
+      "loss": 2.9072,
+      "step": 575
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017189999999999998,
+      "loss": 3.0534,
+      "step": 576
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017219999999999998,
+      "loss": 3.0721,
+      "step": 577
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017249999999999996,
+      "loss": 3.1219,
+      "step": 578
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017279999999999997,
+      "loss": 3.0596,
+      "step": 579
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017309999999999998,
+      "loss": 3.0897,
+      "step": 580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00017339999999999996,
+      "loss": 3.0468,
+      "step": 581
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017369999999999997,
+      "loss": 3.0265,
+      "step": 582
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017399999999999997,
+      "loss": 3.1277,
+      "step": 583
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017429999999999998,
+      "loss": 3.0789,
+      "step": 584
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017459999999999996,
+      "loss": 3.0828,
+      "step": 585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017489999999999997,
+      "loss": 3.0793,
+      "step": 586
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017519999999999998,
+      "loss": 3.0759,
+      "step": 587
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017549999999999998,
+      "loss": 2.9961,
+      "step": 588
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017579999999999996,
+      "loss": 3.1083,
+      "step": 589
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017609999999999997,
+      "loss": 3.0659,
+      "step": 590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017639999999999998,
+      "loss": 3.0407,
+      "step": 591
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00017669999999999999,
+      "loss": 3.0827,
+      "step": 592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00017699999999999997,
+      "loss": 3.098,
+      "step": 593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00017729999999999997,
+      "loss": 2.7888,
+      "step": 594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00017759999999999998,
+      "loss": 3.0416,
+      "step": 595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001779,
+      "loss": 3.1261,
+      "step": 596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00017819999999999997,
+      "loss": 3.0448,
+      "step": 597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00017849999999999997,
+      "loss": 3.0713,
+      "step": 598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00017879999999999998,
+      "loss": 3.0403,
+      "step": 599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001791,
+      "loss": 2.8209,
+      "step": 600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00017939999999999997,
+      "loss": 3.1067,
+      "step": 601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00017969999999999998,
+      "loss": 3.0684,
+      "step": 602
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 3.108,
+      "step": 603
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018029999999999996,
+      "loss": 3.0845,
+      "step": 604
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018059999999999997,
+      "loss": 3.1348,
+      "step": 605
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018089999999999998,
+      "loss": 3.0897,
+      "step": 606
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018119999999999999,
+      "loss": 2.9896,
+      "step": 607
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018149999999999997,
+      "loss": 3.0995,
+      "step": 608
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018179999999999997,
+      "loss": 3.1463,
+      "step": 609
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018209999999999998,
+      "loss": 3.0713,
+      "step": 610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001824,
+      "loss": 3.05,
+      "step": 611
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018269999999999997,
+      "loss": 3.0536,
+      "step": 612
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018299999999999998,
+      "loss": 2.9615,
+      "step": 613
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018329999999999998,
+      "loss": 3.0634,
+      "step": 614
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001836,
+      "loss": 3.0761,
+      "step": 615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018389999999999997,
+      "loss": 3.1137,
+      "step": 616
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018419999999999998,
+      "loss": 3.0851,
+      "step": 617
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018449999999999999,
+      "loss": 3.1056,
+      "step": 618
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001848,
+      "loss": 2.8998,
+      "step": 619
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018509999999999997,
+      "loss": 3.0727,
+      "step": 620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018539999999999998,
+      "loss": 3.0627,
+      "step": 621
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001857,
+      "loss": 3.097,
+      "step": 622
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000186,
+      "loss": 3.0577,
+      "step": 623
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018629999999999997,
+      "loss": 3.0497,
+      "step": 624
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018659999999999998,
+      "loss": 2.9327,
+      "step": 625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001869,
+      "loss": 3.0539,
+      "step": 626
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001872,
+      "loss": 3.0407,
+      "step": 627
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018749999999999998,
+      "loss": 3.0375,
+      "step": 628
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018779999999999998,
+      "loss": 3.0178,
+      "step": 629
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001881,
+      "loss": 3.0922,
+      "step": 630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018839999999999997,
+      "loss": 3.0655,
+      "step": 631
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018869999999999998,
+      "loss": 2.9092,
+      "step": 632
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018899999999999999,
+      "loss": 3.0617,
+      "step": 633
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001893,
+      "loss": 3.0889,
+      "step": 634
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018959999999999997,
+      "loss": 3.094,
+      "step": 635
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018989999999999998,
+      "loss": 3.0646,
+      "step": 636
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001902,
+      "loss": 3.0484,
+      "step": 637
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001905,
+      "loss": 2.9848,
+      "step": 638
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00019079999999999998,
+      "loss": 3.045,
+      "step": 639
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00019109999999999998,
+      "loss": 3.032,
+      "step": 640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001914,
+      "loss": 2.9927,
+      "step": 641
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001917,
+      "loss": 3.0554,
+      "step": 642
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00019199999999999998,
+      "loss": 3.0307,
+      "step": 643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00019229999999999999,
+      "loss": 2.9086,
+      "step": 644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001926,
+      "loss": 3.0333,
+      "step": 645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001929,
+      "loss": 3.0145,
+      "step": 646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00019319999999999998,
+      "loss": 2.9969,
+      "step": 647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001935,
+      "loss": 2.9967,
+      "step": 648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001938,
+      "loss": 2.9584,
+      "step": 649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001941,
+      "loss": 2.901,
+      "step": 650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00019439999999999998,
+      "loss": 2.9662,
+      "step": 651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001947,
+      "loss": 2.9356,
+      "step": 652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000195,
+      "loss": 2.9744,
+      "step": 653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00019529999999999998,
+      "loss": 2.902,
+      "step": 654
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00019559999999999998,
+      "loss": 2.918,
+      "step": 655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001959,
+      "loss": 2.8798,
+      "step": 656
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001962,
+      "loss": 2.9673,
+      "step": 657
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00019649999999999998,
+      "loss": 2.8923,
+      "step": 658
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00019679999999999999,
+      "loss": 2.8036,
+      "step": 659
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001971,
+      "loss": 2.7968,
+      "step": 660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001974,
+      "loss": 2.759,
+      "step": 661
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00019769999999999998,
+      "loss": 2.7425,
+      "step": 662
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.000198,
+      "loss": 2.583,
+      "step": 663
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001983,
+      "loss": 2.7617,
+      "step": 664
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001986,
+      "loss": 2.7101,
+      "step": 665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00019889999999999998,
+      "loss": 2.634,
+      "step": 666
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001992,
+      "loss": 2.6503,
+      "step": 667
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001995,
+      "loss": 2.575,
+      "step": 668
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001998,
+      "loss": 2.3321,
+      "step": 669
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020009999999999998,
+      "loss": 2.4613,
+      "step": 670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002004,
+      "loss": 2.4095,
+      "step": 671
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002007,
+      "loss": 2.5524,
+      "step": 672
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000201,
+      "loss": 2.282,
+      "step": 673
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002013,
+      "loss": 2.3357,
+      "step": 674
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002016,
+      "loss": 2.4709,
+      "step": 675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002019,
+      "loss": 2.4067,
+      "step": 676
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002022,
+      "loss": 2.286,
+      "step": 677
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002025,
+      "loss": 2.2289,
+      "step": 678
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002028,
+      "loss": 2.2339,
+      "step": 679
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002031,
+      "loss": 2.0423,
+      "step": 680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020339999999999998,
+      "loss": 2.1442,
+      "step": 681
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002037,
+      "loss": 2.235,
+      "step": 682
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000204,
+      "loss": 2.1599,
+      "step": 683
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002043,
+      "loss": 1.9735,
+      "step": 684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020459999999999999,
+      "loss": 2.0162,
+      "step": 685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002049,
+      "loss": 1.8074,
+      "step": 686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002052,
+      "loss": 1.8826,
+      "step": 687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002055,
+      "loss": 1.8905,
+      "step": 688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002058,
+      "loss": 1.9072,
+      "step": 689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002061,
+      "loss": 1.8261,
+      "step": 690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020639999999999998,
+      "loss": 1.9367,
+      "step": 691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020669999999999996,
+      "loss": 1.835,
+      "step": 692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020699999999999996,
+      "loss": 1.9254,
+      "step": 693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020729999999999997,
+      "loss": 1.7017,
+      "step": 694
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020759999999999998,
+      "loss": 1.965,
+      "step": 695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00020789999999999996,
+      "loss": 1.713,
+      "step": 696
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00020819999999999996,
+      "loss": 1.5504,
+      "step": 697
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00020849999999999997,
+      "loss": 1.7171,
+      "step": 698
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00020879999999999998,
+      "loss": 1.5673,
+      "step": 699
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00020909999999999996,
+      "loss": 1.9135,
+      "step": 700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00020939999999999997,
+      "loss": 1.7929,
+      "step": 701
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00020969999999999997,
+      "loss": 1.5535,
+      "step": 702
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00020999999999999998,
+      "loss": 1.6549,
+      "step": 703
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00021029999999999996,
+      "loss": 1.6161,
+      "step": 704
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00021059999999999997,
+      "loss": 1.6063,
+      "step": 705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021089999999999998,
+      "loss": 1.5072,
+      "step": 706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021119999999999996,
+      "loss": 1.504,
+      "step": 707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021149999999999996,
+      "loss": 1.3957,
+      "step": 708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021179999999999997,
+      "loss": 1.5295,
+      "step": 709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021209999999999998,
+      "loss": 1.5585,
+      "step": 710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021239999999999996,
+      "loss": 1.5335,
+      "step": 711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021269999999999997,
+      "loss": 1.4745,
+      "step": 712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021299999999999997,
+      "loss": 1.2947,
+      "step": 713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021329999999999998,
+      "loss": 1.461,
+      "step": 714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00021359999999999996,
+      "loss": 1.3703,
+      "step": 715
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021389999999999997,
+      "loss": 1.2951,
+      "step": 716
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021419999999999998,
+      "loss": 1.2391,
+      "step": 717
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021449999999999998,
+      "loss": 1.4724,
+      "step": 718
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021479999999999996,
+      "loss": 1.4623,
+      "step": 719
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021509999999999997,
+      "loss": 1.3787,
+      "step": 720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021539999999999998,
+      "loss": 1.3575,
+      "step": 721
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021569999999999998,
+      "loss": 1.3798,
+      "step": 722
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021599999999999996,
+      "loss": 1.313,
+      "step": 723
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021629999999999997,
+      "loss": 1.3684,
+      "step": 724
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021659999999999998,
+      "loss": 1.2902,
+      "step": 725
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002169,
+      "loss": 1.3924,
+      "step": 726
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021719999999999997,
+      "loss": 1.4578,
+      "step": 727
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021749999999999997,
+      "loss": 1.1712,
+      "step": 728
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021779999999999998,
+      "loss": 1.151,
+      "step": 729
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021809999999999996,
+      "loss": 1.1398,
+      "step": 730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021839999999999997,
+      "loss": 1.294,
+      "step": 731
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021869999999999998,
+      "loss": 1.144,
+      "step": 732
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021899999999999998,
+      "loss": 1.1332,
+      "step": 733
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021929999999999996,
+      "loss": 1.122,
+      "step": 734
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021959999999999997,
+      "loss": 1.2741,
+      "step": 735
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021989999999999998,
+      "loss": 1.2975,
+      "step": 736
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00022019999999999999,
+      "loss": 1.2774,
+      "step": 737
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00022049999999999997,
+      "loss": 1.3293,
+      "step": 738
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00022079999999999997,
+      "loss": 1.2767,
+      "step": 739
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00022109999999999998,
+      "loss": 1.1475,
+      "step": 740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002214,
+      "loss": 1.1118,
+      "step": 741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00022169999999999997,
+      "loss": 1.1057,
+      "step": 742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00022199999999999998,
+      "loss": 1.1188,
+      "step": 743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00022229999999999998,
+      "loss": 1.085,
+      "step": 744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002226,
+      "loss": 1.2239,
+      "step": 745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00022289999999999997,
+      "loss": 1.0318,
+      "step": 746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00022319999999999998,
+      "loss": 1.1194,
+      "step": 747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00022349999999999998,
+      "loss": 1.11,
+      "step": 748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002238,
+      "loss": 1.1376,
+      "step": 749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00022409999999999997,
+      "loss": 1.164,
+      "step": 750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00022439999999999998,
+      "loss": 1.6213,
+      "step": 751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002247,
+      "loss": 1.0801,
+      "step": 752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.000225,
+      "loss": 1.1924,
+      "step": 753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00022529999999999997,
+      "loss": 1.0547,
+      "step": 754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00022559999999999998,
+      "loss": 1.0753,
+      "step": 755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002259,
+      "loss": 1.0703,
+      "step": 756
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00022619999999999997,
+      "loss": 1.0935,
+      "step": 757
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00022649999999999998,
+      "loss": 1.0962,
+      "step": 758
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00022679999999999998,
+      "loss": 1.1274,
+      "step": 759
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002271,
+      "loss": 1.1263,
+      "step": 760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00022739999999999997,
+      "loss": 0.9613,
+      "step": 761
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00022769999999999998,
+      "loss": 0.9693,
+      "step": 762
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 0.9716,
+      "step": 763
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002283,
+      "loss": 0.9967,
+      "step": 764
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00022859999999999997,
+      "loss": 1.0343,
+      "step": 765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00022889999999999998,
+      "loss": 1.2189,
+      "step": 766
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002292,
+      "loss": 1.0123,
+      "step": 767
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002295,
+      "loss": 0.8324,
+      "step": 768
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00022979999999999997,
+      "loss": 1.0937,
+      "step": 769
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00023009999999999998,
+      "loss": 1.1963,
+      "step": 770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002304,
+      "loss": 0.9191,
+      "step": 771
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002307,
+      "loss": 0.916,
+      "step": 772
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00023099999999999998,
+      "loss": 1.1599,
+      "step": 773
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00023129999999999998,
+      "loss": 0.8713,
+      "step": 774
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002316,
+      "loss": 1.0701,
+      "step": 775
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002319,
+      "loss": 0.9891,
+      "step": 776
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00023219999999999998,
+      "loss": 0.8734,
+      "step": 777
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00023249999999999999,
+      "loss": 0.9364,
+      "step": 778
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002328,
+      "loss": 1.0365,
+      "step": 779
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00023309999999999997,
+      "loss": 0.8466,
+      "step": 780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00023339999999999998,
+      "loss": 1.031,
+      "step": 781
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002337,
+      "loss": 1.1263,
+      "step": 782
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000234,
+      "loss": 0.8494,
+      "step": 783
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00023429999999999998,
+      "loss": 0.9926,
+      "step": 784
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00023459999999999998,
+      "loss": 0.8668,
+      "step": 785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002349,
+      "loss": 0.8733,
+      "step": 786
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002352,
+      "loss": 1.0404,
+      "step": 787
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00023549999999999998,
+      "loss": 1.0492,
+      "step": 788
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00023579999999999999,
+      "loss": 1.1289,
+      "step": 789
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002361,
+      "loss": 1.1559,
+      "step": 790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002364,
+      "loss": 0.8558,
+      "step": 791
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00023669999999999998,
+      "loss": 1.0651,
+      "step": 792
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000237,
+      "loss": 0.9266,
+      "step": 793
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002373,
+      "loss": 1.025,
+      "step": 794
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002376,
+      "loss": 1.1001,
+      "step": 795
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00023789999999999998,
+      "loss": 0.9491,
+      "step": 796
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002382,
+      "loss": 1.024,
+      "step": 797
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002385,
+      "loss": 0.936,
+      "step": 798
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002388,
+      "loss": 0.9944,
+      "step": 799
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00023909999999999998,
+      "loss": 1.0683,
+      "step": 800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002394,
+      "loss": 0.9788,
+      "step": 801
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002397,
+      "loss": 0.8131,
+      "step": 802
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 0.875,
+      "step": 803
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00024029999999999999,
+      "loss": 1.1203,
+      "step": 804
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002406,
+      "loss": 0.9198,
+      "step": 805
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002409,
+      "loss": 1.1147,
+      "step": 806
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00024119999999999998,
+      "loss": 1.0,
+      "step": 807
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002415,
+      "loss": 1.0405,
+      "step": 808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002418,
+      "loss": 0.7781,
+      "step": 809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002421,
+      "loss": 1.0149,
+      "step": 810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00024239999999999998,
+      "loss": 1.0773,
+      "step": 811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002427,
+      "loss": 0.8661,
+      "step": 812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.000243,
+      "loss": 0.9988,
+      "step": 813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002433,
+      "loss": 0.9182,
+      "step": 814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00024359999999999999,
+      "loss": 0.931,
+      "step": 815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00024389999999999997,
+      "loss": 0.9233,
+      "step": 816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00024419999999999997,
+      "loss": 0.9828,
+      "step": 817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002445,
+      "loss": 1.0876,
+      "step": 818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002448,
+      "loss": 0.9324,
+      "step": 819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00024509999999999994,
+      "loss": 0.806,
+      "step": 820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00024539999999999995,
+      "loss": 1.0356,
+      "step": 821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00024569999999999995,
+      "loss": 0.8061,
+      "step": 822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00024599999999999996,
+      "loss": 0.9484,
+      "step": 823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00024629999999999997,
+      "loss": 0.8986,
+      "step": 824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002466,
+      "loss": 0.9531,
+      "step": 825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002469,
+      "loss": 0.9663,
+      "step": 826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002472,
+      "loss": 0.9998,
+      "step": 827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00024749999999999994,
+      "loss": 0.8664,
+      "step": 828
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00024779999999999995,
+      "loss": 0.7926,
+      "step": 829
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00024809999999999996,
+      "loss": 0.8046,
+      "step": 830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00024839999999999997,
+      "loss": 0.9695,
+      "step": 831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0002487,
+      "loss": 0.9447,
+      "step": 832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.000249,
+      "loss": 0.8173,
+      "step": 833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0002493,
+      "loss": 0.9194,
+      "step": 834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00024959999999999994,
+      "loss": 0.8333,
+      "step": 835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00024989999999999995,
+      "loss": 0.9356,
+      "step": 836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00025019999999999996,
+      "loss": 0.8172,
+      "step": 837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00025049999999999996,
+      "loss": 0.8352,
+      "step": 838
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00025079999999999997,
+      "loss": 0.8992,
+      "step": 839
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002511,
+      "loss": 0.9217,
+      "step": 840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002514,
+      "loss": 0.8731,
+      "step": 841
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002517,
+      "loss": 0.7937,
+      "step": 842
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00025199999999999995,
+      "loss": 0.76,
+      "step": 843
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00025229999999999995,
+      "loss": 0.8595,
+      "step": 844
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00025259999999999996,
+      "loss": 0.7149,
+      "step": 845
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00025289999999999997,
+      "loss": 0.9245,
+      "step": 846
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002532,
+      "loss": 0.749,
+      "step": 847
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002535,
+      "loss": 0.7589,
+      "step": 848
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002538,
+      "loss": 0.9026,
+      "step": 849
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002541,
+      "loss": 1.1156,
+      "step": 850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00025439999999999995,
+      "loss": 0.8204,
+      "step": 851
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00025469999999999996,
+      "loss": 0.7411,
+      "step": 852
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 0.7564,
+      "step": 853
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00025529999999999997,
+      "loss": 0.8612,
+      "step": 854
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002556,
+      "loss": 0.7596,
+      "step": 855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002559,
+      "loss": 0.832,
+      "step": 856
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002562,
+      "loss": 0.95,
+      "step": 857
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00025649999999999995,
+      "loss": 0.8068,
+      "step": 858
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00025679999999999995,
+      "loss": 0.7972,
+      "step": 859
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00025709999999999996,
+      "loss": 0.8306,
+      "step": 860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00025739999999999997,
+      "loss": 0.8521,
+      "step": 861
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0002577,
+      "loss": 0.9133,
+      "step": 862
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.000258,
+      "loss": 0.8322,
+      "step": 863
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0002583,
+      "loss": 0.8976,
+      "step": 864
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0002586,
+      "loss": 0.709,
+      "step": 865
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00025889999999999995,
+      "loss": 0.8069,
+      "step": 866
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00025919999999999996,
+      "loss": 0.9265,
+      "step": 867
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00025949999999999997,
+      "loss": 0.833,
+      "step": 868
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00025979999999999997,
+      "loss": 0.858,
+      "step": 869
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0002601,
+      "loss": 0.7234,
+      "step": 870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002604,
+      "loss": 0.7724,
+      "step": 871
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002607,
+      "loss": 0.6745,
+      "step": 872
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.000261,
+      "loss": 0.906,
+      "step": 873
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00026129999999999995,
+      "loss": 0.6521,
+      "step": 874
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00026159999999999996,
+      "loss": 0.7834,
+      "step": 875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00026189999999999997,
+      "loss": 0.9081,
+      "step": 876
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002622,
+      "loss": 0.8004,
+      "step": 877
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002625,
+      "loss": 0.7573,
+      "step": 878
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002628,
+      "loss": 0.6556,
+      "step": 879
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002631,
+      "loss": 0.738,
+      "step": 880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00026339999999999995,
+      "loss": 0.9805,
+      "step": 881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00026369999999999996,
+      "loss": 0.8738,
+      "step": 882
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00026399999999999997,
+      "loss": 0.6671,
+      "step": 883
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002643,
+      "loss": 0.5979,
+      "step": 884
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002646,
+      "loss": 0.748,
+      "step": 885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002649,
+      "loss": 0.7585,
+      "step": 886
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002652,
+      "loss": 0.6884,
+      "step": 887
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002655,
+      "loss": 0.897,
+      "step": 888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00026579999999999996,
+      "loss": 0.848,
+      "step": 889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00026609999999999996,
+      "loss": 0.8246,
+      "step": 890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00026639999999999997,
+      "loss": 0.7677,
+      "step": 891
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002667,
+      "loss": 0.6112,
+      "step": 892
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000267,
+      "loss": 0.9802,
+      "step": 893
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002673,
+      "loss": 0.8253,
+      "step": 894
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002676,
+      "loss": 0.8606,
+      "step": 895
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002679,
+      "loss": 0.6891,
+      "step": 896
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00026819999999999996,
+      "loss": 0.7888,
+      "step": 897
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00026849999999999997,
+      "loss": 0.6734,
+      "step": 898
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002688,
+      "loss": 0.7022,
+      "step": 899
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002691,
+      "loss": 0.8038,
+      "step": 900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002694,
+      "loss": 0.7869,
+      "step": 901
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002697,
+      "loss": 0.6056,
+      "step": 902
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00027,
+      "loss": 0.687,
+      "step": 903
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00027029999999999996,
+      "loss": 0.6524,
+      "step": 904
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00027059999999999996,
+      "loss": 0.6669,
+      "step": 905
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00027089999999999997,
+      "loss": 0.8012,
+      "step": 906
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002712,
+      "loss": 0.6878,
+      "step": 907
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002715,
+      "loss": 0.6391,
+      "step": 908
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002718,
+      "loss": 0.7272,
+      "step": 909
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002721,
+      "loss": 0.7261,
+      "step": 910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002724,
+      "loss": 0.6685,
+      "step": 911
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00027269999999999996,
+      "loss": 0.6956,
+      "step": 912
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00027299999999999997,
+      "loss": 0.7676,
+      "step": 913
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002733,
+      "loss": 0.7311,
+      "step": 914
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002736,
+      "loss": 0.6671,
+      "step": 915
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002739,
+      "loss": 0.5879,
+      "step": 916
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002742,
+      "loss": 0.6921,
+      "step": 917
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002745,
+      "loss": 0.6178,
+      "step": 918
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002748,
+      "loss": 0.6722,
+      "step": 919
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00027509999999999996,
+      "loss": 0.6394,
+      "step": 920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00027539999999999997,
+      "loss": 0.7072,
+      "step": 921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002757,
+      "loss": 0.6387,
+      "step": 922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000276,
+      "loss": 0.5964,
+      "step": 923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002763,
+      "loss": 0.598,
+      "step": 924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002766,
+      "loss": 0.8506,
+      "step": 925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002769,
+      "loss": 0.6464,
+      "step": 926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002772,
+      "loss": 0.737,
+      "step": 927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00027749999999999997,
+      "loss": 0.7997,
+      "step": 928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002778,
+      "loss": 0.6367,
+      "step": 929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002781,
+      "loss": 0.6816,
+      "step": 930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002784,
+      "loss": 0.7854,
+      "step": 931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002787,
+      "loss": 0.6057,
+      "step": 932
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.000279,
+      "loss": 0.6954,
+      "step": 933
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002793,
+      "loss": 0.5555,
+      "step": 934
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00027959999999999997,
+      "loss": 0.6514,
+      "step": 935
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002799,
+      "loss": 0.541,
+      "step": 936
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002802,
+      "loss": 0.6165,
+      "step": 937
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002805,
+      "loss": 0.8988,
+      "step": 938
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002808,
+      "loss": 0.5916,
+      "step": 939
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002811,
+      "loss": 0.6339,
+      "step": 940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00028139999999999996,
+      "loss": 0.735,
+      "step": 941
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00028169999999999996,
+      "loss": 0.6257,
+      "step": 942
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 0.6529,
+      "step": 943
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002823,
+      "loss": 0.5977,
+      "step": 944
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002826,
+      "loss": 0.725,
+      "step": 945
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00028289999999999994,
+      "loss": 0.5425,
+      "step": 946
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00028319999999999994,
+      "loss": 0.6691,
+      "step": 947
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00028349999999999995,
+      "loss": 0.7081,
+      "step": 948
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00028379999999999996,
+      "loss": 0.5961,
+      "step": 949
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00028409999999999997,
+      "loss": 0.617,
+      "step": 950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002844,
+      "loss": 0.5945,
+      "step": 951
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002847,
+      "loss": 0.5942,
+      "step": 952
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000285,
+      "loss": 0.5564,
+      "step": 953
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00028529999999999994,
+      "loss": 0.6063,
+      "step": 954
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00028559999999999995,
+      "loss": 0.6293,
+      "step": 955
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00028589999999999996,
+      "loss": 0.6933,
+      "step": 956
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00028619999999999996,
+      "loss": 0.7187,
+      "step": 957
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00028649999999999997,
+      "loss": 0.6056,
+      "step": 958
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0002868,
+      "loss": 0.6234,
+      "step": 959
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0002871,
+      "loss": 0.6112,
+      "step": 960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00028739999999999994,
+      "loss": 0.5691,
+      "step": 961
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00028769999999999995,
+      "loss": 0.6282,
+      "step": 962
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00028799999999999995,
+      "loss": 0.7851,
+      "step": 963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00028829999999999996,
+      "loss": 0.6414,
+      "step": 964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00028859999999999997,
+      "loss": 0.6063,
+      "step": 965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0002889,
+      "loss": 0.6099,
+      "step": 966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0002892,
+      "loss": 0.7376,
+      "step": 967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0002895,
+      "loss": 0.7043,
+      "step": 968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00028979999999999994,
+      "loss": 0.6468,
+      "step": 969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00029009999999999995,
+      "loss": 0.6013,
+      "step": 970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00029039999999999996,
+      "loss": 0.5917,
+      "step": 971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00029069999999999996,
+      "loss": 0.6539,
+      "step": 972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 0.5583,
+      "step": 973
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002913,
+      "loss": 0.5624,
+      "step": 974
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002916,
+      "loss": 0.564,
+      "step": 975
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002919,
+      "loss": 0.558,
+      "step": 976
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00029219999999999995,
+      "loss": 0.6725,
+      "step": 977
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00029249999999999995,
+      "loss": 0.6443,
+      "step": 978
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00029279999999999996,
+      "loss": 0.6424,
+      "step": 979
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00029309999999999997,
+      "loss": 0.6341,
+      "step": 980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002934,
+      "loss": 0.5475,
+      "step": 981
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002937,
+      "loss": 0.7556,
+      "step": 982
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.000294,
+      "loss": 0.5838,
+      "step": 983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029429999999999994,
+      "loss": 0.5473,
+      "step": 984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029459999999999995,
+      "loss": 0.6598,
+      "step": 985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029489999999999996,
+      "loss": 0.4789,
+      "step": 986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029519999999999997,
+      "loss": 0.679,
+      "step": 987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029549999999999997,
+      "loss": 0.6608,
+      "step": 988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0002958,
+      "loss": 0.581,
+      "step": 989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0002961,
+      "loss": 0.4937,
+      "step": 990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0002964,
+      "loss": 0.6421,
+      "step": 991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029669999999999995,
+      "loss": 0.502,
+      "step": 992
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029699999999999996,
+      "loss": 0.5557,
+      "step": 993
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00029729999999999996,
+      "loss": 0.5581,
+      "step": 994
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00029759999999999997,
+      "loss": 0.4898,
+      "step": 995
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002979,
+      "loss": 0.5372,
+      "step": 996
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002982,
+      "loss": 0.5339,
+      "step": 997
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002985,
+      "loss": 0.5124,
+      "step": 998
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002988,
+      "loss": 0.4837,
+      "step": 999
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00029909999999999995,
+      "loss": 0.6397,
+      "step": 1000
+    },
+    {
+      "epoch": 0.97,
+      "eval_cer": 0.2862226856272025,
+      "eval_loss": 0.9038726091384888,
+      "eval_runtime": 227.493,
+      "eval_samples_per_second": 38.744,
+      "eval_steps_per_second": 4.844,
+      "eval_wer": 0.7170791725713783,
+      "step": 1000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00029939999999999996,
+      "loss": 0.5287,
+      "step": 1001
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00029969999999999997,
+      "loss": 0.6291,
+      "step": 1002
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003,
+      "loss": 0.5124,
+      "step": 1003
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.000299984693877551,
+      "loss": 0.4534,
+      "step": 1004
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029996938775510203,
+      "loss": 0.5156,
+      "step": 1005
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029995408163265307,
+      "loss": 0.6214,
+      "step": 1006
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029993877551020404,
+      "loss": 0.6877,
+      "step": 1007
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029992346938775507,
+      "loss": 0.5647,
+      "step": 1008
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0002999081632653061,
+      "loss": 0.6335,
+      "step": 1009
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029989285714285713,
+      "loss": 0.5297,
+      "step": 1010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0002998775510204081,
+      "loss": 0.5316,
+      "step": 1011
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029986224489795914,
+      "loss": 0.6434,
+      "step": 1012
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029984693877551017,
+      "loss": 0.6867,
+      "step": 1013
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0002998316326530612,
+      "loss": 0.4849,
+      "step": 1014
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00029981632653061223,
+      "loss": 0.6898,
+      "step": 1015
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002998010204081632,
+      "loss": 0.4438,
+      "step": 1016
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002997857142857143,
+      "loss": 0.6986,
+      "step": 1017
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00029977040816326527,
+      "loss": 0.542,
+      "step": 1018
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002997551020408163,
+      "loss": 0.6904,
+      "step": 1019
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00029973979591836733,
+      "loss": 0.5405,
+      "step": 1020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00029972448979591836,
+      "loss": 0.5512,
+      "step": 1021
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002997091836734694,
+      "loss": 0.5744,
+      "step": 1022
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00029969387755102037,
+      "loss": 0.523,
+      "step": 1023
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002996785714285714,
+      "loss": 0.6039,
+      "step": 1024
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00029966326530612243,
+      "loss": 0.779,
+      "step": 1025
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00029964795918367346,
+      "loss": 0.6198,
+      "step": 1026
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00029963265306122444,
+      "loss": 0.611,
+      "step": 1027
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00029961734693877547,
+      "loss": 0.4502,
+      "step": 1028
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002996020408163265,
+      "loss": 0.5414,
+      "step": 1029
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00029958673469387753,
+      "loss": 0.5267,
+      "step": 1030
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00029957142857142856,
+      "loss": 0.4332,
+      "step": 1031
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00029955612244897953,
+      "loss": 0.6271,
+      "step": 1032
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002995408163265306,
+      "loss": 0.4924,
+      "step": 1033
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002995255102040816,
+      "loss": 0.456,
+      "step": 1034
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002995102040816326,
+      "loss": 0.578,
+      "step": 1035
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00029949489795918366,
+      "loss": 0.4645,
+      "step": 1036
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002994795918367347,
+      "loss": 0.6642,
+      "step": 1037
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00029946428571428566,
+      "loss": 0.526,
+      "step": 1038
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002994489795918367,
+      "loss": 0.3732,
+      "step": 1039
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002994336734693877,
+      "loss": 0.4387,
+      "step": 1040
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00029941836734693876,
+      "loss": 0.4159,
+      "step": 1041
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002994030612244898,
+      "loss": 0.4911,
+      "step": 1042
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00029938775510204076,
+      "loss": 0.6693,
+      "step": 1043
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002993724489795918,
+      "loss": 0.4473,
+      "step": 1044
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002993571428571428,
+      "loss": 0.5325,
+      "step": 1045
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00029934183673469385,
+      "loss": 0.4118,
+      "step": 1046
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002993265306122449,
+      "loss": 0.5709,
+      "step": 1047
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00029931122448979586,
+      "loss": 0.4963,
+      "step": 1048
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00029929591836734695,
+      "loss": 0.608,
+      "step": 1049
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002992806122448979,
+      "loss": 0.5106,
+      "step": 1050
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00029926530612244895,
+      "loss": 0.4957,
+      "step": 1051
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00029925,
+      "loss": 0.5696,
+      "step": 1052
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000299234693877551,
+      "loss": 0.4621,
+      "step": 1053
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000299219387755102,
+      "loss": 0.3859,
+      "step": 1054
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000299204081632653,
+      "loss": 0.5419,
+      "step": 1055
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00029918877551020405,
+      "loss": 0.605,
+      "step": 1056
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002991734693877551,
+      "loss": 0.4808,
+      "step": 1057
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002991581632653061,
+      "loss": 0.4454,
+      "step": 1058
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002991428571428571,
+      "loss": 0.4573,
+      "step": 1059
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00029912755102040817,
+      "loss": 0.4189,
+      "step": 1060
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00029911224489795915,
+      "loss": 0.4822,
+      "step": 1061
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002990969387755102,
+      "loss": 0.5744,
+      "step": 1062
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002990816326530612,
+      "loss": 0.561,
+      "step": 1063
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00029906632653061224,
+      "loss": 0.5636,
+      "step": 1064
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00029905102040816327,
+      "loss": 0.5313,
+      "step": 1065
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00029903571428571425,
+      "loss": 0.5876,
+      "step": 1066
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002990204081632653,
+      "loss": 0.4708,
+      "step": 1067
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002990051020408163,
+      "loss": 0.79,
+      "step": 1068
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00029898979591836734,
+      "loss": 0.5442,
+      "step": 1069
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002989744897959183,
+      "loss": 0.4536,
+      "step": 1070
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00029895918367346935,
+      "loss": 0.5527,
+      "step": 1071
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002989438775510204,
+      "loss": 0.5393,
+      "step": 1072
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002989285714285714,
+      "loss": 0.5077,
+      "step": 1073
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00029891326530612244,
+      "loss": 0.4196,
+      "step": 1074
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002988979591836734,
+      "loss": 0.5824,
+      "step": 1075
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002988826530612245,
+      "loss": 0.4785,
+      "step": 1076
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002988673469387755,
+      "loss": 0.5385,
+      "step": 1077
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002988520408163265,
+      "loss": 0.5016,
+      "step": 1078
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00029883673469387754,
+      "loss": 0.524,
+      "step": 1079
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00029882142857142857,
+      "loss": 0.3923,
+      "step": 1080
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00029880612244897954,
+      "loss": 0.5762,
+      "step": 1081
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002987908163265306,
+      "loss": 0.4825,
+      "step": 1082
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002987755102040816,
+      "loss": 0.538,
+      "step": 1083
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00029876020408163263,
+      "loss": 0.3656,
+      "step": 1084
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00029874489795918367,
+      "loss": 0.4784,
+      "step": 1085
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00029872959183673464,
+      "loss": 0.5563,
+      "step": 1086
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00029871428571428567,
+      "loss": 0.6652,
+      "step": 1087
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002986989795918367,
+      "loss": 0.5741,
+      "step": 1088
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00029868367346938773,
+      "loss": 0.4646,
+      "step": 1089
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00029866836734693876,
+      "loss": 0.5654,
+      "step": 1090
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00029865306122448974,
+      "loss": 0.4527,
+      "step": 1091
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002986377551020408,
+      "loss": 0.4517,
+      "step": 1092
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002986224489795918,
+      "loss": 0.6242,
+      "step": 1093
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00029860714285714283,
+      "loss": 0.4826,
+      "step": 1094
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00029859183673469386,
+      "loss": 0.4871,
+      "step": 1095
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002985765306122449,
+      "loss": 0.4693,
+      "step": 1096
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00029856122448979587,
+      "loss": 0.5032,
+      "step": 1097
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002985459183673469,
+      "loss": 0.5669,
+      "step": 1098
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00029853061224489793,
+      "loss": 0.5017,
+      "step": 1099
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00029851530612244896,
+      "loss": 0.4086,
+      "step": 1100
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002985,
+      "loss": 0.4666,
+      "step": 1101
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00029848469387755097,
+      "loss": 0.4675,
+      "step": 1102
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00029846938775510205,
+      "loss": 0.4891,
+      "step": 1103
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00029845408163265303,
+      "loss": 0.501,
+      "step": 1104
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00029843877551020406,
+      "loss": 0.5004,
+      "step": 1105
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002984234693877551,
+      "loss": 0.5413,
+      "step": 1106
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002984081632653061,
+      "loss": 0.4712,
+      "step": 1107
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002983928571428571,
+      "loss": 0.4411,
+      "step": 1108
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00029837755102040813,
+      "loss": 0.4955,
+      "step": 1109
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00029836224489795916,
+      "loss": 0.3788,
+      "step": 1110
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002983469387755102,
+      "loss": 0.5948,
+      "step": 1111
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002983316326530612,
+      "loss": 0.5657,
+      "step": 1112
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002983163265306122,
+      "loss": 0.457,
+      "step": 1113
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002983010204081632,
+      "loss": 0.3877,
+      "step": 1114
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00029828571428571426,
+      "loss": 0.4318,
+      "step": 1115
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002982704081632653,
+      "loss": 0.4143,
+      "step": 1116
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002982551020408163,
+      "loss": 0.3926,
+      "step": 1117
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002982397959183673,
+      "loss": 0.5049,
+      "step": 1118
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002982244897959184,
+      "loss": 0.641,
+      "step": 1119
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00029820918367346936,
+      "loss": 0.4603,
+      "step": 1120
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002981938775510204,
+      "loss": 0.3469,
+      "step": 1121
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00029817857142857136,
+      "loss": 0.4874,
+      "step": 1122
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00029816326530612245,
+      "loss": 0.6017,
+      "step": 1123
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002981479591836734,
+      "loss": 0.6193,
+      "step": 1124
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00029813265306122445,
+      "loss": 0.5417,
+      "step": 1125
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002981173469387755,
+      "loss": 0.4847,
+      "step": 1126
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002981020408163265,
+      "loss": 0.454,
+      "step": 1127
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00029808673469387755,
+      "loss": 0.4219,
+      "step": 1128
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002980714285714285,
+      "loss": 0.4325,
+      "step": 1129
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00029805612244897955,
+      "loss": 0.5875,
+      "step": 1130
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002980408163265306,
+      "loss": 0.6195,
+      "step": 1131
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002980255102040816,
+      "loss": 0.4198,
+      "step": 1132
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00029801020408163264,
+      "loss": 0.5196,
+      "step": 1133
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002979948979591836,
+      "loss": 0.4299,
+      "step": 1134
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00029797959183673465,
+      "loss": 0.4803,
+      "step": 1135
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002979642857142857,
+      "loss": 0.536,
+      "step": 1136
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002979489795918367,
+      "loss": 0.6851,
+      "step": 1137
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002979336734693877,
+      "loss": 0.5323,
+      "step": 1138
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00029791836734693877,
+      "loss": 0.4237,
+      "step": 1139
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00029790306122448975,
+      "loss": 0.4849,
+      "step": 1140
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002978877551020408,
+      "loss": 0.4757,
+      "step": 1141
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002978724489795918,
+      "loss": 0.3958,
+      "step": 1142
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00029785714285714284,
+      "loss": 0.6944,
+      "step": 1143
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00029784183673469387,
+      "loss": 0.5348,
+      "step": 1144
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00029782653061224485,
+      "loss": 0.4493,
+      "step": 1145
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00029781122448979593,
+      "loss": 0.5448,
+      "step": 1146
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002977959183673469,
+      "loss": 0.4325,
+      "step": 1147
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00029778061224489794,
+      "loss": 0.5076,
+      "step": 1148
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00029776530612244897,
+      "loss": 0.626,
+      "step": 1149
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00029775,
+      "loss": 0.5235,
+      "step": 1150
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.000297734693877551,
+      "loss": 0.4362,
+      "step": 1151
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.000297719387755102,
+      "loss": 0.3866,
+      "step": 1152
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00029770408163265304,
+      "loss": 0.4474,
+      "step": 1153
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00029768877551020407,
+      "loss": 0.4396,
+      "step": 1154
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002976734693877551,
+      "loss": 0.4894,
+      "step": 1155
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002976581632653061,
+      "loss": 0.4751,
+      "step": 1156
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002976428571428571,
+      "loss": 0.4552,
+      "step": 1157
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00029762755102040814,
+      "loss": 0.4182,
+      "step": 1158
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00029761224489795917,
+      "loss": 0.4199,
+      "step": 1159
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002975969387755102,
+      "loss": 0.556,
+      "step": 1160
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002975816326530612,
+      "loss": 0.5279,
+      "step": 1161
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002975663265306122,
+      "loss": 0.4545,
+      "step": 1162
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00029755102040816323,
+      "loss": 0.5037,
+      "step": 1163
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00029753571428571427,
+      "loss": 0.4686,
+      "step": 1164
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00029752040816326524,
+      "loss": 0.4407,
+      "step": 1165
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002975051020408163,
+      "loss": 0.4537,
+      "step": 1166
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002974897959183673,
+      "loss": 0.5339,
+      "step": 1167
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00029747448979591833,
+      "loss": 0.6183,
+      "step": 1168
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00029745918367346936,
+      "loss": 0.4069,
+      "step": 1169
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002974438775510204,
+      "loss": 0.4307,
+      "step": 1170
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002974285714285714,
+      "loss": 0.5071,
+      "step": 1171
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002974132653061224,
+      "loss": 0.4526,
+      "step": 1172
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00029739795918367343,
+      "loss": 0.3862,
+      "step": 1173
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00029738265306122446,
+      "loss": 0.5373,
+      "step": 1174
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002973673469387755,
+      "loss": 0.4575,
+      "step": 1175
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002973520408163265,
+      "loss": 0.4543,
+      "step": 1176
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002973367346938775,
+      "loss": 0.4114,
+      "step": 1177
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00029732142857142853,
+      "loss": 0.5166,
+      "step": 1178
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00029730612244897956,
+      "loss": 0.4034,
+      "step": 1179
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002972908163265306,
+      "loss": 0.4874,
+      "step": 1180
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029727551020408157,
+      "loss": 0.5105,
+      "step": 1181
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029726020408163265,
+      "loss": 0.529,
+      "step": 1182
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029724489795918363,
+      "loss": 0.3806,
+      "step": 1183
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029722959183673466,
+      "loss": 0.4215,
+      "step": 1184
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002972142857142857,
+      "loss": 0.3813,
+      "step": 1185
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002971989795918367,
+      "loss": 0.4029,
+      "step": 1186
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029718367346938775,
+      "loss": 0.4726,
+      "step": 1187
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029716836734693873,
+      "loss": 0.3654,
+      "step": 1188
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002971530612244898,
+      "loss": 0.4241,
+      "step": 1189
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002971377551020408,
+      "loss": 0.4322,
+      "step": 1190
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002971224489795918,
+      "loss": 0.4369,
+      "step": 1191
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002971071428571428,
+      "loss": 0.4406,
+      "step": 1192
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002970918367346939,
+      "loss": 0.4798,
+      "step": 1193
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00029707653061224486,
+      "loss": 0.4536,
+      "step": 1194
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002970612244897959,
+      "loss": 0.3943,
+      "step": 1195
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002970459183673469,
+      "loss": 0.4072,
+      "step": 1196
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00029703061224489795,
+      "loss": 0.4637,
+      "step": 1197
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000297015306122449,
+      "loss": 0.5925,
+      "step": 1198
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00029699999999999996,
+      "loss": 0.5876,
+      "step": 1199
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000296984693877551,
+      "loss": 0.5588,
+      "step": 1200
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000296969387755102,
+      "loss": 0.5087,
+      "step": 1201
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029695408163265305,
+      "loss": 0.379,
+      "step": 1202
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002969387755102041,
+      "loss": 0.4279,
+      "step": 1203
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029692346938775505,
+      "loss": 0.4144,
+      "step": 1204
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002969081632653061,
+      "loss": 0.4508,
+      "step": 1205
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002968928571428571,
+      "loss": 0.5336,
+      "step": 1206
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029687755102040815,
+      "loss": 0.4398,
+      "step": 1207
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002968622448979591,
+      "loss": 0.3459,
+      "step": 1208
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002968469387755102,
+      "loss": 0.4466,
+      "step": 1209
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002968316326530612,
+      "loss": 0.4715,
+      "step": 1210
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002968163265306122,
+      "loss": 0.5584,
+      "step": 1211
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00029680102040816324,
+      "loss": 0.6002,
+      "step": 1212
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002967857142857143,
+      "loss": 0.4422,
+      "step": 1213
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002967704081632653,
+      "loss": 0.475,
+      "step": 1214
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002967551020408163,
+      "loss": 0.4418,
+      "step": 1215
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002967397959183673,
+      "loss": 0.3879,
+      "step": 1216
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00029672448979591834,
+      "loss": 0.4332,
+      "step": 1217
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00029670918367346937,
+      "loss": 0.5488,
+      "step": 1218
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00029669387755102035,
+      "loss": 0.4526,
+      "step": 1219
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002966785714285714,
+      "loss": 0.4717,
+      "step": 1220
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002966632653061224,
+      "loss": 0.472,
+      "step": 1221
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00029664795918367344,
+      "loss": 0.5275,
+      "step": 1222
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00029663265306122447,
+      "loss": 0.4128,
+      "step": 1223
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00029661734693877545,
+      "loss": 0.4087,
+      "step": 1224
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00029660204081632653,
+      "loss": 0.5032,
+      "step": 1225
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002965867346938775,
+      "loss": 0.4543,
+      "step": 1226
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00029657142857142854,
+      "loss": 0.4565,
+      "step": 1227
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00029655612244897957,
+      "loss": 0.3639,
+      "step": 1228
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002965408163265306,
+      "loss": 0.3386,
+      "step": 1229
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00029652551020408163,
+      "loss": 0.5796,
+      "step": 1230
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002965102040816326,
+      "loss": 0.4214,
+      "step": 1231
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029649489795918364,
+      "loss": 0.3924,
+      "step": 1232
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029647959183673467,
+      "loss": 0.4435,
+      "step": 1233
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002964642857142857,
+      "loss": 0.4521,
+      "step": 1234
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002964489795918367,
+      "loss": 0.3463,
+      "step": 1235
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029643367346938776,
+      "loss": 0.4938,
+      "step": 1236
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029641836734693874,
+      "loss": 0.5405,
+      "step": 1237
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029640306122448977,
+      "loss": 0.5669,
+      "step": 1238
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002963877551020408,
+      "loss": 0.4264,
+      "step": 1239
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029637244897959183,
+      "loss": 0.4566,
+      "step": 1240
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029635714285714286,
+      "loss": 0.455,
+      "step": 1241
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00029634183673469383,
+      "loss": 0.4771,
+      "step": 1242
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00029632653061224487,
+      "loss": 0.4822,
+      "step": 1243
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002963112244897959,
+      "loss": 0.5074,
+      "step": 1244
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002962959183673469,
+      "loss": 0.4053,
+      "step": 1245
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00029628061224489796,
+      "loss": 0.3539,
+      "step": 1246
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00029626530612244893,
+      "loss": 0.3407,
+      "step": 1247
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00029624999999999996,
+      "loss": 0.3777,
+      "step": 1248
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.000296234693877551,
+      "loss": 0.5174,
+      "step": 1249
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.000296219387755102,
+      "loss": 0.4031,
+      "step": 1250
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.000296204081632653,
+      "loss": 0.4643,
+      "step": 1251
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002961887755102041,
+      "loss": 0.4014,
+      "step": 1252
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00029617346938775506,
+      "loss": 0.425,
+      "step": 1253
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002961581632653061,
+      "loss": 0.4333,
+      "step": 1254
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002961428571428571,
+      "loss": 0.4212,
+      "step": 1255
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00029612755102040815,
+      "loss": 0.5081,
+      "step": 1256
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002961122448979592,
+      "loss": 0.3142,
+      "step": 1257
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00029609693877551016,
+      "loss": 0.3606,
+      "step": 1258
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002960816326530612,
+      "loss": 0.3598,
+      "step": 1259
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002960663265306122,
+      "loss": 0.3474,
+      "step": 1260
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00029605102040816325,
+      "loss": 0.4307,
+      "step": 1261
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00029603571428571423,
+      "loss": 0.6452,
+      "step": 1262
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00029602040816326526,
+      "loss": 0.4021,
+      "step": 1263
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002960051020408163,
+      "loss": 0.3773,
+      "step": 1264
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002959897959183673,
+      "loss": 0.4006,
+      "step": 1265
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00029597448979591835,
+      "loss": 0.4084,
+      "step": 1266
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00029595918367346933,
+      "loss": 0.3901,
+      "step": 1267
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002959438775510204,
+      "loss": 0.607,
+      "step": 1268
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002959285714285714,
+      "loss": 0.4327,
+      "step": 1269
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002959132653061224,
+      "loss": 0.3977,
+      "step": 1270
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00029589795918367345,
+      "loss": 0.3289,
+      "step": 1271
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002958826530612245,
+      "loss": 0.3555,
+      "step": 1272
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002958673469387755,
+      "loss": 0.3671,
+      "step": 1273
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002958520408163265,
+      "loss": 0.6874,
+      "step": 1274
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002958367346938775,
+      "loss": 0.4517,
+      "step": 1275
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00029582142857142855,
+      "loss": 0.3341,
+      "step": 1276
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002958061224489796,
+      "loss": 0.4745,
+      "step": 1277
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00029579081632653056,
+      "loss": 0.361,
+      "step": 1278
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00029577551020408164,
+      "loss": 0.4251,
+      "step": 1279
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002957602040816326,
+      "loss": 0.4385,
+      "step": 1280
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00029574489795918365,
+      "loss": 0.5018,
+      "step": 1281
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002957295918367347,
+      "loss": 0.5373,
+      "step": 1282
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002957142857142857,
+      "loss": 0.3317,
+      "step": 1283
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00029569897959183674,
+      "loss": 0.5335,
+      "step": 1284
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002956836734693877,
+      "loss": 0.3658,
+      "step": 1285
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00029566836734693875,
+      "loss": 0.4459,
+      "step": 1286
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002956530612244898,
+      "loss": 0.3711,
+      "step": 1287
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002956377551020408,
+      "loss": 0.3423,
+      "step": 1288
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002956224489795918,
+      "loss": 0.4239,
+      "step": 1289
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002956071428571428,
+      "loss": 0.3226,
+      "step": 1290
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00029559183673469384,
+      "loss": 0.3799,
+      "step": 1291
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002955765306122449,
+      "loss": 0.4328,
+      "step": 1292
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002955612244897959,
+      "loss": 0.4355,
+      "step": 1293
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002955459183673469,
+      "loss": 0.3688,
+      "step": 1294
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00029553061224489797,
+      "loss": 0.4291,
+      "step": 1295
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00029551530612244894,
+      "loss": 0.382,
+      "step": 1296
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00029549999999999997,
+      "loss": 0.4235,
+      "step": 1297
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.000295484693877551,
+      "loss": 0.3609,
+      "step": 1298
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00029546938775510203,
+      "loss": 0.4707,
+      "step": 1299
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00029545408163265306,
+      "loss": 0.4623,
+      "step": 1300
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00029543877551020404,
+      "loss": 0.4506,
+      "step": 1301
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00029542346938775507,
+      "loss": 0.3528,
+      "step": 1302
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002954081632653061,
+      "loss": 0.4012,
+      "step": 1303
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00029539285714285713,
+      "loss": 0.4231,
+      "step": 1304
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002953775510204081,
+      "loss": 0.4393,
+      "step": 1305
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00029536224489795914,
+      "loss": 0.4874,
+      "step": 1306
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00029534693877551017,
+      "loss": 0.419,
+      "step": 1307
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002953316326530612,
+      "loss": 0.4074,
+      "step": 1308
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00029531632653061223,
+      "loss": 0.373,
+      "step": 1309
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002953010204081632,
+      "loss": 0.4718,
+      "step": 1310
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002952857142857143,
+      "loss": 0.4094,
+      "step": 1311
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00029527040816326527,
+      "loss": 0.4488,
+      "step": 1312
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002952551020408163,
+      "loss": 0.4059,
+      "step": 1313
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00029523979591836733,
+      "loss": 0.3891,
+      "step": 1314
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00029522448979591836,
+      "loss": 0.4319,
+      "step": 1315
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00029520918367346934,
+      "loss": 0.4409,
+      "step": 1316
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00029519387755102037,
+      "loss": 0.5127,
+      "step": 1317
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002951785714285714,
+      "loss": 0.4426,
+      "step": 1318
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00029516326530612243,
+      "loss": 0.396,
+      "step": 1319
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00029514795918367346,
+      "loss": 0.4301,
+      "step": 1320
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00029513265306122443,
+      "loss": 0.4053,
+      "step": 1321
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002951173469387755,
+      "loss": 0.3911,
+      "step": 1322
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002951020408163265,
+      "loss": 0.3486,
+      "step": 1323
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002950867346938775,
+      "loss": 0.4487,
+      "step": 1324
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00029507142857142856,
+      "loss": 0.424,
+      "step": 1325
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002950561224489796,
+      "loss": 0.4204,
+      "step": 1326
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002950408163265306,
+      "loss": 0.3914,
+      "step": 1327
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002950255102040816,
+      "loss": 0.3891,
+      "step": 1328
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002950102040816326,
+      "loss": 0.4438,
+      "step": 1329
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00029499489795918366,
+      "loss": 0.5454,
+      "step": 1330
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002949795918367347,
+      "loss": 0.5435,
+      "step": 1331
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00029496428571428566,
+      "loss": 0.3935,
+      "step": 1332
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002949489795918367,
+      "loss": 0.4876,
+      "step": 1333
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002949336734693877,
+      "loss": 0.3919,
+      "step": 1334
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00029491836734693875,
+      "loss": 0.4437,
+      "step": 1335
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002949030612244898,
+      "loss": 0.3475,
+      "step": 1336
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00029488775510204076,
+      "loss": 0.5409,
+      "step": 1337
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00029487244897959185,
+      "loss": 0.391,
+      "step": 1338
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002948571428571428,
+      "loss": 0.337,
+      "step": 1339
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00029484183673469385,
+      "loss": 0.497,
+      "step": 1340
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002948265306122449,
+      "loss": 0.381,
+      "step": 1341
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002948112244897959,
+      "loss": 0.3733,
+      "step": 1342
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00029479591836734694,
+      "loss": 0.4175,
+      "step": 1343
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002947806122448979,
+      "loss": 0.4208,
+      "step": 1344
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00029476530612244895,
+      "loss": 0.3666,
+      "step": 1345
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00029475,
+      "loss": 0.3221,
+      "step": 1346
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.000294734693877551,
+      "loss": 0.4908,
+      "step": 1347
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.000294719387755102,
+      "loss": 0.3367,
+      "step": 1348
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.000294704081632653,
+      "loss": 0.5281,
+      "step": 1349
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00029468877551020405,
+      "loss": 0.4529,
+      "step": 1350
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002946734693877551,
+      "loss": 0.3962,
+      "step": 1351
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002946581632653061,
+      "loss": 0.4011,
+      "step": 1352
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002946428571428571,
+      "loss": 0.3474,
+      "step": 1353
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00029462755102040817,
+      "loss": 0.3771,
+      "step": 1354
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00029461224489795915,
+      "loss": 0.4941,
+      "step": 1355
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002945969387755102,
+      "loss": 0.5028,
+      "step": 1356
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002945816326530612,
+      "loss": 0.4014,
+      "step": 1357
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00029456632653061224,
+      "loss": 0.3263,
+      "step": 1358
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002945510204081632,
+      "loss": 0.3522,
+      "step": 1359
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00029453571428571425,
+      "loss": 0.4936,
+      "step": 1360
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002945204081632653,
+      "loss": 0.4518,
+      "step": 1361
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002945051020408163,
+      "loss": 0.4543,
+      "step": 1362
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00029448979591836734,
+      "loss": 0.4066,
+      "step": 1363
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002944744897959183,
+      "loss": 0.3374,
+      "step": 1364
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00029445918367346935,
+      "loss": 0.339,
+      "step": 1365
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002944438775510204,
+      "loss": 0.3234,
+      "step": 1366
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002944285714285714,
+      "loss": 0.441,
+      "step": 1367
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00029441326530612244,
+      "loss": 0.3553,
+      "step": 1368
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00029439795918367347,
+      "loss": 0.4364,
+      "step": 1369
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002943826530612245,
+      "loss": 0.391,
+      "step": 1370
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002943673469387755,
+      "loss": 0.4322,
+      "step": 1371
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002943520408163265,
+      "loss": 0.3469,
+      "step": 1372
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00029433673469387754,
+      "loss": 0.4353,
+      "step": 1373
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00029432142857142857,
+      "loss": 0.444,
+      "step": 1374
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00029430612244897954,
+      "loss": 0.4781,
+      "step": 1375
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00029429081632653057,
+      "loss": 0.3953,
+      "step": 1376
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002942755102040816,
+      "loss": 0.3827,
+      "step": 1377
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00029426020408163263,
+      "loss": 0.4478,
+      "step": 1378
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00029424489795918366,
+      "loss": 0.3325,
+      "step": 1379
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00029422959183673464,
+      "loss": 0.4576,
+      "step": 1380
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002942142857142857,
+      "loss": 0.4209,
+      "step": 1381
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002941989795918367,
+      "loss": 0.4684,
+      "step": 1382
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00029418367346938773,
+      "loss": 0.3909,
+      "step": 1383
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00029416836734693876,
+      "loss": 0.4127,
+      "step": 1384
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002941530612244898,
+      "loss": 0.2793,
+      "step": 1385
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00029413775510204077,
+      "loss": 0.4008,
+      "step": 1386
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002941224489795918,
+      "loss": 0.4498,
+      "step": 1387
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00029410714285714283,
+      "loss": 0.3577,
+      "step": 1388
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00029409183673469386,
+      "loss": 0.4247,
+      "step": 1389
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002940765306122449,
+      "loss": 0.3033,
+      "step": 1390
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00029406122448979587,
+      "loss": 0.4677,
+      "step": 1391
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002940459183673469,
+      "loss": 0.3738,
+      "step": 1392
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00029403061224489793,
+      "loss": 0.4846,
+      "step": 1393
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00029401530612244896,
+      "loss": 0.4101,
+      "step": 1394
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.000294,
+      "loss": 0.3656,
+      "step": 1395
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00029398469387755097,
+      "loss": 0.4741,
+      "step": 1396
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00029396938775510205,
+      "loss": 0.3422,
+      "step": 1397
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00029395408163265303,
+      "loss": 0.3364,
+      "step": 1398
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00029393877551020406,
+      "loss": 0.4804,
+      "step": 1399
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00029392346938775503,
+      "loss": 0.3988,
+      "step": 1400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002939081632653061,
+      "loss": 0.4726,
+      "step": 1401
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002938928571428571,
+      "loss": 0.3449,
+      "step": 1402
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002938775510204081,
+      "loss": 0.4281,
+      "step": 1403
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00029386224489795916,
+      "loss": 0.3477,
+      "step": 1404
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002938469387755102,
+      "loss": 0.5089,
+      "step": 1405
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002938316326530612,
+      "loss": 0.4936,
+      "step": 1406
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002938163265306122,
+      "loss": 0.5278,
+      "step": 1407
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002938010204081632,
+      "loss": 0.4245,
+      "step": 1408
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00029378571428571426,
+      "loss": 0.2914,
+      "step": 1409
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002937704081632653,
+      "loss": 0.3553,
+      "step": 1410
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002937551020408163,
+      "loss": 0.3562,
+      "step": 1411
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00029373979591836735,
+      "loss": 0.3768,
+      "step": 1412
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002937244897959183,
+      "loss": 0.523,
+      "step": 1413
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00029370918367346935,
+      "loss": 0.363,
+      "step": 1414
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002936938775510204,
+      "loss": 0.374,
+      "step": 1415
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002936785714285714,
+      "loss": 0.3975,
+      "step": 1416
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00029366326530612245,
+      "loss": 0.3482,
+      "step": 1417
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002936479591836734,
+      "loss": 0.4919,
+      "step": 1418
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00029363265306122445,
+      "loss": 0.4643,
+      "step": 1419
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002936173469387755,
+      "loss": 0.4458,
+      "step": 1420
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002936020408163265,
+      "loss": 0.3605,
+      "step": 1421
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00029358673469387754,
+      "loss": 0.3585,
+      "step": 1422
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002935714285714285,
+      "loss": 0.3322,
+      "step": 1423
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002935561224489796,
+      "loss": 0.6029,
+      "step": 1424
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002935408163265306,
+      "loss": 0.3884,
+      "step": 1425
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002935255102040816,
+      "loss": 0.4141,
+      "step": 1426
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00029351020408163264,
+      "loss": 0.4133,
+      "step": 1427
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002934948979591837,
+      "loss": 0.3631,
+      "step": 1428
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00029347959183673465,
+      "loss": 0.3228,
+      "step": 1429
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002934642857142857,
+      "loss": 0.4858,
+      "step": 1430
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002934489795918367,
+      "loss": 0.3762,
+      "step": 1431
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00029343367346938774,
+      "loss": 0.4793,
+      "step": 1432
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00029341836734693877,
+      "loss": 0.3388,
+      "step": 1433
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00029340306122448975,
+      "loss": 0.3882,
+      "step": 1434
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002933877551020408,
+      "loss": 0.4615,
+      "step": 1435
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002933724489795918,
+      "loss": 0.358,
+      "step": 1436
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00029335714285714284,
+      "loss": 0.4512,
+      "step": 1437
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00029334183673469387,
+      "loss": 0.3134,
+      "step": 1438
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00029332653061224485,
+      "loss": 0.3368,
+      "step": 1439
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002933112244897959,
+      "loss": 0.2841,
+      "step": 1440
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002932959183673469,
+      "loss": 0.3778,
+      "step": 1441
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00029328061224489794,
+      "loss": 0.3123,
+      "step": 1442
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002932653061224489,
+      "loss": 0.4466,
+      "step": 1443
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00029325,
+      "loss": 0.5081,
+      "step": 1444
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.000293234693877551,
+      "loss": 0.4311,
+      "step": 1445
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.000293219387755102,
+      "loss": 0.3259,
+      "step": 1446
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00029320408163265304,
+      "loss": 0.3641,
+      "step": 1447
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00029318877551020407,
+      "loss": 0.4351,
+      "step": 1448
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002931734693877551,
+      "loss": 0.5116,
+      "step": 1449
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002931581632653061,
+      "loss": 0.4372,
+      "step": 1450
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002931428571428571,
+      "loss": 0.4321,
+      "step": 1451
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00029312755102040814,
+      "loss": 0.3144,
+      "step": 1452
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00029311224489795917,
+      "loss": 0.383,
+      "step": 1453
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002930969387755102,
+      "loss": 0.4199,
+      "step": 1454
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00029308163265306117,
+      "loss": 0.4625,
+      "step": 1455
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002930663265306122,
+      "loss": 0.4363,
+      "step": 1456
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00029305102040816323,
+      "loss": 0.4366,
+      "step": 1457
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00029303571428571426,
+      "loss": 0.3162,
+      "step": 1458
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002930204081632653,
+      "loss": 0.4124,
+      "step": 1459
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002930051020408163,
+      "loss": 0.42,
+      "step": 1460
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002929897959183673,
+      "loss": 0.491,
+      "step": 1461
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00029297448979591833,
+      "loss": 0.4307,
+      "step": 1462
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00029295918367346936,
+      "loss": 0.3227,
+      "step": 1463
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002929438775510204,
+      "loss": 0.4863,
+      "step": 1464
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002929285714285714,
+      "loss": 0.3463,
+      "step": 1465
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002929132653061224,
+      "loss": 0.4043,
+      "step": 1466
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002928979591836735,
+      "loss": 0.3177,
+      "step": 1467
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00029288265306122446,
+      "loss": 0.5416,
+      "step": 1468
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002928673469387755,
+      "loss": 0.5129,
+      "step": 1469
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00029285204081632647,
+      "loss": 0.3627,
+      "step": 1470
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00029283673469387755,
+      "loss": 0.4244,
+      "step": 1471
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00029282142857142853,
+      "loss": 0.2804,
+      "step": 1472
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00029280612244897956,
+      "loss": 0.307,
+      "step": 1473
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002927908163265306,
+      "loss": 0.4363,
+      "step": 1474
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002927755102040816,
+      "loss": 0.4126,
+      "step": 1475
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00029276020408163265,
+      "loss": 0.3994,
+      "step": 1476
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00029274489795918363,
+      "loss": 0.3058,
+      "step": 1477
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00029272959183673466,
+      "loss": 0.4873,
+      "step": 1478
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002927142857142857,
+      "loss": 0.3728,
+      "step": 1479
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002926989795918367,
+      "loss": 0.6182,
+      "step": 1480
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00029268367346938775,
+      "loss": 0.4203,
+      "step": 1481
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002926683673469387,
+      "loss": 0.4629,
+      "step": 1482
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00029265306122448976,
+      "loss": 0.424,
+      "step": 1483
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002926377551020408,
+      "loss": 0.3484,
+      "step": 1484
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002926224489795918,
+      "loss": 0.348,
+      "step": 1485
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002926071428571428,
+      "loss": 0.3735,
+      "step": 1486
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002925918367346939,
+      "loss": 0.3693,
+      "step": 1487
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00029257653061224486,
+      "loss": 0.3353,
+      "step": 1488
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002925612244897959,
+      "loss": 0.2981,
+      "step": 1489
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002925459183673469,
+      "loss": 0.4627,
+      "step": 1490
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00029253061224489795,
+      "loss": 0.3897,
+      "step": 1491
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000292515306122449,
+      "loss": 0.3262,
+      "step": 1492
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00029249999999999995,
+      "loss": 0.3559,
+      "step": 1493
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000292484693877551,
+      "loss": 0.4617,
+      "step": 1494
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000292469387755102,
+      "loss": 0.3359,
+      "step": 1495
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00029245408163265305,
+      "loss": 0.4352,
+      "step": 1496
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000292438775510204,
+      "loss": 0.4065,
+      "step": 1497
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00029242346938775505,
+      "loss": 0.3429,
+      "step": 1498
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002924081632653061,
+      "loss": 0.571,
+      "step": 1499
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002923928571428571,
+      "loss": 0.4033,
+      "step": 1500
+    },
+    {
+      "epoch": 1.46,
+      "eval_cer": 0.2762843314478978,
+      "eval_loss": 0.891418993473053,
+      "eval_runtime": 227.2445,
+      "eval_samples_per_second": 38.786,
+      "eval_steps_per_second": 4.849,
+      "eval_wer": 0.686209889156203,
+      "step": 1500
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00029237755102040814,
+      "loss": 0.4125,
+      "step": 1501
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002923622448979591,
+      "loss": 0.4302,
+      "step": 1502
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002923469387755102,
+      "loss": 0.4181,
+      "step": 1503
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002923316326530612,
+      "loss": 0.331,
+      "step": 1504
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002923163265306122,
+      "loss": 0.2892,
+      "step": 1505
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00029230102040816324,
+      "loss": 0.3893,
+      "step": 1506
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002922857142857143,
+      "loss": 0.4904,
+      "step": 1507
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002922704081632653,
+      "loss": 0.4246,
+      "step": 1508
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002922551020408163,
+      "loss": 0.4147,
+      "step": 1509
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002922397959183673,
+      "loss": 0.48,
+      "step": 1510
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00029222448979591834,
+      "loss": 0.4196,
+      "step": 1511
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00029220918367346937,
+      "loss": 0.3782,
+      "step": 1512
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00029219387755102035,
+      "loss": 0.2884,
+      "step": 1513
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00029217857142857143,
+      "loss": 0.333,
+      "step": 1514
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002921632653061224,
+      "loss": 0.4469,
+      "step": 1515
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00029214795918367344,
+      "loss": 0.3573,
+      "step": 1516
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00029213265306122447,
+      "loss": 0.3617,
+      "step": 1517
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002921173469387755,
+      "loss": 0.4426,
+      "step": 1518
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00029210204081632653,
+      "loss": 0.4295,
+      "step": 1519
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002920867346938775,
+      "loss": 0.35,
+      "step": 1520
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00029207142857142854,
+      "loss": 0.4735,
+      "step": 1521
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00029205612244897957,
+      "loss": 0.3261,
+      "step": 1522
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002920408163265306,
+      "loss": 0.4538,
+      "step": 1523
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00029202551020408163,
+      "loss": 0.4225,
+      "step": 1524
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002920102040816326,
+      "loss": 0.4102,
+      "step": 1525
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00029199489795918364,
+      "loss": 0.3392,
+      "step": 1526
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00029197959183673467,
+      "loss": 0.3895,
+      "step": 1527
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002919642857142857,
+      "loss": 0.4037,
+      "step": 1528
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002919489795918367,
+      "loss": 0.417,
+      "step": 1529
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00029193367346938776,
+      "loss": 0.4405,
+      "step": 1530
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00029191836734693874,
+      "loss": 0.4585,
+      "step": 1531
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00029190306122448977,
+      "loss": 0.3567,
+      "step": 1532
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002918877551020408,
+      "loss": 0.4421,
+      "step": 1533
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002918724489795918,
+      "loss": 0.4376,
+      "step": 1534
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00029185714285714286,
+      "loss": 0.3195,
+      "step": 1535
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00029184183673469383,
+      "loss": 0.4108,
+      "step": 1536
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00029182653061224486,
+      "loss": 0.6341,
+      "step": 1537
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002918112244897959,
+      "loss": 0.4583,
+      "step": 1538
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002917959183673469,
+      "loss": 0.4582,
+      "step": 1539
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002917806122448979,
+      "loss": 0.4151,
+      "step": 1540
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00029176530612244893,
+      "loss": 0.4372,
+      "step": 1541
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00029174999999999996,
+      "loss": 0.4155,
+      "step": 1542
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.000291734693877551,
+      "loss": 0.3573,
+      "step": 1543
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.000291719387755102,
+      "loss": 0.3987,
+      "step": 1544
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.000291704081632653,
+      "loss": 0.3299,
+      "step": 1545
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002916887755102041,
+      "loss": 0.3403,
+      "step": 1546
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00029167346938775506,
+      "loss": 0.3966,
+      "step": 1547
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002916581632653061,
+      "loss": 0.2886,
+      "step": 1548
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002916428571428571,
+      "loss": 0.5182,
+      "step": 1549
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00029162755102040815,
+      "loss": 0.3993,
+      "step": 1550
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002916122448979592,
+      "loss": 0.3405,
+      "step": 1551
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00029159693877551016,
+      "loss": 0.4495,
+      "step": 1552
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002915816326530612,
+      "loss": 0.4853,
+      "step": 1553
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002915663265306122,
+      "loss": 0.4293,
+      "step": 1554
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00029155102040816325,
+      "loss": 0.5109,
+      "step": 1555
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00029153571428571423,
+      "loss": 0.4931,
+      "step": 1556
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002915204081632653,
+      "loss": 0.3834,
+      "step": 1557
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002915051020408163,
+      "loss": 0.3866,
+      "step": 1558
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002914897959183673,
+      "loss": 0.3726,
+      "step": 1559
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00029147448979591835,
+      "loss": 0.3942,
+      "step": 1560
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002914591836734694,
+      "loss": 0.402,
+      "step": 1561
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002914438775510204,
+      "loss": 0.4476,
+      "step": 1562
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002914285714285714,
+      "loss": 0.4097,
+      "step": 1563
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002914132653061224,
+      "loss": 0.3496,
+      "step": 1564
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00029139795918367345,
+      "loss": 0.2944,
+      "step": 1565
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002913826530612245,
+      "loss": 0.4046,
+      "step": 1566
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00029136734693877546,
+      "loss": 0.4398,
+      "step": 1567
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002913520408163265,
+      "loss": 0.6117,
+      "step": 1568
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002913367346938775,
+      "loss": 0.3212,
+      "step": 1569
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00029132142857142855,
+      "loss": 0.3595,
+      "step": 1570
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002913061224489796,
+      "loss": 0.3827,
+      "step": 1571
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00029129081632653055,
+      "loss": 0.3525,
+      "step": 1572
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00029127551020408164,
+      "loss": 0.3736,
+      "step": 1573
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002912602040816326,
+      "loss": 0.547,
+      "step": 1574
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00029124489795918365,
+      "loss": 0.4328,
+      "step": 1575
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002912295918367347,
+      "loss": 0.4725,
+      "step": 1576
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002912142857142857,
+      "loss": 0.3305,
+      "step": 1577
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00029119897959183674,
+      "loss": 0.3924,
+      "step": 1578
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002911836734693877,
+      "loss": 0.3973,
+      "step": 1579
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00029116836734693874,
+      "loss": 0.4493,
+      "step": 1580
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002911530612244898,
+      "loss": 0.4565,
+      "step": 1581
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002911377551020408,
+      "loss": 0.4532,
+      "step": 1582
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002911224489795918,
+      "loss": 0.5118,
+      "step": 1583
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002911071428571428,
+      "loss": 0.3861,
+      "step": 1584
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00029109183673469384,
+      "loss": 0.2904,
+      "step": 1585
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002910765306122449,
+      "loss": 0.4696,
+      "step": 1586
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002910612244897959,
+      "loss": 0.467,
+      "step": 1587
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002910459183673469,
+      "loss": 0.3722,
+      "step": 1588
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00029103061224489796,
+      "loss": 0.3988,
+      "step": 1589
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00029101530612244894,
+      "loss": 0.3426,
+      "step": 1590
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 0.3537,
+      "step": 1591
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.000290984693877551,
+      "loss": 0.4486,
+      "step": 1592
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00029096938775510203,
+      "loss": 0.4366,
+      "step": 1593
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.000290954081632653,
+      "loss": 0.4121,
+      "step": 1594
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00029093877551020404,
+      "loss": 0.3847,
+      "step": 1595
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00029092346938775507,
+      "loss": 0.3204,
+      "step": 1596
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002909081632653061,
+      "loss": 0.2926,
+      "step": 1597
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00029089285714285713,
+      "loss": 0.3824,
+      "step": 1598
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002908775510204081,
+      "loss": 0.5966,
+      "step": 1599
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002908622448979592,
+      "loss": 0.36,
+      "step": 1600
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00029084693877551017,
+      "loss": 0.4555,
+      "step": 1601
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002908316326530612,
+      "loss": 0.3789,
+      "step": 1602
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00029081632653061223,
+      "loss": 0.3393,
+      "step": 1603
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00029080102040816326,
+      "loss": 0.3364,
+      "step": 1604
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002907857142857143,
+      "loss": 0.4189,
+      "step": 1605
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00029077040816326527,
+      "loss": 0.461,
+      "step": 1606
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002907551020408163,
+      "loss": 0.5019,
+      "step": 1607
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00029073979591836733,
+      "loss": 0.3613,
+      "step": 1608
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00029072448979591836,
+      "loss": 0.3972,
+      "step": 1609
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00029070918367346934,
+      "loss": 0.3675,
+      "step": 1610
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00029069387755102037,
+      "loss": 0.4722,
+      "step": 1611
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002906785714285714,
+      "loss": 0.4333,
+      "step": 1612
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002906632653061224,
+      "loss": 0.3555,
+      "step": 1613
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00029064795918367346,
+      "loss": 0.3947,
+      "step": 1614
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00029063265306122443,
+      "loss": 0.3812,
+      "step": 1615
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002906173469387755,
+      "loss": 0.408,
+      "step": 1616
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002906020408163265,
+      "loss": 0.3473,
+      "step": 1617
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002905867346938775,
+      "loss": 0.61,
+      "step": 1618
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00029057142857142856,
+      "loss": 0.3768,
+      "step": 1619
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002905561224489796,
+      "loss": 0.3443,
+      "step": 1620
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00029054081632653056,
+      "loss": 0.3404,
+      "step": 1621
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002905255102040816,
+      "loss": 0.4339,
+      "step": 1622
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002905102040816326,
+      "loss": 0.4037,
+      "step": 1623
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00029049489795918365,
+      "loss": 0.4873,
+      "step": 1624
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002904795918367347,
+      "loss": 0.4416,
+      "step": 1625
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00029046428571428566,
+      "loss": 0.3432,
+      "step": 1626
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002904489795918367,
+      "loss": 0.3448,
+      "step": 1627
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002904336734693877,
+      "loss": 0.3376,
+      "step": 1628
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00029041836734693875,
+      "loss": 0.291,
+      "step": 1629
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002904030612244898,
+      "loss": 0.4964,
+      "step": 1630
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00029038775510204076,
+      "loss": 0.5184,
+      "step": 1631
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00029037244897959184,
+      "loss": 0.4615,
+      "step": 1632
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002903571428571428,
+      "loss": 0.3317,
+      "step": 1633
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00029034183673469385,
+      "loss": 0.3525,
+      "step": 1634
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002903265306122449,
+      "loss": 0.3445,
+      "step": 1635
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002903112244897959,
+      "loss": 0.3321,
+      "step": 1636
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002902959183673469,
+      "loss": 0.4441,
+      "step": 1637
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002902806122448979,
+      "loss": 0.3641,
+      "step": 1638
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00029026530612244895,
+      "loss": 0.342,
+      "step": 1639
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00029025,
+      "loss": 0.4448,
+      "step": 1640
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.000290234693877551,
+      "loss": 0.4592,
+      "step": 1641
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.000290219387755102,
+      "loss": 0.4299,
+      "step": 1642
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00029020408163265307,
+      "loss": 0.4145,
+      "step": 1643
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00029018877551020405,
+      "loss": 0.3747,
+      "step": 1644
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0002901734693877551,
+      "loss": 0.4597,
+      "step": 1645
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0002901581632653061,
+      "loss": 0.4239,
+      "step": 1646
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00029014285714285714,
+      "loss": 0.4063,
+      "step": 1647
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00029012755102040817,
+      "loss": 0.3828,
+      "step": 1648
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00029011224489795915,
+      "loss": 0.6534,
+      "step": 1649
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0002900969387755102,
+      "loss": 0.336,
+      "step": 1650
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0002900816326530612,
+      "loss": 0.3778,
+      "step": 1651
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00029006632653061224,
+      "loss": 0.3372,
+      "step": 1652
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0002900510204081632,
+      "loss": 0.3545,
+      "step": 1653
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00029003571428571425,
+      "loss": 0.3993,
+      "step": 1654
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002900204081632653,
+      "loss": 0.5592,
+      "step": 1655
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002900051020408163,
+      "loss": 0.4311,
+      "step": 1656
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00028998979591836734,
+      "loss": 0.4169,
+      "step": 1657
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002899744897959183,
+      "loss": 0.2771,
+      "step": 1658
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002899591836734694,
+      "loss": 0.3345,
+      "step": 1659
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002899438775510204,
+      "loss": 0.38,
+      "step": 1660
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002899285714285714,
+      "loss": 0.3052,
+      "step": 1661
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00028991326530612244,
+      "loss": 0.413,
+      "step": 1662
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00028989795918367347,
+      "loss": 0.4144,
+      "step": 1663
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00028988265306122444,
+      "loss": 0.3482,
+      "step": 1664
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0002898673469387755,
+      "loss": 0.3186,
+      "step": 1665
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0002898520408163265,
+      "loss": 0.3103,
+      "step": 1666
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00028983673469387753,
+      "loss": 0.3618,
+      "step": 1667
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00028982142857142856,
+      "loss": 0.3772,
+      "step": 1668
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00028980612244897954,
+      "loss": 0.2942,
+      "step": 1669
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00028979081632653057,
+      "loss": 0.377,
+      "step": 1670
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0002897755102040816,
+      "loss": 0.3688,
+      "step": 1671
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00028976020408163263,
+      "loss": 0.3322,
+      "step": 1672
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00028974489795918366,
+      "loss": 0.3443,
+      "step": 1673
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00028972959183673464,
+      "loss": 0.344,
+      "step": 1674
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002897142857142857,
+      "loss": 0.484,
+      "step": 1675
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002896989795918367,
+      "loss": 0.3975,
+      "step": 1676
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00028968367346938773,
+      "loss": 0.3884,
+      "step": 1677
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002896683673469387,
+      "loss": 0.259,
+      "step": 1678
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002896530612244898,
+      "loss": 0.3518,
+      "step": 1679
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00028963775510204077,
+      "loss": 0.4155,
+      "step": 1680
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002896224489795918,
+      "loss": 0.4012,
+      "step": 1681
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00028960714285714283,
+      "loss": 0.3946,
+      "step": 1682
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00028959183673469386,
+      "loss": 0.3895,
+      "step": 1683
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002895765306122449,
+      "loss": 0.3985,
+      "step": 1684
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00028956122448979587,
+      "loss": 0.3904,
+      "step": 1685
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00028954591836734695,
+      "loss": 0.3308,
+      "step": 1686
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00028953061224489793,
+      "loss": 0.5378,
+      "step": 1687
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00028951530612244896,
+      "loss": 0.3255,
+      "step": 1688
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0002895,
+      "loss": 0.4573,
+      "step": 1689
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000289484693877551,
+      "loss": 0.3845,
+      "step": 1690
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000289469387755102,
+      "loss": 0.272,
+      "step": 1691
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000289454081632653,
+      "loss": 0.3449,
+      "step": 1692
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00028943877551020406,
+      "loss": 0.5548,
+      "step": 1693
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0002894234693877551,
+      "loss": 0.3349,
+      "step": 1694
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002894081632653061,
+      "loss": 0.4248,
+      "step": 1695
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002893928571428571,
+      "loss": 0.2978,
+      "step": 1696
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002893775510204081,
+      "loss": 0.323,
+      "step": 1697
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00028936224489795916,
+      "loss": 0.3159,
+      "step": 1698
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002893469387755102,
+      "loss": 0.3773,
+      "step": 1699
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002893316326530612,
+      "loss": 0.3856,
+      "step": 1700
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002893163265306122,
+      "loss": 0.3386,
+      "step": 1701
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002893010204081633,
+      "loss": 0.3436,
+      "step": 1702
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00028928571428571425,
+      "loss": 0.4372,
+      "step": 1703
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002892704081632653,
+      "loss": 0.2893,
+      "step": 1704
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002892551020408163,
+      "loss": 0.48,
+      "step": 1705
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00028923979591836735,
+      "loss": 0.4507,
+      "step": 1706
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002892244897959183,
+      "loss": 0.4994,
+      "step": 1707
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00028920918367346935,
+      "loss": 0.3333,
+      "step": 1708
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002891938775510204,
+      "loss": 0.4061,
+      "step": 1709
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002891785714285714,
+      "loss": 0.3792,
+      "step": 1710
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00028916326530612244,
+      "loss": 0.306,
+      "step": 1711
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002891479591836734,
+      "loss": 0.4058,
+      "step": 1712
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00028913265306122445,
+      "loss": 0.3389,
+      "step": 1713
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002891173469387755,
+      "loss": 0.3418,
+      "step": 1714
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002891020408163265,
+      "loss": 0.3501,
+      "step": 1715
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00028908673469387754,
+      "loss": 0.3541,
+      "step": 1716
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002890714285714285,
+      "loss": 0.3076,
+      "step": 1717
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00028905612244897955,
+      "loss": 0.5332,
+      "step": 1718
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002890408163265306,
+      "loss": 0.3963,
+      "step": 1719
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002890255102040816,
+      "loss": 0.2998,
+      "step": 1720
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002890102040816326,
+      "loss": 0.2761,
+      "step": 1721
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00028899489795918367,
+      "loss": 0.2884,
+      "step": 1722
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00028897959183673465,
+      "loss": 0.407,
+      "step": 1723
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002889642857142857,
+      "loss": 0.4129,
+      "step": 1724
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002889489795918367,
+      "loss": 0.3253,
+      "step": 1725
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00028893367346938774,
+      "loss": 0.3903,
+      "step": 1726
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00028891836734693877,
+      "loss": 0.3839,
+      "step": 1727
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00028890306122448975,
+      "loss": 0.3844,
+      "step": 1728
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00028888775510204083,
+      "loss": 0.2242,
+      "step": 1729
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0002888724489795918,
+      "loss": 0.3584,
+      "step": 1730
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00028885714285714284,
+      "loss": 0.3453,
+      "step": 1731
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00028884183673469387,
+      "loss": 0.3844,
+      "step": 1732
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0002888265306122449,
+      "loss": 0.298,
+      "step": 1733
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0002888112244897959,
+      "loss": 0.3397,
+      "step": 1734
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0002887959183673469,
+      "loss": 0.3119,
+      "step": 1735
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00028878061224489794,
+      "loss": 0.4061,
+      "step": 1736
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00028876530612244897,
+      "loss": 0.645,
+      "step": 1737
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00028875,
+      "loss": 0.3954,
+      "step": 1738
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.000288734693877551,
+      "loss": 0.3602,
+      "step": 1739
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.000288719387755102,
+      "loss": 0.3129,
+      "step": 1740
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00028870408163265304,
+      "loss": 0.3352,
+      "step": 1741
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00028868877551020407,
+      "loss": 0.4124,
+      "step": 1742
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0002886734693877551,
+      "loss": 0.3374,
+      "step": 1743
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0002886581632653061,
+      "loss": 0.3841,
+      "step": 1744
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00028864285714285716,
+      "loss": 0.4022,
+      "step": 1745
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00028862755102040813,
+      "loss": 0.3779,
+      "step": 1746
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00028861224489795916,
+      "loss": 0.3153,
+      "step": 1747
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00028859693877551014,
+      "loss": 0.256,
+      "step": 1748
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002885816326530612,
+      "loss": 0.4732,
+      "step": 1749
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002885663265306122,
+      "loss": 0.418,
+      "step": 1750
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00028855102040816323,
+      "loss": 0.3531,
+      "step": 1751
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00028853571428571426,
+      "loss": 0.2997,
+      "step": 1752
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002885204081632653,
+      "loss": 0.3335,
+      "step": 1753
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002885051020408163,
+      "loss": 0.4207,
+      "step": 1754
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002884897959183673,
+      "loss": 0.6745,
+      "step": 1755
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00028847448979591833,
+      "loss": 0.4455,
+      "step": 1756
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00028845918367346936,
+      "loss": 0.3713,
+      "step": 1757
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0002884438775510204,
+      "loss": 0.3768,
+      "step": 1758
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0002884285714285714,
+      "loss": 0.3771,
+      "step": 1759
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0002884132653061224,
+      "loss": 0.2816,
+      "step": 1760
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00028839795918367343,
+      "loss": 0.4291,
+      "step": 1761
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00028838265306122446,
+      "loss": 0.3703,
+      "step": 1762
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0002883673469387755,
+      "loss": 0.4222,
+      "step": 1763
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00028835204081632647,
+      "loss": 0.3015,
+      "step": 1764
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00028833673469387755,
+      "loss": 0.2849,
+      "step": 1765
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00028832142857142853,
+      "loss": 0.3241,
+      "step": 1766
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00028830612244897956,
+      "loss": 0.3467,
+      "step": 1767
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002882908163265306,
+      "loss": 0.5758,
+      "step": 1768
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002882755102040816,
+      "loss": 0.493,
+      "step": 1769
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00028826020408163265,
+      "loss": 0.3098,
+      "step": 1770
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002882448979591836,
+      "loss": 0.2969,
+      "step": 1771
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00028822959183673466,
+      "loss": 0.3334,
+      "step": 1772
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002882142857142857,
+      "loss": 0.3645,
+      "step": 1773
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002881989795918367,
+      "loss": 0.4345,
+      "step": 1774
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002881836734693877,
+      "loss": 0.3513,
+      "step": 1775
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002881683673469388,
+      "loss": 0.3769,
+      "step": 1776
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00028815306122448976,
+      "loss": 0.4886,
+      "step": 1777
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002881377551020408,
+      "loss": 0.31,
+      "step": 1778
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002881224489795918,
+      "loss": 0.3562,
+      "step": 1779
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00028810714285714285,
+      "loss": 0.4081,
+      "step": 1780
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002880918367346939,
+      "loss": 0.401,
+      "step": 1781
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00028807653061224485,
+      "loss": 0.392,
+      "step": 1782
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002880612244897959,
+      "loss": 0.3885,
+      "step": 1783
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002880459183673469,
+      "loss": 0.2313,
+      "step": 1784
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00028803061224489795,
+      "loss": 0.4525,
+      "step": 1785
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.000288015306122449,
+      "loss": 0.4344,
+      "step": 1786
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00028799999999999995,
+      "loss": 0.4263,
+      "step": 1787
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.000287984693877551,
+      "loss": 0.3308,
+      "step": 1788
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.000287969387755102,
+      "loss": 0.3049,
+      "step": 1789
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00028795408163265304,
+      "loss": 0.3081,
+      "step": 1790
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.000287938775510204,
+      "loss": 0.4167,
+      "step": 1791
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002879234693877551,
+      "loss": 0.3712,
+      "step": 1792
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002879081632653061,
+      "loss": 0.4905,
+      "step": 1793
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002878928571428571,
+      "loss": 0.396,
+      "step": 1794
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00028787755102040814,
+      "loss": 0.3582,
+      "step": 1795
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002878622448979592,
+      "loss": 0.4114,
+      "step": 1796
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002878469387755102,
+      "loss": 0.3786,
+      "step": 1797
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002878316326530612,
+      "loss": 0.3737,
+      "step": 1798
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002878163265306122,
+      "loss": 0.3669,
+      "step": 1799
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00028780102040816324,
+      "loss": 0.3989,
+      "step": 1800
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00028778571428571427,
+      "loss": 0.3378,
+      "step": 1801
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002877704081632653,
+      "loss": 0.3235,
+      "step": 1802
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002877551020408163,
+      "loss": 0.2748,
+      "step": 1803
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002877397959183673,
+      "loss": 0.3023,
+      "step": 1804
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00028772448979591834,
+      "loss": 0.4853,
+      "step": 1805
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00028770918367346937,
+      "loss": 0.5106,
+      "step": 1806
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00028769387755102035,
+      "loss": 0.2893,
+      "step": 1807
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00028767857142857143,
+      "loss": 0.3564,
+      "step": 1808
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0002876632653061224,
+      "loss": 0.3108,
+      "step": 1809
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00028764795918367344,
+      "loss": 0.3476,
+      "step": 1810
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00028763265306122447,
+      "loss": 0.5123,
+      "step": 1811
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0002876173469387755,
+      "loss": 0.4979,
+      "step": 1812
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00028760204081632653,
+      "loss": 0.3111,
+      "step": 1813
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0002875867346938775,
+      "loss": 0.3764,
+      "step": 1814
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00028757142857142854,
+      "loss": 0.4089,
+      "step": 1815
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00028755612244897957,
+      "loss": 0.4025,
+      "step": 1816
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0002875408163265306,
+      "loss": 0.3544,
+      "step": 1817
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002875255102040816,
+      "loss": 0.63,
+      "step": 1818
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002875102040816326,
+      "loss": 0.4435,
+      "step": 1819
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00028749489795918364,
+      "loss": 0.2975,
+      "step": 1820
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00028747959183673467,
+      "loss": 0.3663,
+      "step": 1821
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002874642857142857,
+      "loss": 0.3037,
+      "step": 1822
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00028744897959183673,
+      "loss": 0.3741,
+      "step": 1823
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00028743367346938776,
+      "loss": 0.2852,
+      "step": 1824
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00028741836734693873,
+      "loss": 0.3643,
+      "step": 1825
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00028740306122448976,
+      "loss": 0.2889,
+      "step": 1826
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002873877551020408,
+      "loss": 0.3321,
+      "step": 1827
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002873724489795918,
+      "loss": 0.3748,
+      "step": 1828
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00028735714285714286,
+      "loss": 0.3419,
+      "step": 1829
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00028734183673469383,
+      "loss": 0.3627,
+      "step": 1830
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00028732653061224486,
+      "loss": 0.4849,
+      "step": 1831
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0002873112244897959,
+      "loss": 0.4187,
+      "step": 1832
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0002872959183673469,
+      "loss": 0.3044,
+      "step": 1833
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0002872806122448979,
+      "loss": 0.3322,
+      "step": 1834
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.000287265306122449,
+      "loss": 0.3335,
+      "step": 1835
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00028724999999999996,
+      "loss": 0.3871,
+      "step": 1836
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.000287234693877551,
+      "loss": 0.4598,
+      "step": 1837
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.000287219387755102,
+      "loss": 0.3844,
+      "step": 1838
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00028720408163265305,
+      "loss": 0.3587,
+      "step": 1839
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002871887755102041,
+      "loss": 0.3018,
+      "step": 1840
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00028717346938775506,
+      "loss": 0.3235,
+      "step": 1841
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002871581632653061,
+      "loss": 0.3369,
+      "step": 1842
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002871428571428571,
+      "loss": 0.5679,
+      "step": 1843
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00028712755102040815,
+      "loss": 0.3614,
+      "step": 1844
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00028711224489795913,
+      "loss": 0.3337,
+      "step": 1845
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00028709693877551016,
+      "loss": 0.3459,
+      "step": 1846
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002870816326530612,
+      "loss": 0.318,
+      "step": 1847
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002870663265306122,
+      "loss": 0.3137,
+      "step": 1848
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00028705102040816325,
+      "loss": 0.3941,
+      "step": 1849
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002870357142857142,
+      "loss": 0.3366,
+      "step": 1850
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002870204081632653,
+      "loss": 0.3361,
+      "step": 1851
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002870051020408163,
+      "loss": 0.3062,
+      "step": 1852
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002869897959183673,
+      "loss": 0.2812,
+      "step": 1853
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00028697448979591835,
+      "loss": 0.3966,
+      "step": 1854
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002869591836734694,
+      "loss": 0.4479,
+      "step": 1855
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002869438775510204,
+      "loss": 0.4475,
+      "step": 1856
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002869285714285714,
+      "loss": 0.3168,
+      "step": 1857
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002869132653061224,
+      "loss": 0.3246,
+      "step": 1858
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00028689795918367345,
+      "loss": 0.339,
+      "step": 1859
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002868826530612245,
+      "loss": 0.3372,
+      "step": 1860
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00028686734693877545,
+      "loss": 0.3421,
+      "step": 1861
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002868520408163265,
+      "loss": 0.4156,
+      "step": 1862
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002868367346938775,
+      "loss": 0.3892,
+      "step": 1863
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00028682142857142855,
+      "loss": 0.371,
+      "step": 1864
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002868061224489796,
+      "loss": 0.3015,
+      "step": 1865
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002867908163265306,
+      "loss": 0.3207,
+      "step": 1866
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00028677551020408164,
+      "loss": 0.3544,
+      "step": 1867
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002867602040816326,
+      "loss": 0.3825,
+      "step": 1868
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00028674489795918364,
+      "loss": 0.3459,
+      "step": 1869
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002867295918367347,
+      "loss": 0.2874,
+      "step": 1870
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002867142857142857,
+      "loss": 0.395,
+      "step": 1871
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002866989795918367,
+      "loss": 0.3028,
+      "step": 1872
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002866836734693877,
+      "loss": 0.3399,
+      "step": 1873
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00028666836734693874,
+      "loss": 0.3393,
+      "step": 1874
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002866530612244898,
+      "loss": 0.3139,
+      "step": 1875
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002866377551020408,
+      "loss": 0.3371,
+      "step": 1876
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002866224489795918,
+      "loss": 0.3377,
+      "step": 1877
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00028660714285714287,
+      "loss": 0.3019,
+      "step": 1878
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00028659183673469384,
+      "loss": 0.3044,
+      "step": 1879
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00028657653061224487,
+      "loss": 0.2605,
+      "step": 1880
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0002865612244897959,
+      "loss": 0.4519,
+      "step": 1881
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00028654591836734693,
+      "loss": 0.3462,
+      "step": 1882
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00028653061224489796,
+      "loss": 0.2652,
+      "step": 1883
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00028651530612244894,
+      "loss": 0.4372,
+      "step": 1884
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00028649999999999997,
+      "loss": 0.3518,
+      "step": 1885
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.000286484693877551,
+      "loss": 0.3898,
+      "step": 1886
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00028646938775510203,
+      "loss": 0.4147,
+      "step": 1887
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.000286454081632653,
+      "loss": 0.4163,
+      "step": 1888
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00028643877551020404,
+      "loss": 0.3575,
+      "step": 1889
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00028642346938775507,
+      "loss": 0.3798,
+      "step": 1890
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002864081632653061,
+      "loss": 0.2866,
+      "step": 1891
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00028639285714285713,
+      "loss": 0.3655,
+      "step": 1892
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002863775510204081,
+      "loss": 0.3957,
+      "step": 1893
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002863622448979592,
+      "loss": 0.2813,
+      "step": 1894
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00028634693877551017,
+      "loss": 0.329,
+      "step": 1895
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002863316326530612,
+      "loss": 0.3922,
+      "step": 1896
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00028631632653061223,
+      "loss": 0.3426,
+      "step": 1897
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00028630102040816326,
+      "loss": 0.2647,
+      "step": 1898
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00028628571428571424,
+      "loss": 0.3749,
+      "step": 1899
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00028627040816326527,
+      "loss": 0.395,
+      "step": 1900
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002862551020408163,
+      "loss": 0.3523,
+      "step": 1901
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00028623979591836733,
+      "loss": 0.2907,
+      "step": 1902
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00028622448979591836,
+      "loss": 0.4316,
+      "step": 1903
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00028620918367346933,
+      "loss": 0.3186,
+      "step": 1904
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00028619387755102036,
+      "loss": 0.3377,
+      "step": 1905
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002861785714285714,
+      "loss": 0.3875,
+      "step": 1906
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002861632653061224,
+      "loss": 0.3239,
+      "step": 1907
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00028614795918367346,
+      "loss": 0.4099,
+      "step": 1908
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00028613265306122443,
+      "loss": 0.3017,
+      "step": 1909
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002861173469387755,
+      "loss": 0.282,
+      "step": 1910
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002861020408163265,
+      "loss": 0.3541,
+      "step": 1911
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002860867346938775,
+      "loss": 0.4359,
+      "step": 1912
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00028607142857142856,
+      "loss": 0.43,
+      "step": 1913
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002860561224489796,
+      "loss": 0.2748,
+      "step": 1914
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00028604081632653056,
+      "loss": 0.3332,
+      "step": 1915
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002860255102040816,
+      "loss": 0.5066,
+      "step": 1916
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002860102040816326,
+      "loss": 0.2922,
+      "step": 1917
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00028599489795918365,
+      "loss": 0.4995,
+      "step": 1918
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002859795918367347,
+      "loss": 0.3786,
+      "step": 1919
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00028596428571428566,
+      "loss": 0.3636,
+      "step": 1920
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00028594897959183675,
+      "loss": 0.3805,
+      "step": 1921
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002859336734693877,
+      "loss": 0.271,
+      "step": 1922
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00028591836734693875,
+      "loss": 0.3563,
+      "step": 1923
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002859030612244898,
+      "loss": 0.5531,
+      "step": 1924
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002858877551020408,
+      "loss": 0.4314,
+      "step": 1925
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00028587244897959184,
+      "loss": 0.3113,
+      "step": 1926
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002858571428571428,
+      "loss": 0.3389,
+      "step": 1927
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00028584183673469385,
+      "loss": 0.3368,
+      "step": 1928
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002858265306122449,
+      "loss": 0.2972,
+      "step": 1929
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002858112244897959,
+      "loss": 0.3607,
+      "step": 1930
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002857959183673469,
+      "loss": 0.3759,
+      "step": 1931
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002857806122448979,
+      "loss": 0.4351,
+      "step": 1932
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00028576530612244895,
+      "loss": 0.3314,
+      "step": 1933
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00028575,
+      "loss": 0.3571,
+      "step": 1934
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.000285734693877551,
+      "loss": 0.3475,
+      "step": 1935
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.000285719387755102,
+      "loss": 0.3681,
+      "step": 1936
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00028570408163265307,
+      "loss": 0.4378,
+      "step": 1937
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00028568877551020405,
+      "loss": 0.3932,
+      "step": 1938
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002856734693877551,
+      "loss": 0.3377,
+      "step": 1939
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002856581632653061,
+      "loss": 0.4631,
+      "step": 1940
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00028564285714285714,
+      "loss": 0.314,
+      "step": 1941
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002856275510204081,
+      "loss": 0.3122,
+      "step": 1942
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00028561224489795915,
+      "loss": 0.4968,
+      "step": 1943
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002855969387755102,
+      "loss": 0.3938,
+      "step": 1944
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002855816326530612,
+      "loss": 0.3618,
+      "step": 1945
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00028556632653061224,
+      "loss": 0.2365,
+      "step": 1946
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002855510204081632,
+      "loss": 0.295,
+      "step": 1947
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00028553571428571424,
+      "loss": 0.3582,
+      "step": 1948
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002855204081632653,
+      "loss": 0.4467,
+      "step": 1949
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002855051020408163,
+      "loss": 0.4616,
+      "step": 1950
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00028548979591836734,
+      "loss": 0.2552,
+      "step": 1951
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002854744897959183,
+      "loss": 0.3507,
+      "step": 1952
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002854591836734694,
+      "loss": 0.3299,
+      "step": 1953
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002854438775510204,
+      "loss": 0.3182,
+      "step": 1954
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002854285714285714,
+      "loss": 0.3618,
+      "step": 1955
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00028541326530612243,
+      "loss": 0.4392,
+      "step": 1956
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00028539795918367347,
+      "loss": 0.3072,
+      "step": 1957
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00028538265306122444,
+      "loss": 0.3393,
+      "step": 1958
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00028536734693877547,
+      "loss": 0.3972,
+      "step": 1959
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002853520408163265,
+      "loss": 0.3651,
+      "step": 1960
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00028533673469387753,
+      "loss": 0.453,
+      "step": 1961
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00028532142857142856,
+      "loss": 0.4849,
+      "step": 1962
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00028530612244897954,
+      "loss": 0.3734,
+      "step": 1963
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0002852908163265306,
+      "loss": 0.3251,
+      "step": 1964
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0002852755102040816,
+      "loss": 0.3593,
+      "step": 1965
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00028526020408163263,
+      "loss": 0.3509,
+      "step": 1966
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00028524489795918366,
+      "loss": 0.3848,
+      "step": 1967
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0002852295918367347,
+      "loss": 0.4321,
+      "step": 1968
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00028521428571428567,
+      "loss": 0.3828,
+      "step": 1969
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0002851989795918367,
+      "loss": 0.3775,
+      "step": 1970
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00028518367346938773,
+      "loss": 0.324,
+      "step": 1971
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00028516836734693876,
+      "loss": 0.282,
+      "step": 1972
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0002851530612244898,
+      "loss": 0.3336,
+      "step": 1973
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00028513775510204077,
+      "loss": 0.4351,
+      "step": 1974
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0002851224489795918,
+      "loss": 0.2822,
+      "step": 1975
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00028510714285714283,
+      "loss": 0.3836,
+      "step": 1976
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00028509183673469386,
+      "loss": 0.2759,
+      "step": 1977
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0002850765306122449,
+      "loss": 0.3398,
+      "step": 1978
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00028506122448979587,
+      "loss": 0.2904,
+      "step": 1979
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00028504591836734695,
+      "loss": 0.2595,
+      "step": 1980
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00028503061224489793,
+      "loss": 0.3562,
+      "step": 1981
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00028501530612244896,
+      "loss": 0.2799,
+      "step": 1982
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.000285,
+      "loss": 0.3014,
+      "step": 1983
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.000284984693877551,
+      "loss": 0.2423,
+      "step": 1984
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.000284969387755102,
+      "loss": 0.3152,
+      "step": 1985
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.000284954081632653,
+      "loss": 0.4409,
+      "step": 1986
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00028493877551020406,
+      "loss": 0.3615,
+      "step": 1987
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002849234693877551,
+      "loss": 0.3282,
+      "step": 1988
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002849081632653061,
+      "loss": 0.3346,
+      "step": 1989
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002848928571428571,
+      "loss": 0.3153,
+      "step": 1990
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002848775510204081,
+      "loss": 0.4243,
+      "step": 1991
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00028486224489795916,
+      "loss": 0.3841,
+      "step": 1992
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002848469387755102,
+      "loss": 0.3671,
+      "step": 1993
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002848316326530612,
+      "loss": 0.3346,
+      "step": 1994
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002848163265306122,
+      "loss": 0.324,
+      "step": 1995
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002848010204081632,
+      "loss": 0.2714,
+      "step": 1996
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00028478571428571425,
+      "loss": 0.3404,
+      "step": 1997
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002847704081632653,
+      "loss": 0.3068,
+      "step": 1998
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00028475510204081626,
+      "loss": 0.3235,
+      "step": 1999
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00028473979591836735,
+      "loss": 0.3473,
+      "step": 2000
+    },
+    {
+      "epoch": 1.94,
+      "eval_cer": 0.25362864376583466,
+      "eval_loss": 0.8016663789749146,
+      "eval_runtime": 227.1779,
+      "eval_samples_per_second": 38.798,
+      "eval_steps_per_second": 4.851,
+      "eval_wer": 0.6504740247613398,
+      "step": 2000
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002847244897959183,
+      "loss": 0.3243,
+      "step": 2001
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00028470918367346935,
+      "loss": 0.3979,
+      "step": 2002
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002846938775510204,
+      "loss": 0.2793,
+      "step": 2003
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0002846785714285714,
+      "loss": 0.3854,
+      "step": 2004
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00028466326530612244,
+      "loss": 0.338,
+      "step": 2005
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0002846479591836734,
+      "loss": 0.4094,
+      "step": 2006
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0002846326530612245,
+      "loss": 0.2985,
+      "step": 2007
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0002846173469387755,
+      "loss": 0.3279,
+      "step": 2008
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0002846020408163265,
+      "loss": 0.2861,
+      "step": 2009
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00028458673469387754,
+      "loss": 0.353,
+      "step": 2010
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00028457142857142857,
+      "loss": 0.4341,
+      "step": 2011
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00028455612244897955,
+      "loss": 0.4717,
+      "step": 2012
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0002845408163265306,
+      "loss": 0.3717,
+      "step": 2013
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0002845255102040816,
+      "loss": 0.3574,
+      "step": 2014
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00028451020408163264,
+      "loss": 0.3226,
+      "step": 2015
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00028449489795918367,
+      "loss": 0.26,
+      "step": 2016
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00028447959183673465,
+      "loss": 0.4076,
+      "step": 2017
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0002844642857142857,
+      "loss": 0.3277,
+      "step": 2018
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0002844489795918367,
+      "loss": 0.3353,
+      "step": 2019
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00028443367346938774,
+      "loss": 0.3895,
+      "step": 2020
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00028441836734693877,
+      "loss": 0.2533,
+      "step": 2021
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00028440306122448975,
+      "loss": 0.4252,
+      "step": 2022
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00028438775510204083,
+      "loss": 0.3748,
+      "step": 2023
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0002843724489795918,
+      "loss": 0.3881,
+      "step": 2024
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00028435714285714284,
+      "loss": 0.3759,
+      "step": 2025
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0002843418367346938,
+      "loss": 0.3573,
+      "step": 2026
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0002843265306122449,
+      "loss": 0.3868,
+      "step": 2027
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0002843112244897959,
+      "loss": 0.2471,
+      "step": 2028
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0002842959183673469,
+      "loss": 0.2891,
+      "step": 2029
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00028428061224489794,
+      "loss": 0.3787,
+      "step": 2030
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00028426530612244897,
+      "loss": 0.366,
+      "step": 2031
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00028425,
+      "loss": 0.3993,
+      "step": 2032
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.000284234693877551,
+      "loss": 0.3082,
+      "step": 2033
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.000284219387755102,
+      "loss": 0.3164,
+      "step": 2034
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00028420408163265303,
+      "loss": 0.2487,
+      "step": 2035
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00028418877551020407,
+      "loss": 0.3759,
+      "step": 2036
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0002841734693877551,
+      "loss": 0.5681,
+      "step": 2037
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00028415816326530607,
+      "loss": 0.2938,
+      "step": 2038
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0002841428571428571,
+      "loss": 0.3144,
+      "step": 2039
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00028412755102040813,
+      "loss": 0.3668,
+      "step": 2040
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00028411224489795916,
+      "loss": 0.2883,
+      "step": 2041
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00028409693877551014,
+      "loss": 0.3611,
+      "step": 2042
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0002840816326530612,
+      "loss": 0.3789,
+      "step": 2043
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0002840663265306122,
+      "loss": 0.4117,
+      "step": 2044
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00028405102040816323,
+      "loss": 0.3564,
+      "step": 2045
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00028403571428571426,
+      "loss": 0.3149,
+      "step": 2046
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0002840204081632653,
+      "loss": 0.2385,
+      "step": 2047
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0002840051020408163,
+      "loss": 0.2888,
+      "step": 2048
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0002839897959183673,
+      "loss": 0.4265,
+      "step": 2049
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0002839744897959184,
+      "loss": 0.2751,
+      "step": 2050
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00028395918367346936,
+      "loss": 0.3415,
+      "step": 2051
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0002839438775510204,
+      "loss": 0.288,
+      "step": 2052
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00028392857142857137,
+      "loss": 0.2886,
+      "step": 2053
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00028391326530612245,
+      "loss": 0.3286,
+      "step": 2054
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00028389795918367343,
+      "loss": 0.3832,
+      "step": 2055
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00028388265306122446,
+      "loss": 0.4207,
+      "step": 2056
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002838673469387755,
+      "loss": 0.3111,
+      "step": 2057
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002838520408163265,
+      "loss": 0.269,
+      "step": 2058
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00028383673469387755,
+      "loss": 0.3279,
+      "step": 2059
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00028382142857142853,
+      "loss": 0.3363,
+      "step": 2060
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00028380612244897956,
+      "loss": 0.3996,
+      "step": 2061
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002837908163265306,
+      "loss": 0.3716,
+      "step": 2062
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002837755102040816,
+      "loss": 0.2672,
+      "step": 2063
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00028376020408163265,
+      "loss": 0.2447,
+      "step": 2064
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002837448979591836,
+      "loss": 0.2671,
+      "step": 2065
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.00028372959183673466,
+      "loss": 0.2275,
+      "step": 2066
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002837142857142857,
+      "loss": 0.4111,
+      "step": 2067
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002836989795918367,
+      "loss": 0.2704,
+      "step": 2068
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002836836734693877,
+      "loss": 0.3551,
+      "step": 2069
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002836683673469388,
+      "loss": 0.285,
+      "step": 2070
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.00028365306122448976,
+      "loss": 0.2043,
+      "step": 2071
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002836377551020408,
+      "loss": 0.2296,
+      "step": 2072
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002836224489795918,
+      "loss": 0.3299,
+      "step": 2073
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.00028360714285714285,
+      "loss": 0.2493,
+      "step": 2074
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002835918367346939,
+      "loss": 0.1972,
+      "step": 2075
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.00028357653061224485,
+      "loss": 0.2601,
+      "step": 2076
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0002835612244897959,
+      "loss": 0.1871,
+      "step": 2077
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0002835459183673469,
+      "loss": 0.1956,
+      "step": 2078
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.00028353061224489795,
+      "loss": 0.4653,
+      "step": 2079
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0002835153061224489,
+      "loss": 0.311,
+      "step": 2080
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.00028349999999999995,
+      "loss": 0.2743,
+      "step": 2081
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.000283484693877551,
+      "loss": 0.2377,
+      "step": 2082
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.000283469387755102,
+      "loss": 0.2784,
+      "step": 2083
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.00028345408163265304,
+      "loss": 0.1885,
+      "step": 2084
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.000283438775510204,
+      "loss": 0.3399,
+      "step": 2085
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.0002834234693877551,
+      "loss": 0.2028,
+      "step": 2086
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.0002834081632653061,
+      "loss": 0.2833,
+      "step": 2087
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.0002833928571428571,
+      "loss": 0.2231,
+      "step": 2088
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00028337755102040814,
+      "loss": 0.2052,
+      "step": 2089
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00028336224489795917,
+      "loss": 0.1796,
+      "step": 2090
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.0002833469387755102,
+      "loss": 0.2817,
+      "step": 2091
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.0002833316326530612,
+      "loss": 0.2857,
+      "step": 2092
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.0002833163265306122,
+      "loss": 0.2624,
+      "step": 2093
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00028330102040816324,
+      "loss": 0.2489,
+      "step": 2094
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00028328571428571427,
+      "loss": 0.2768,
+      "step": 2095
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00028327040816326525,
+      "loss": 0.3034,
+      "step": 2096
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00028325510204081633,
+      "loss": 0.2607,
+      "step": 2097
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0002832397959183673,
+      "loss": 0.3883,
+      "step": 2098
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00028322448979591834,
+      "loss": 0.2821,
+      "step": 2099
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00028320918367346937,
+      "loss": 0.2709,
+      "step": 2100
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0002831938775510204,
+      "loss": 0.2346,
+      "step": 2101
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00028317857142857143,
+      "loss": 0.2555,
+      "step": 2102
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0002831632653061224,
+      "loss": 0.2741,
+      "step": 2103
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00028314795918367344,
+      "loss": 0.3692,
+      "step": 2104
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00028313265306122447,
+      "loss": 0.2982,
+      "step": 2105
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0002831173469387755,
+      "loss": 0.2436,
+      "step": 2106
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00028310204081632653,
+      "loss": 0.2549,
+      "step": 2107
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.0002830867346938775,
+      "loss": 0.2223,
+      "step": 2108
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00028307142857142854,
+      "loss": 0.2143,
+      "step": 2109
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00028305612244897957,
+      "loss": 0.3206,
+      "step": 2110
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.0002830408163265306,
+      "loss": 0.324,
+      "step": 2111
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.0002830255102040816,
+      "loss": 0.244,
+      "step": 2112
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00028301020408163266,
+      "loss": 0.2907,
+      "step": 2113
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00028299489795918363,
+      "loss": 0.3155,
+      "step": 2114
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00028297959183673467,
+      "loss": 0.167,
+      "step": 2115
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.0002829642857142857,
+      "loss": 0.2015,
+      "step": 2116
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0002829489795918367,
+      "loss": 0.5478,
+      "step": 2117
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00028293367346938776,
+      "loss": 0.3433,
+      "step": 2118
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00028291836734693873,
+      "loss": 0.2215,
+      "step": 2119
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00028290306122448976,
+      "loss": 0.2445,
+      "step": 2120
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0002828877551020408,
+      "loss": 0.2795,
+      "step": 2121
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0002828724489795918,
+      "loss": 0.2613,
+      "step": 2122
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0002828571428571428,
+      "loss": 0.3764,
+      "step": 2123
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00028284183673469383,
+      "loss": 0.3167,
+      "step": 2124
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00028282653061224486,
+      "loss": 0.1982,
+      "step": 2125
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0002828112244897959,
+      "loss": 0.2568,
+      "step": 2126
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0002827959183673469,
+      "loss": 0.2072,
+      "step": 2127
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0002827806122448979,
+      "loss": 0.2697,
+      "step": 2128
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.000282765306122449,
+      "loss": 0.3946,
+      "step": 2129
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.00028274999999999996,
+      "loss": 0.2966,
+      "step": 2130
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.000282734693877551,
+      "loss": 0.3236,
+      "step": 2131
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.000282719387755102,
+      "loss": 0.2371,
+      "step": 2132
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.00028270408163265305,
+      "loss": 0.3094,
+      "step": 2133
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0002826887755102041,
+      "loss": 0.2605,
+      "step": 2134
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.00028267346938775506,
+      "loss": 0.3053,
+      "step": 2135
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0002826581632653061,
+      "loss": 0.3219,
+      "step": 2136
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0002826428571428571,
+      "loss": 0.2471,
+      "step": 2137
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.00028262755102040815,
+      "loss": 0.2058,
+      "step": 2138
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.00028261224489795913,
+      "loss": 0.2256,
+      "step": 2139
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0002825969387755102,
+      "loss": 0.2301,
+      "step": 2140
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0002825816326530612,
+      "loss": 0.2849,
+      "step": 2141
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0002825663265306122,
+      "loss": 0.5457,
+      "step": 2142
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.00028255102040816325,
+      "loss": 0.2764,
+      "step": 2143
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0002825357142857143,
+      "loss": 0.2432,
+      "step": 2144
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0002825204081632653,
+      "loss": 0.239,
+      "step": 2145
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0002825051020408163,
+      "loss": 0.2157,
+      "step": 2146
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0002824897959183673,
+      "loss": 0.306,
+      "step": 2147
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00028247448979591835,
+      "loss": 0.2545,
+      "step": 2148
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0002824591836734694,
+      "loss": 0.445,
+      "step": 2149
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00028244387755102036,
+      "loss": 0.3191,
+      "step": 2150
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0002824285714285714,
+      "loss": 0.2979,
+      "step": 2151
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0002824132653061224,
+      "loss": 0.378,
+      "step": 2152
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00028239795918367345,
+      "loss": 0.2016,
+      "step": 2153
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0002823826530612245,
+      "loss": 0.4719,
+      "step": 2154
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00028236734693877545,
+      "loss": 0.3133,
+      "step": 2155
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00028235204081632654,
+      "loss": 0.2234,
+      "step": 2156
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0002823367346938775,
+      "loss": 0.2488,
+      "step": 2157
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00028232142857142855,
+      "loss": 0.2325,
+      "step": 2158
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0002823061224489796,
+      "loss": 0.2004,
+      "step": 2159
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0002822908163265306,
+      "loss": 0.3543,
+      "step": 2160
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00028227551020408164,
+      "loss": 0.3477,
+      "step": 2161
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0002822602040816326,
+      "loss": 0.4152,
+      "step": 2162
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00028224489795918364,
+      "loss": 0.2263,
+      "step": 2163
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0002822295918367347,
+      "loss": 0.208,
+      "step": 2164
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0002822142857142857,
+      "loss": 0.2073,
+      "step": 2165
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0002821989795918367,
+      "loss": 0.2868,
+      "step": 2166
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0002821836734693877,
+      "loss": 0.3914,
+      "step": 2167
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00028216836734693874,
+      "loss": 0.2257,
+      "step": 2168
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00028215306122448977,
+      "loss": 0.2013,
+      "step": 2169
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0002821377551020408,
+      "loss": 0.2792,
+      "step": 2170
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0002821224489795918,
+      "loss": 0.2774,
+      "step": 2171
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00028210714285714286,
+      "loss": 0.2441,
+      "step": 2172
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00028209183673469384,
+      "loss": 0.3072,
+      "step": 2173
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00028207653061224487,
+      "loss": 0.4356,
+      "step": 2174
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0002820612244897959,
+      "loss": 0.2363,
+      "step": 2175
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00028204591836734693,
+      "loss": 0.1976,
+      "step": 2176
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0002820306122448979,
+      "loss": 0.2325,
+      "step": 2177
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00028201530612244894,
+      "loss": 0.1957,
+      "step": 2178
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 0.3738,
+      "step": 2179
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.000281984693877551,
+      "loss": 0.2964,
+      "step": 2180
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.00028196938775510203,
+      "loss": 0.229,
+      "step": 2181
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.000281954081632653,
+      "loss": 0.2416,
+      "step": 2182
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0002819387755102041,
+      "loss": 0.2055,
+      "step": 2183
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.00028192346938775507,
+      "loss": 0.2587,
+      "step": 2184
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0002819081632653061,
+      "loss": 0.4339,
+      "step": 2185
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.00028189285714285713,
+      "loss": 0.3598,
+      "step": 2186
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.00028187755102040816,
+      "loss": 0.3641,
+      "step": 2187
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0002818622448979592,
+      "loss": 0.2462,
+      "step": 2188
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00028184693877551017,
+      "loss": 0.2362,
+      "step": 2189
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0002818316326530612,
+      "loss": 0.292,
+      "step": 2190
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00028181632653061223,
+      "loss": 0.305,
+      "step": 2191
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00028180102040816326,
+      "loss": 0.301,
+      "step": 2192
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00028178571428571423,
+      "loss": 0.2295,
+      "step": 2193
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00028177040816326527,
+      "loss": 0.2383,
+      "step": 2194
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0002817551020408163,
+      "loss": 0.3015,
+      "step": 2195
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0002817397959183673,
+      "loss": 0.2843,
+      "step": 2196
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00028172448979591836,
+      "loss": 0.2218,
+      "step": 2197
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00028170918367346933,
+      "loss": 0.3291,
+      "step": 2198
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0002816938775510204,
+      "loss": 0.2496,
+      "step": 2199
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0002816785714285714,
+      "loss": 0.2943,
+      "step": 2200
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0002816632653061224,
+      "loss": 0.2457,
+      "step": 2201
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.00028164795918367346,
+      "loss": 0.2675,
+      "step": 2202
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0002816326530612245,
+      "loss": 0.287,
+      "step": 2203
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0002816173469387755,
+      "loss": 0.3859,
+      "step": 2204
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0002816020408163265,
+      "loss": 0.3797,
+      "step": 2205
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0002815867346938775,
+      "loss": 0.3023,
+      "step": 2206
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.00028157142857142855,
+      "loss": 0.2418,
+      "step": 2207
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0002815561224489796,
+      "loss": 0.3694,
+      "step": 2208
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.00028154081632653056,
+      "loss": 0.2216,
+      "step": 2209
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0002815255102040816,
+      "loss": 0.2025,
+      "step": 2210
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0002815102040816326,
+      "loss": 0.3164,
+      "step": 2211
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.00028149489795918365,
+      "loss": 0.2272,
+      "step": 2212
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0002814795918367347,
+      "loss": 0.3026,
+      "step": 2213
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.00028146428571428566,
+      "loss": 0.2151,
+      "step": 2214
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.00028144897959183674,
+      "loss": 0.1917,
+      "step": 2215
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0002814336734693877,
+      "loss": 0.2613,
+      "step": 2216
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.00028141836734693875,
+      "loss": 0.3633,
+      "step": 2217
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0002814030612244898,
+      "loss": 0.2224,
+      "step": 2218
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0002813877551020408,
+      "loss": 0.2937,
+      "step": 2219
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0002813724489795918,
+      "loss": 0.227,
+      "step": 2220
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0002813571428571428,
+      "loss": 0.3035,
+      "step": 2221
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.00028134183673469385,
+      "loss": 0.184,
+      "step": 2222
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0002813265306122449,
+      "loss": 0.3798,
+      "step": 2223
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0002813112244897959,
+      "loss": 0.3368,
+      "step": 2224
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0002812959183673469,
+      "loss": 0.2401,
+      "step": 2225
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0002812806122448979,
+      "loss": 0.2699,
+      "step": 2226
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.00028126530612244895,
+      "loss": 0.2718,
+      "step": 2227
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.00028125,
+      "loss": 0.2324,
+      "step": 2228
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.000281234693877551,
+      "loss": 0.3446,
+      "step": 2229
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00028121938775510204,
+      "loss": 0.2903,
+      "step": 2230
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00028120408163265307,
+      "loss": 0.2269,
+      "step": 2231
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00028118877551020405,
+      "loss": 0.3391,
+      "step": 2232
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0002811734693877551,
+      "loss": 0.2336,
+      "step": 2233
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0002811581632653061,
+      "loss": 0.3101,
+      "step": 2234
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00028114285714285714,
+      "loss": 0.3203,
+      "step": 2235
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0002811275510204081,
+      "loss": 0.3428,
+      "step": 2236
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00028111224489795915,
+      "loss": 0.361,
+      "step": 2237
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0002810969387755102,
+      "loss": 0.2423,
+      "step": 2238
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0002810816326530612,
+      "loss": 0.1973,
+      "step": 2239
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00028106632653061224,
+      "loss": 0.2373,
+      "step": 2240
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0002810510204081632,
+      "loss": 0.3645,
+      "step": 2241
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0002810357142857143,
+      "loss": 0.3513,
+      "step": 2242
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0002810204081632653,
+      "loss": 0.3364,
+      "step": 2243
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0002810051020408163,
+      "loss": 0.278,
+      "step": 2244
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00028098979591836734,
+      "loss": 0.2323,
+      "step": 2245
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00028097448979591837,
+      "loss": 0.3501,
+      "step": 2246
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00028095918367346934,
+      "loss": 0.242,
+      "step": 2247
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00028094387755102037,
+      "loss": 0.2591,
+      "step": 2248
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0002809285714285714,
+      "loss": 0.2756,
+      "step": 2249
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00028091326530612243,
+      "loss": 0.3181,
+      "step": 2250
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00028089795918367346,
+      "loss": 0.2752,
+      "step": 2251
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00028088265306122444,
+      "loss": 0.1798,
+      "step": 2252
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00028086734693877547,
+      "loss": 0.1877,
+      "step": 2253
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.0002808520408163265,
+      "loss": 0.5239,
+      "step": 2254
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00028083673469387753,
+      "loss": 0.2255,
+      "step": 2255
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00028082142857142856,
+      "loss": 0.251,
+      "step": 2256
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00028080612244897954,
+      "loss": 0.2671,
+      "step": 2257
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.0002807908163265306,
+      "loss": 0.2989,
+      "step": 2258
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.0002807755102040816,
+      "loss": 0.3062,
+      "step": 2259
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00028076020408163263,
+      "loss": 0.2705,
+      "step": 2260
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.00028074489795918366,
+      "loss": 0.3999,
+      "step": 2261
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0002807295918367347,
+      "loss": 0.2998,
+      "step": 2262
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.00028071428571428567,
+      "loss": 0.2601,
+      "step": 2263
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0002806989795918367,
+      "loss": 0.2314,
+      "step": 2264
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.00028068367346938773,
+      "loss": 0.1871,
+      "step": 2265
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.00028066836734693876,
+      "loss": 0.2531,
+      "step": 2266
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0002806530612244898,
+      "loss": 0.3784,
+      "step": 2267
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.00028063775510204077,
+      "loss": 0.3719,
+      "step": 2268
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0002806224489795918,
+      "loss": 0.1862,
+      "step": 2269
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.00028060714285714283,
+      "loss": 0.2668,
+      "step": 2270
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.00028059183673469386,
+      "loss": 0.237,
+      "step": 2271
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0002805765306122449,
+      "loss": 0.3499,
+      "step": 2272
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.00028056122448979587,
+      "loss": 0.3146,
+      "step": 2273
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0002805459183673469,
+      "loss": 0.371,
+      "step": 2274
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0002805306122448979,
+      "loss": 0.2605,
+      "step": 2275
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.00028051530612244896,
+      "loss": 0.3747,
+      "step": 2276
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0002805,
+      "loss": 0.2581,
+      "step": 2277
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.000280484693877551,
+      "loss": 0.1882,
+      "step": 2278
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.000280469387755102,
+      "loss": 0.444,
+      "step": 2279
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.000280454081632653,
+      "loss": 0.3686,
+      "step": 2280
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.00028043877551020406,
+      "loss": 0.3017,
+      "step": 2281
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0002804234693877551,
+      "loss": 0.2303,
+      "step": 2282
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0002804081632653061,
+      "loss": 0.2129,
+      "step": 2283
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0002803928571428571,
+      "loss": 0.3515,
+      "step": 2284
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0002803775510204082,
+      "loss": 0.2485,
+      "step": 2285
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.00028036224489795915,
+      "loss": 0.3587,
+      "step": 2286
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0002803469387755102,
+      "loss": 0.2244,
+      "step": 2287
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0002803316326530612,
+      "loss": 0.2325,
+      "step": 2288
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.00028031632653061225,
+      "loss": 0.2495,
+      "step": 2289
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0002803010204081632,
+      "loss": 0.2218,
+      "step": 2290
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.00028028571428571425,
+      "loss": 0.219,
+      "step": 2291
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0002802704081632653,
+      "loss": 0.4858,
+      "step": 2292
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0002802551020408163,
+      "loss": 0.256,
+      "step": 2293
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00028023979591836734,
+      "loss": 0.3327,
+      "step": 2294
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0002802244897959183,
+      "loss": 0.3336,
+      "step": 2295
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00028020918367346935,
+      "loss": 0.2146,
+      "step": 2296
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0002801938775510204,
+      "loss": 0.2719,
+      "step": 2297
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0002801785714285714,
+      "loss": 0.3293,
+      "step": 2298
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00028016326530612244,
+      "loss": 0.3061,
+      "step": 2299
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0002801479591836734,
+      "loss": 0.2855,
+      "step": 2300
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0002801326530612245,
+      "loss": 0.27,
+      "step": 2301
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0002801173469387755,
+      "loss": 0.2376,
+      "step": 2302
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0002801020408163265,
+      "loss": 0.282,
+      "step": 2303
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0002800867346938775,
+      "loss": 0.2737,
+      "step": 2304
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.00028007142857142857,
+      "loss": 0.3019,
+      "step": 2305
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.00028005612244897955,
+      "loss": 0.3667,
+      "step": 2306
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0002800408163265306,
+      "loss": 0.2194,
+      "step": 2307
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0002800255102040816,
+      "loss": 0.2946,
+      "step": 2308
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.00028001020408163264,
+      "loss": 0.1802,
+      "step": 2309
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.00027999489795918367,
+      "loss": 0.4232,
+      "step": 2310
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.00027997959183673465,
+      "loss": 0.3912,
+      "step": 2311
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0002799642857142857,
+      "loss": 0.2648,
+      "step": 2312
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.0002799489795918367,
+      "loss": 0.2205,
+      "step": 2313
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00027993367346938774,
+      "loss": 0.2478,
+      "step": 2314
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00027991836734693877,
+      "loss": 0.3266,
+      "step": 2315
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00027990306122448975,
+      "loss": 0.2614,
+      "step": 2316
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.0002798877551020408,
+      "loss": 0.3709,
+      "step": 2317
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.0002798724489795918,
+      "loss": 0.2531,
+      "step": 2318
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00027985714285714284,
+      "loss": 0.2604,
+      "step": 2319
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00027984183673469387,
+      "loss": 0.2793,
+      "step": 2320
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.0002798265306122449,
+      "loss": 0.2075,
+      "step": 2321
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.0002798112244897959,
+      "loss": 0.3236,
+      "step": 2322
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0002797959183673469,
+      "loss": 0.3723,
+      "step": 2323
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00027978061224489794,
+      "loss": 0.2991,
+      "step": 2324
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00027976530612244897,
+      "loss": 0.2997,
+      "step": 2325
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00027975,
+      "loss": 0.3167,
+      "step": 2326
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00027973469387755097,
+      "loss": 0.2967,
+      "step": 2327
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00027971938775510206,
+      "loss": 0.2224,
+      "step": 2328
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00027970408163265303,
+      "loss": 0.4141,
+      "step": 2329
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00027968877551020406,
+      "loss": 0.3723,
+      "step": 2330
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00027967346938775504,
+      "loss": 0.2252,
+      "step": 2331
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0002796581632653061,
+      "loss": 0.2537,
+      "step": 2332
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0002796428571428571,
+      "loss": 0.2485,
+      "step": 2333
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00027962755102040813,
+      "loss": 0.1665,
+      "step": 2334
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00027961224489795916,
+      "loss": 0.2618,
+      "step": 2335
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0002795969387755102,
+      "loss": 0.3141,
+      "step": 2336
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0002795816326530612,
+      "loss": 0.264,
+      "step": 2337
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0002795663265306122,
+      "loss": 0.3138,
+      "step": 2338
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00027955102040816323,
+      "loss": 0.2292,
+      "step": 2339
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00027953571428571426,
+      "loss": 0.2559,
+      "step": 2340
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0002795204081632653,
+      "loss": 0.2715,
+      "step": 2341
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0002795051020408163,
+      "loss": 0.2646,
+      "step": 2342
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0002794897959183673,
+      "loss": 0.383,
+      "step": 2343
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00027947448979591833,
+      "loss": 0.2635,
+      "step": 2344
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00027945918367346936,
+      "loss": 0.2364,
+      "step": 2345
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.0002794438775510204,
+      "loss": 0.307,
+      "step": 2346
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00027942857142857137,
+      "loss": 0.2552,
+      "step": 2347
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00027941326530612245,
+      "loss": 0.465,
+      "step": 2348
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00027939795918367343,
+      "loss": 0.2638,
+      "step": 2349
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00027938265306122446,
+      "loss": 0.2627,
+      "step": 2350
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.0002793673469387755,
+      "loss": 0.2641,
+      "step": 2351
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.0002793520408163265,
+      "loss": 0.335,
+      "step": 2352
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00027933673469387755,
+      "loss": 0.2815,
+      "step": 2353
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0002793214285714285,
+      "loss": 0.4087,
+      "step": 2354
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.00027930612244897956,
+      "loss": 0.3297,
+      "step": 2355
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0002792908163265306,
+      "loss": 0.2358,
+      "step": 2356
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0002792755102040816,
+      "loss": 0.2739,
+      "step": 2357
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0002792602040816326,
+      "loss": 0.3117,
+      "step": 2358
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0002792448979591836,
+      "loss": 0.2978,
+      "step": 2359
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.00027922959183673466,
+      "loss": 0.3719,
+      "step": 2360
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0002792142857142857,
+      "loss": 0.3505,
+      "step": 2361
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0002791989795918367,
+      "loss": 0.3048,
+      "step": 2362
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0002791836734693877,
+      "loss": 0.174,
+      "step": 2363
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002791683673469388,
+      "loss": 0.2955,
+      "step": 2364
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.00027915306122448975,
+      "loss": 0.2077,
+      "step": 2365
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002791377551020408,
+      "loss": 0.3167,
+      "step": 2366
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002791224489795918,
+      "loss": 0.5187,
+      "step": 2367
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.00027910714285714285,
+      "loss": 0.2751,
+      "step": 2368
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002790918367346939,
+      "loss": 0.4296,
+      "step": 2369
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.00027907653061224485,
+      "loss": 0.3747,
+      "step": 2370
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002790612244897959,
+      "loss": 0.3132,
+      "step": 2371
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002790459183673469,
+      "loss": 0.2005,
+      "step": 2372
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.00027903061224489794,
+      "loss": 0.2206,
+      "step": 2373
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002790153061224489,
+      "loss": 0.2897,
+      "step": 2374
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.000279,
+      "loss": 0.2571,
+      "step": 2375
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.000278984693877551,
+      "loss": 0.3027,
+      "step": 2376
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.000278969387755102,
+      "loss": 0.206,
+      "step": 2377
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.00027895408163265304,
+      "loss": 0.2038,
+      "step": 2378
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0002789387755102041,
+      "loss": 0.3688,
+      "step": 2379
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0002789234693877551,
+      "loss": 0.3059,
+      "step": 2380
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0002789081632653061,
+      "loss": 0.2737,
+      "step": 2381
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0002788928571428571,
+      "loss": 0.3181,
+      "step": 2382
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.00027887755102040814,
+      "loss": 0.2892,
+      "step": 2383
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.00027886224489795917,
+      "loss": 0.2972,
+      "step": 2384
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.0002788469387755102,
+      "loss": 0.2763,
+      "step": 2385
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.0002788316326530612,
+      "loss": 0.3311,
+      "step": 2386
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.0002788163265306122,
+      "loss": 0.3034,
+      "step": 2387
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00027880102040816324,
+      "loss": 0.2754,
+      "step": 2388
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00027878571428571427,
+      "loss": 0.2628,
+      "step": 2389
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00027877040816326525,
+      "loss": 0.2275,
+      "step": 2390
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00027875510204081633,
+      "loss": 0.2598,
+      "step": 2391
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.0002787397959183673,
+      "loss": 0.4426,
+      "step": 2392
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00027872448979591834,
+      "loss": 0.342,
+      "step": 2393
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00027870918367346937,
+      "loss": 0.3208,
+      "step": 2394
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0002786938775510204,
+      "loss": 0.2416,
+      "step": 2395
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00027867857142857143,
+      "loss": 0.3248,
+      "step": 2396
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0002786632653061224,
+      "loss": 0.2589,
+      "step": 2397
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00027864795918367344,
+      "loss": 0.4687,
+      "step": 2398
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00027863265306122447,
+      "loss": 0.2551,
+      "step": 2399
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0002786173469387755,
+      "loss": 0.2596,
+      "step": 2400
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0002786020408163265,
+      "loss": 0.2439,
+      "step": 2401
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0002785867346938775,
+      "loss": 0.2351,
+      "step": 2402
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00027857142857142854,
+      "loss": 0.194,
+      "step": 2403
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00027855612244897957,
+      "loss": 0.3401,
+      "step": 2404
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0002785408163265306,
+      "loss": 0.2411,
+      "step": 2405
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00027852551020408157,
+      "loss": 0.3318,
+      "step": 2406
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00027851020408163266,
+      "loss": 0.2561,
+      "step": 2407
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00027849489795918363,
+      "loss": 0.1767,
+      "step": 2408
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00027847959183673466,
+      "loss": 0.2114,
+      "step": 2409
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0002784642857142857,
+      "loss": 0.2662,
+      "step": 2410
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0002784489795918367,
+      "loss": 0.4353,
+      "step": 2411
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00027843367346938776,
+      "loss": 0.3372,
+      "step": 2412
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00027841836734693873,
+      "loss": 0.2768,
+      "step": 2413
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00027840306122448976,
+      "loss": 0.2273,
+      "step": 2414
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0002783877551020408,
+      "loss": 0.3257,
+      "step": 2415
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.0002783724489795918,
+      "loss": 0.2579,
+      "step": 2416
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.0002783571428571428,
+      "loss": 0.3396,
+      "step": 2417
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.0002783418367346939,
+      "loss": 0.2742,
+      "step": 2418
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.00027832653061224486,
+      "loss": 0.2556,
+      "step": 2419
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.0002783112244897959,
+      "loss": 0.2146,
+      "step": 2420
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.0002782959183673469,
+      "loss": 0.3247,
+      "step": 2421
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.00027828061224489795,
+      "loss": 0.2744,
+      "step": 2422
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.000278265306122449,
+      "loss": 0.343,
+      "step": 2423
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.00027824999999999996,
+      "loss": 0.2353,
+      "step": 2424
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.000278234693877551,
+      "loss": 0.2998,
+      "step": 2425
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.000278219387755102,
+      "loss": 0.1981,
+      "step": 2426
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.00027820408163265305,
+      "loss": 0.203,
+      "step": 2427
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.00027818877551020403,
+      "loss": 0.2141,
+      "step": 2428
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.00027817346938775506,
+      "loss": 0.4036,
+      "step": 2429
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0002781581632653061,
+      "loss": 0.3286,
+      "step": 2430
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0002781428571428571,
+      "loss": 0.2609,
+      "step": 2431
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.00027812755102040815,
+      "loss": 0.2463,
+      "step": 2432
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0002781122448979591,
+      "loss": 0.3002,
+      "step": 2433
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0002780969387755102,
+      "loss": 0.223,
+      "step": 2434
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0002780816326530612,
+      "loss": 0.3475,
+      "step": 2435
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002780663265306122,
+      "loss": 0.3089,
+      "step": 2436
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.00027805102040816325,
+      "loss": 0.3054,
+      "step": 2437
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002780357142857143,
+      "loss": 0.3019,
+      "step": 2438
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002780204081632653,
+      "loss": 0.2684,
+      "step": 2439
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002780051020408163,
+      "loss": 0.3273,
+      "step": 2440
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002779897959183673,
+      "loss": 0.3845,
+      "step": 2441
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.00027797448979591835,
+      "loss": 0.4026,
+      "step": 2442
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002779591836734694,
+      "loss": 0.2952,
+      "step": 2443
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.00027794387755102035,
+      "loss": 0.2568,
+      "step": 2444
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002779285714285714,
+      "loss": 0.2634,
+      "step": 2445
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002779132653061224,
+      "loss": 0.2621,
+      "step": 2446
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.00027789795918367345,
+      "loss": 0.212,
+      "step": 2447
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0002778826530612245,
+      "loss": 0.3619,
+      "step": 2448
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.00027786734693877545,
+      "loss": 0.2881,
+      "step": 2449
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.00027785204081632654,
+      "loss": 0.3416,
+      "step": 2450
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0002778367346938775,
+      "loss": 0.2911,
+      "step": 2451
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.00027782142857142854,
+      "loss": 0.2759,
+      "step": 2452
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0002778061224489796,
+      "loss": 0.2464,
+      "step": 2453
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0002777908163265306,
+      "loss": 0.3289,
+      "step": 2454
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0002777755102040816,
+      "loss": 0.4159,
+      "step": 2455
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0002777602040816326,
+      "loss": 0.2302,
+      "step": 2456
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00027774489795918364,
+      "loss": 0.2395,
+      "step": 2457
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.0002777295918367347,
+      "loss": 0.2154,
+      "step": 2458
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.0002777142857142857,
+      "loss": 0.2024,
+      "step": 2459
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.0002776989795918367,
+      "loss": 0.4432,
+      "step": 2460
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00027768367346938776,
+      "loss": 0.2854,
+      "step": 2461
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00027766836734693874,
+      "loss": 0.2749,
+      "step": 2462
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00027765306122448977,
+      "loss": 0.3099,
+      "step": 2463
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.0002776377551020408,
+      "loss": 0.2224,
+      "step": 2464
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00027762244897959183,
+      "loss": 0.2378,
+      "step": 2465
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00027760714285714286,
+      "loss": 0.3284,
+      "step": 2466
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00027759183673469384,
+      "loss": 0.4741,
+      "step": 2467
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00027757653061224487,
+      "loss": 0.3134,
+      "step": 2468
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.0002775612244897959,
+      "loss": 0.2225,
+      "step": 2469
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00027754591836734693,
+      "loss": 0.2444,
+      "step": 2470
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.0002775306122448979,
+      "loss": 0.2345,
+      "step": 2471
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00027751530612244894,
+      "loss": 0.1776,
+      "step": 2472
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00027749999999999997,
+      "loss": 0.2986,
+      "step": 2473
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.000277484693877551,
+      "loss": 0.326,
+      "step": 2474
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00027746938775510203,
+      "loss": 0.2883,
+      "step": 2475
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.000277454081632653,
+      "loss": 0.2785,
+      "step": 2476
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.0002774387755102041,
+      "loss": 0.2246,
+      "step": 2477
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00027742346938775507,
+      "loss": 0.2042,
+      "step": 2478
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.0002774081632653061,
+      "loss": 0.4325,
+      "step": 2479
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00027739285714285713,
+      "loss": 0.2943,
+      "step": 2480
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00027737755102040816,
+      "loss": 0.2557,
+      "step": 2481
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.0002773622448979592,
+      "loss": 0.2175,
+      "step": 2482
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00027734693877551017,
+      "loss": 0.2124,
+      "step": 2483
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.0002773316326530612,
+      "loss": 0.3286,
+      "step": 2484
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.0002773163265306122,
+      "loss": 0.2584,
+      "step": 2485
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00027730102040816326,
+      "loss": 0.4328,
+      "step": 2486
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00027728571428571423,
+      "loss": 0.2197,
+      "step": 2487
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00027727040816326526,
+      "loss": 0.1889,
+      "step": 2488
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0002772551020408163,
+      "loss": 0.2637,
+      "step": 2489
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0002772397959183673,
+      "loss": 0.273,
+      "step": 2490
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00027722448979591836,
+      "loss": 0.2502,
+      "step": 2491
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00027720918367346933,
+      "loss": 0.2883,
+      "step": 2492
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0002771938775510204,
+      "loss": 0.3199,
+      "step": 2493
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0002771785714285714,
+      "loss": 0.1877,
+      "step": 2494
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0002771632653061224,
+      "loss": 0.285,
+      "step": 2495
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00027714795918367345,
+      "loss": 0.2226,
+      "step": 2496
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0002771326530612245,
+      "loss": 0.2375,
+      "step": 2497
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.00027711734693877546,
+      "loss": 0.324,
+      "step": 2498
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0002771020408163265,
+      "loss": 0.3729,
+      "step": 2499
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0002770867346938775,
+      "loss": 0.3143,
+      "step": 2500
+    },
+    {
+      "epoch": 2.43,
+      "eval_cer": 0.2627236434307557,
+      "eval_loss": 0.8568186163902283,
+      "eval_runtime": 225.8627,
+      "eval_samples_per_second": 39.024,
+      "eval_steps_per_second": 4.879,
+      "eval_wer": 0.6565929538938141,
+      "step": 2500
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.00027707142857142855,
+      "loss": 0.2677,
+      "step": 2501
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0002770561224489796,
+      "loss": 0.3665,
+      "step": 2502
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.00027704081632653056,
+      "loss": 0.4049,
+      "step": 2503
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.00027702551020408164,
+      "loss": 0.3673,
+      "step": 2504
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0002770102040816326,
+      "loss": 0.2952,
+      "step": 2505
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.00027699489795918365,
+      "loss": 0.2496,
+      "step": 2506
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0002769795918367347,
+      "loss": 0.1879,
+      "step": 2507
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0002769642857142857,
+      "loss": 0.226,
+      "step": 2508
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.00027694897959183674,
+      "loss": 0.2552,
+      "step": 2509
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0002769336734693877,
+      "loss": 0.3509,
+      "step": 2510
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.00027691836734693875,
+      "loss": 0.3546,
+      "step": 2511
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0002769030612244898,
+      "loss": 0.1751,
+      "step": 2512
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0002768877551020408,
+      "loss": 0.2471,
+      "step": 2513
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0002768724489795918,
+      "loss": 0.1893,
+      "step": 2514
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0002768571428571428,
+      "loss": 0.2334,
+      "step": 2515
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.00027684183673469385,
+      "loss": 0.2885,
+      "step": 2516
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0002768265306122449,
+      "loss": 0.4141,
+      "step": 2517
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0002768112244897959,
+      "loss": 0.276,
+      "step": 2518
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0002767959183673469,
+      "loss": 0.2629,
+      "step": 2519
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.00027678061224489797,
+      "loss": 0.302,
+      "step": 2520
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.00027676530612244895,
+      "loss": 0.2482,
+      "step": 2521
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.00027675,
+      "loss": 0.2822,
+      "step": 2522
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.000276734693877551,
+      "loss": 0.2962,
+      "step": 2523
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.00027671938775510204,
+      "loss": 0.3274,
+      "step": 2524
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.000276704081632653,
+      "loss": 0.2586,
+      "step": 2525
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.00027668877551020405,
+      "loss": 0.2392,
+      "step": 2526
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0002766734693877551,
+      "loss": 0.1762,
+      "step": 2527
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0002766581632653061,
+      "loss": 0.2725,
+      "step": 2528
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.00027664285714285714,
+      "loss": 0.3432,
+      "step": 2529
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002766275510204081,
+      "loss": 0.3554,
+      "step": 2530
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.00027661224489795914,
+      "loss": 0.3058,
+      "step": 2531
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002765969387755102,
+      "loss": 0.2802,
+      "step": 2532
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002765816326530612,
+      "loss": 0.3717,
+      "step": 2533
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.00027656632653061224,
+      "loss": 0.2332,
+      "step": 2534
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002765510204081632,
+      "loss": 0.2557,
+      "step": 2535
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002765357142857143,
+      "loss": 0.3836,
+      "step": 2536
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002765204081632653,
+      "loss": 0.3155,
+      "step": 2537
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002765051020408163,
+      "loss": 0.2025,
+      "step": 2538
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0002764897959183673,
+      "loss": 0.2855,
+      "step": 2539
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00027647448979591836,
+      "loss": 0.2126,
+      "step": 2540
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00027645918367346934,
+      "loss": 0.3135,
+      "step": 2541
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00027644387755102037,
+      "loss": 0.2367,
+      "step": 2542
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0002764285714285714,
+      "loss": 0.3551,
+      "step": 2543
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00027641326530612243,
+      "loss": 0.2622,
+      "step": 2544
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00027639795918367346,
+      "loss": 0.2011,
+      "step": 2545
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00027638265306122444,
+      "loss": 0.2335,
+      "step": 2546
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0002763673469387755,
+      "loss": 0.266,
+      "step": 2547
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0002763520408163265,
+      "loss": 0.4024,
+      "step": 2548
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00027633673469387753,
+      "loss": 0.2741,
+      "step": 2549
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.00027632142857142856,
+      "loss": 0.2224,
+      "step": 2550
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.0002763061224489796,
+      "loss": 0.21,
+      "step": 2551
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.00027629081632653057,
+      "loss": 0.267,
+      "step": 2552
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.0002762755102040816,
+      "loss": 0.3431,
+      "step": 2553
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.00027626020408163263,
+      "loss": 0.2971,
+      "step": 2554
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.00027624489795918366,
+      "loss": 0.2861,
+      "step": 2555
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.0002762295918367347,
+      "loss": 0.3051,
+      "step": 2556
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.00027621428571428567,
+      "loss": 0.2624,
+      "step": 2557
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.0002761989795918367,
+      "loss": 0.2709,
+      "step": 2558
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.00027618367346938773,
+      "loss": 0.2589,
+      "step": 2559
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.00027616836734693876,
+      "loss": 0.2991,
+      "step": 2560
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0002761530612244898,
+      "loss": 0.3618,
+      "step": 2561
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.00027613775510204077,
+      "loss": 0.3802,
+      "step": 2562
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.00027612244897959185,
+      "loss": 0.2244,
+      "step": 2563
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0002761071428571428,
+      "loss": 0.301,
+      "step": 2564
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.00027609183673469386,
+      "loss": 0.3049,
+      "step": 2565
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0002760765306122449,
+      "loss": 0.1928,
+      "step": 2566
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0002760612244897959,
+      "loss": 0.3174,
+      "step": 2567
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0002760459183673469,
+      "loss": 0.3465,
+      "step": 2568
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0002760306122448979,
+      "loss": 0.2959,
+      "step": 2569
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.00027601530612244896,
+      "loss": 0.267,
+      "step": 2570
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.000276,
+      "loss": 0.2056,
+      "step": 2571
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.000275984693877551,
+      "loss": 0.2465,
+      "step": 2572
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.000275969387755102,
+      "loss": 0.3312,
+      "step": 2573
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.000275954081632653,
+      "loss": 0.2967,
+      "step": 2574
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.00027593877551020405,
+      "loss": 0.2288,
+      "step": 2575
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0002759234693877551,
+      "loss": 0.2399,
+      "step": 2576
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0002759081632653061,
+      "loss": 0.2994,
+      "step": 2577
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0002758928571428571,
+      "loss": 0.1885,
+      "step": 2578
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0002758775510204081,
+      "loss": 0.314,
+      "step": 2579
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.00027586224489795915,
+      "loss": 0.3208,
+      "step": 2580
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0002758469387755102,
+      "loss": 0.2504,
+      "step": 2581
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00027583163265306116,
+      "loss": 0.2667,
+      "step": 2582
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00027581632653061224,
+      "loss": 0.2108,
+      "step": 2583
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0002758010204081632,
+      "loss": 0.1894,
+      "step": 2584
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00027578571428571425,
+      "loss": 0.3224,
+      "step": 2585
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0002757704081632653,
+      "loss": 0.4494,
+      "step": 2586
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0002757551020408163,
+      "loss": 0.3348,
+      "step": 2587
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00027573979591836734,
+      "loss": 0.2683,
+      "step": 2588
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0002757244897959183,
+      "loss": 0.2227,
+      "step": 2589
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00027570918367346935,
+      "loss": 0.2683,
+      "step": 2590
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0002756938775510204,
+      "loss": 0.2318,
+      "step": 2591
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0002756785714285714,
+      "loss": 0.3354,
+      "step": 2592
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00027566326530612244,
+      "loss": 0.3021,
+      "step": 2593
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00027564795918367347,
+      "loss": 0.3448,
+      "step": 2594
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00027563265306122445,
+      "loss": 0.2286,
+      "step": 2595
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0002756173469387755,
+      "loss": 0.2563,
+      "step": 2596
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0002756020408163265,
+      "loss": 0.2589,
+      "step": 2597
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00027558673469387754,
+      "loss": 0.2806,
+      "step": 2598
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00027557142857142857,
+      "loss": 0.2258,
+      "step": 2599
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00027555612244897955,
+      "loss": 0.2359,
+      "step": 2600
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0002755408163265306,
+      "loss": 0.2813,
+      "step": 2601
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0002755255102040816,
+      "loss": 0.2079,
+      "step": 2602
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00027551020408163264,
+      "loss": 0.3531,
+      "step": 2603
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00027549489795918367,
+      "loss": 0.3408,
+      "step": 2604
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00027547959183673465,
+      "loss": 0.3151,
+      "step": 2605
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00027546428571428573,
+      "loss": 0.2239,
+      "step": 2606
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0002754489795918367,
+      "loss": 0.2399,
+      "step": 2607
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00027543367346938774,
+      "loss": 0.264,
+      "step": 2608
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0002754183673469387,
+      "loss": 0.2584,
+      "step": 2609
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0002754030612244898,
+      "loss": 0.3461,
+      "step": 2610
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0002753877551020408,
+      "loss": 0.3525,
+      "step": 2611
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0002753724489795918,
+      "loss": 0.2988,
+      "step": 2612
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00027535714285714284,
+      "loss": 0.2042,
+      "step": 2613
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00027534183673469387,
+      "loss": 0.1768,
+      "step": 2614
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0002753265306122449,
+      "loss": 0.2544,
+      "step": 2615
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0002753112244897959,
+      "loss": 0.2128,
+      "step": 2616
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0002752959183673469,
+      "loss": 0.3831,
+      "step": 2617
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00027528061224489793,
+      "loss": 0.2606,
+      "step": 2618
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00027526530612244896,
+      "loss": 0.235,
+      "step": 2619
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00027525,
+      "loss": 0.3203,
+      "step": 2620
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00027523469387755097,
+      "loss": 0.3,
+      "step": 2621
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.000275219387755102,
+      "loss": 0.2233,
+      "step": 2622
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.00027520408163265303,
+      "loss": 0.272,
+      "step": 2623
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.00027518877551020406,
+      "loss": 0.4575,
+      "step": 2624
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.00027517346938775504,
+      "loss": 0.2976,
+      "step": 2625
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.0002751581632653061,
+      "loss": 0.2647,
+      "step": 2626
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.0002751428571428571,
+      "loss": 0.1926,
+      "step": 2627
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.00027512755102040813,
+      "loss": 0.2389,
+      "step": 2628
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.00027511224489795916,
+      "loss": 0.3516,
+      "step": 2629
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.0002750969387755102,
+      "loss": 0.299,
+      "step": 2630
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.0002750816326530612,
+      "loss": 0.2382,
+      "step": 2631
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0002750663265306122,
+      "loss": 0.2291,
+      "step": 2632
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00027505102040816323,
+      "loss": 0.2304,
+      "step": 2633
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00027503571428571426,
+      "loss": 0.1947,
+      "step": 2634
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0002750204081632653,
+      "loss": 0.3139,
+      "step": 2635
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00027500510204081627,
+      "loss": 0.3585,
+      "step": 2636
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00027498979591836735,
+      "loss": 0.2192,
+      "step": 2637
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00027497448979591833,
+      "loss": 0.2555,
+      "step": 2638
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00027495918367346936,
+      "loss": 0.279,
+      "step": 2639
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0002749438775510204,
+      "loss": 0.2115,
+      "step": 2640
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0002749285714285714,
+      "loss": 0.2128,
+      "step": 2641
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.00027491326530612245,
+      "loss": 0.3006,
+      "step": 2642
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0002748979591836734,
+      "loss": 0.2145,
+      "step": 2643
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.00027488265306122446,
+      "loss": 0.1985,
+      "step": 2644
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0002748673469387755,
+      "loss": 0.2499,
+      "step": 2645
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0002748520408163265,
+      "loss": 0.2891,
+      "step": 2646
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.00027483673469387755,
+      "loss": 0.2785,
+      "step": 2647
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0002748214285714285,
+      "loss": 0.3807,
+      "step": 2648
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.00027480612244897956,
+      "loss": 0.2999,
+      "step": 2649
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0002747908163265306,
+      "loss": 0.3015,
+      "step": 2650
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0002747755102040816,
+      "loss": 0.2281,
+      "step": 2651
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0002747602040816326,
+      "loss": 0.3064,
+      "step": 2652
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0002747448979591837,
+      "loss": 0.2624,
+      "step": 2653
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.00027472959183673465,
+      "loss": 0.4973,
+      "step": 2654
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0002747142857142857,
+      "loss": 0.3118,
+      "step": 2655
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0002746989795918367,
+      "loss": 0.3628,
+      "step": 2656
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.00027468367346938775,
+      "loss": 0.2835,
+      "step": 2657
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0002746683673469388,
+      "loss": 0.271,
+      "step": 2658
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.00027465306122448975,
+      "loss": 0.1479,
+      "step": 2659
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0002746377551020408,
+      "loss": 0.4031,
+      "step": 2660
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0002746224489795918,
+      "loss": 0.2562,
+      "step": 2661
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.00027460714285714284,
+      "loss": 0.2532,
+      "step": 2662
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0002745918367346939,
+      "loss": 0.1653,
+      "step": 2663
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.00027457653061224485,
+      "loss": 0.2611,
+      "step": 2664
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0002745612244897959,
+      "loss": 0.3098,
+      "step": 2665
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0002745459183673469,
+      "loss": 0.278,
+      "step": 2666
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.00027453061224489794,
+      "loss": 0.2831,
+      "step": 2667
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0002745153061224489,
+      "loss": 0.2892,
+      "step": 2668
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0002745,
+      "loss": 0.2752,
+      "step": 2669
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.000274484693877551,
+      "loss": 0.356,
+      "step": 2670
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.000274469387755102,
+      "loss": 0.2488,
+      "step": 2671
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.00027445408163265304,
+      "loss": 0.2278,
+      "step": 2672
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00027443877551020407,
+      "loss": 0.2095,
+      "step": 2673
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.0002744234693877551,
+      "loss": 0.3026,
+      "step": 2674
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.0002744081632653061,
+      "loss": 0.2274,
+      "step": 2675
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.0002743928571428571,
+      "loss": 0.2013,
+      "step": 2676
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00027437755102040814,
+      "loss": 0.3169,
+      "step": 2677
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00027436224489795917,
+      "loss": 0.3105,
+      "step": 2678
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00027434693877551015,
+      "loss": 0.2005,
+      "step": 2679
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.0002743316326530612,
+      "loss": 0.2506,
+      "step": 2680
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.0002743163265306122,
+      "loss": 0.2979,
+      "step": 2681
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00027430102040816324,
+      "loss": 0.2298,
+      "step": 2682
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00027428571428571427,
+      "loss": 0.27,
+      "step": 2683
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0002742704081632653,
+      "loss": 0.2612,
+      "step": 2684
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00027425510204081633,
+      "loss": 0.3723,
+      "step": 2685
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0002742397959183673,
+      "loss": 0.2397,
+      "step": 2686
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00027422448979591834,
+      "loss": 0.2884,
+      "step": 2687
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00027420918367346937,
+      "loss": 0.279,
+      "step": 2688
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0002741938775510204,
+      "loss": 0.3548,
+      "step": 2689
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00027417857142857143,
+      "loss": 0.2423,
+      "step": 2690
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0002741632653061224,
+      "loss": 0.2953,
+      "step": 2691
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00027414795918367344,
+      "loss": 0.3866,
+      "step": 2692
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00027413265306122447,
+      "loss": 0.2496,
+      "step": 2693
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.0002741173469387755,
+      "loss": 0.2424,
+      "step": 2694
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.0002741020408163265,
+      "loss": 0.3627,
+      "step": 2695
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00027408673469387756,
+      "loss": 0.1989,
+      "step": 2696
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00027407142857142853,
+      "loss": 0.2199,
+      "step": 2697
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00027405612244897956,
+      "loss": 0.403,
+      "step": 2698
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.0002740408163265306,
+      "loss": 0.312,
+      "step": 2699
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.0002740255102040816,
+      "loss": 0.3419,
+      "step": 2700
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00027401020408163266,
+      "loss": 0.3048,
+      "step": 2701
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00027399489795918363,
+      "loss": 0.2095,
+      "step": 2702
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00027397959183673466,
+      "loss": 0.2565,
+      "step": 2703
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002739642857142857,
+      "loss": 0.2394,
+      "step": 2704
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002739489795918367,
+      "loss": 0.3607,
+      "step": 2705
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002739336734693877,
+      "loss": 0.2187,
+      "step": 2706
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00027391836734693873,
+      "loss": 0.259,
+      "step": 2707
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00027390306122448976,
+      "loss": 0.3212,
+      "step": 2708
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002738877551020408,
+      "loss": 0.2381,
+      "step": 2709
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002738724489795918,
+      "loss": 0.4045,
+      "step": 2710
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002738571428571428,
+      "loss": 0.2705,
+      "step": 2711
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002738418367346939,
+      "loss": 0.285,
+      "step": 2712
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00027382653061224486,
+      "loss": 0.274,
+      "step": 2713
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002738112244897959,
+      "loss": 0.2335,
+      "step": 2714
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.0002737959183673469,
+      "loss": 0.2712,
+      "step": 2715
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00027378061224489795,
+      "loss": 0.2297,
+      "step": 2716
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.000273765306122449,
+      "loss": 0.4466,
+      "step": 2717
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00027374999999999996,
+      "loss": 0.226,
+      "step": 2718
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.000273734693877551,
+      "loss": 0.3058,
+      "step": 2719
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.000273719387755102,
+      "loss": 0.1864,
+      "step": 2720
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00027370408163265305,
+      "loss": 0.3453,
+      "step": 2721
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.000273688775510204,
+      "loss": 0.2716,
+      "step": 2722
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00027367346938775506,
+      "loss": 0.4372,
+      "step": 2723
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.0002736581632653061,
+      "loss": 0.254,
+      "step": 2724
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0002736428571428571,
+      "loss": 0.3475,
+      "step": 2725
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.00027362755102040815,
+      "loss": 0.2283,
+      "step": 2726
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0002736122448979591,
+      "loss": 0.2086,
+      "step": 2727
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0002735969387755102,
+      "loss": 0.2574,
+      "step": 2728
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0002735816326530612,
+      "loss": 0.3635,
+      "step": 2729
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0002735663265306122,
+      "loss": 0.3381,
+      "step": 2730
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.00027355102040816325,
+      "loss": 0.1949,
+      "step": 2731
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0002735357142857143,
+      "loss": 0.2286,
+      "step": 2732
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.00027352040816326525,
+      "loss": 0.2398,
+      "step": 2733
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0002735051020408163,
+      "loss": 0.2495,
+      "step": 2734
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0002734897959183673,
+      "loss": 0.2639,
+      "step": 2735
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00027347448979591835,
+      "loss": 0.3045,
+      "step": 2736
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0002734591836734694,
+      "loss": 0.2634,
+      "step": 2737
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00027344387755102035,
+      "loss": 0.2189,
+      "step": 2738
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00027342857142857144,
+      "loss": 0.2652,
+      "step": 2739
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0002734132653061224,
+      "loss": 0.3017,
+      "step": 2740
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00027339795918367344,
+      "loss": 0.1916,
+      "step": 2741
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0002733826530612245,
+      "loss": 0.4158,
+      "step": 2742
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0002733673469387755,
+      "loss": 0.2441,
+      "step": 2743
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00027335204081632654,
+      "loss": 0.2397,
+      "step": 2744
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0002733367346938775,
+      "loss": 0.3411,
+      "step": 2745
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00027332142857142854,
+      "loss": 0.2695,
+      "step": 2746
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0002733061224489796,
+      "loss": 0.285,
+      "step": 2747
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0002732908163265306,
+      "loss": 0.3783,
+      "step": 2748
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0002732755102040816,
+      "loss": 0.2644,
+      "step": 2749
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0002732602040816326,
+      "loss": 0.2539,
+      "step": 2750
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00027324489795918364,
+      "loss": 0.2274,
+      "step": 2751
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00027322959183673467,
+      "loss": 0.3215,
+      "step": 2752
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0002732142857142857,
+      "loss": 0.222,
+      "step": 2753
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0002731989795918367,
+      "loss": 0.358,
+      "step": 2754
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00027318367346938776,
+      "loss": 0.2993,
+      "step": 2755
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00027316836734693874,
+      "loss": 0.2215,
+      "step": 2756
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00027315306122448977,
+      "loss": 0.2945,
+      "step": 2757
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.0002731377551020408,
+      "loss": 0.2566,
+      "step": 2758
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00027312244897959183,
+      "loss": 0.2064,
+      "step": 2759
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00027310714285714286,
+      "loss": 0.3353,
+      "step": 2760
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00027309183673469384,
+      "loss": 0.3072,
+      "step": 2761
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00027307653061224487,
+      "loss": 0.2666,
+      "step": 2762
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.0002730612244897959,
+      "loss": 0.2143,
+      "step": 2763
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00027304591836734693,
+      "loss": 0.2198,
+      "step": 2764
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.0002730306122448979,
+      "loss": 0.2355,
+      "step": 2765
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00027301530612244894,
+      "loss": 0.3376,
+      "step": 2766
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00027299999999999997,
+      "loss": 0.3431,
+      "step": 2767
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.000272984693877551,
+      "loss": 0.2731,
+      "step": 2768
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00027296938775510203,
+      "loss": 0.248,
+      "step": 2769
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.000272954081632653,
+      "loss": 0.2666,
+      "step": 2770
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.0002729387755102041,
+      "loss": 0.2824,
+      "step": 2771
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00027292346938775507,
+      "loss": 0.2317,
+      "step": 2772
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.0002729081632653061,
+      "loss": 0.4028,
+      "step": 2773
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00027289285714285713,
+      "loss": 0.3108,
+      "step": 2774
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00027287755102040816,
+      "loss": 0.3017,
+      "step": 2775
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00027286224489795913,
+      "loss": 0.2261,
+      "step": 2776
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00027284693877551016,
+      "loss": 0.2332,
+      "step": 2777
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0002728316326530612,
+      "loss": 0.2589,
+      "step": 2778
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0002728163265306122,
+      "loss": 0.3654,
+      "step": 2779
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00027280102040816326,
+      "loss": 0.2969,
+      "step": 2780
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00027278571428571423,
+      "loss": 0.3923,
+      "step": 2781
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0002727704081632653,
+      "loss": 0.2309,
+      "step": 2782
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0002727551020408163,
+      "loss": 0.3168,
+      "step": 2783
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0002727397959183673,
+      "loss": 0.219,
+      "step": 2784
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00027272448979591836,
+      "loss": 0.264,
+      "step": 2785
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0002727091836734694,
+      "loss": 0.3253,
+      "step": 2786
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0002726938775510204,
+      "loss": 0.3041,
+      "step": 2787
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0002726785714285714,
+      "loss": 0.2268,
+      "step": 2788
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0002726632653061224,
+      "loss": 0.26,
+      "step": 2789
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00027264795918367345,
+      "loss": 0.2083,
+      "step": 2790
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0002726326530612245,
+      "loss": 0.3155,
+      "step": 2791
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00027261734693877546,
+      "loss": 0.3261,
+      "step": 2792
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0002726020408163265,
+      "loss": 0.2327,
+      "step": 2793
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0002725867346938775,
+      "loss": 0.2659,
+      "step": 2794
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00027257142857142855,
+      "loss": 0.2434,
+      "step": 2795
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0002725561224489796,
+      "loss": 0.2927,
+      "step": 2796
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00027254081632653056,
+      "loss": 0.2259,
+      "step": 2797
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00027252551020408164,
+      "loss": 0.3793,
+      "step": 2798
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0002725102040816326,
+      "loss": 0.3228,
+      "step": 2799
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00027249489795918365,
+      "loss": 0.2216,
+      "step": 2800
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0002724795918367347,
+      "loss": 0.3222,
+      "step": 2801
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0002724642857142857,
+      "loss": 0.2283,
+      "step": 2802
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0002724489795918367,
+      "loss": 0.3045,
+      "step": 2803
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0002724336734693877,
+      "loss": 0.2366,
+      "step": 2804
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00027241836734693875,
+      "loss": 0.2927,
+      "step": 2805
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0002724030612244898,
+      "loss": 0.2106,
+      "step": 2806
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0002723877551020408,
+      "loss": 0.1759,
+      "step": 2807
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0002723724489795918,
+      "loss": 0.2221,
+      "step": 2808
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0002723571428571428,
+      "loss": 0.2399,
+      "step": 2809
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00027234183673469385,
+      "loss": 0.4398,
+      "step": 2810
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0002723265306122449,
+      "loss": 0.4132,
+      "step": 2811
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0002723112244897959,
+      "loss": 0.2881,
+      "step": 2812
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0002722959183673469,
+      "loss": 0.1907,
+      "step": 2813
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00027228061224489797,
+      "loss": 0.2337,
+      "step": 2814
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00027226530612244895,
+      "loss": 0.1947,
+      "step": 2815
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00027225,
+      "loss": 0.2597,
+      "step": 2816
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00027223469387755095,
+      "loss": 0.46,
+      "step": 2817
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00027221938775510204,
+      "loss": 0.3026,
+      "step": 2818
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.000272204081632653,
+      "loss": 0.2226,
+      "step": 2819
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00027218877551020404,
+      "loss": 0.325,
+      "step": 2820
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0002721734693877551,
+      "loss": 0.2212,
+      "step": 2821
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0002721581632653061,
+      "loss": 0.2379,
+      "step": 2822
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00027214285714285714,
+      "loss": 0.2196,
+      "step": 2823
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0002721275510204081,
+      "loss": 0.2984,
+      "step": 2824
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0002721122448979592,
+      "loss": 0.2645,
+      "step": 2825
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0002720969387755102,
+      "loss": 0.2596,
+      "step": 2826
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0002720816326530612,
+      "loss": 0.2064,
+      "step": 2827
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00027206632653061223,
+      "loss": 0.4106,
+      "step": 2828
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00027205102040816327,
+      "loss": 0.2886,
+      "step": 2829
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00027203571428571424,
+      "loss": 0.3232,
+      "step": 2830
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00027202040816326527,
+      "loss": 0.3054,
+      "step": 2831
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.0002720051020408163,
+      "loss": 0.3122,
+      "step": 2832
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00027198979591836733,
+      "loss": 0.2521,
+      "step": 2833
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00027197448979591836,
+      "loss": 0.2307,
+      "step": 2834
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00027195918367346934,
+      "loss": 0.2409,
+      "step": 2835
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00027194387755102037,
+      "loss": 0.2881,
+      "step": 2836
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.0002719285714285714,
+      "loss": 0.2762,
+      "step": 2837
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.00027191326530612243,
+      "loss": 0.2731,
+      "step": 2838
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.00027189795918367346,
+      "loss": 0.2283,
+      "step": 2839
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.00027188265306122444,
+      "loss": 0.2029,
+      "step": 2840
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0002718673469387755,
+      "loss": 0.2896,
+      "step": 2841
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0002718520408163265,
+      "loss": 0.2143,
+      "step": 2842
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.00027183673469387753,
+      "loss": 0.3075,
+      "step": 2843
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.00027182142857142856,
+      "loss": 0.3962,
+      "step": 2844
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0002718061224489796,
+      "loss": 0.2211,
+      "step": 2845
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.00027179081632653057,
+      "loss": 0.2394,
+      "step": 2846
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0002717755102040816,
+      "loss": 0.2365,
+      "step": 2847
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00027176020408163263,
+      "loss": 0.398,
+      "step": 2848
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00027174489795918366,
+      "loss": 0.3604,
+      "step": 2849
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0002717295918367347,
+      "loss": 0.2657,
+      "step": 2850
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00027171428571428567,
+      "loss": 0.2315,
+      "step": 2851
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0002716989795918367,
+      "loss": 0.2613,
+      "step": 2852
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00027168367346938773,
+      "loss": 0.2827,
+      "step": 2853
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00027166836734693876,
+      "loss": 0.3051,
+      "step": 2854
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0002716530612244898,
+      "loss": 0.3186,
+      "step": 2855
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00027163775510204076,
+      "loss": 0.3971,
+      "step": 2856
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0002716224489795918,
+      "loss": 0.2117,
+      "step": 2857
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0002716071428571428,
+      "loss": 0.2142,
+      "step": 2858
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.00027159183673469386,
+      "loss": 0.25,
+      "step": 2859
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.00027157653061224483,
+      "loss": 0.2889,
+      "step": 2860
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0002715612244897959,
+      "loss": 0.3914,
+      "step": 2861
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0002715459183673469,
+      "loss": 0.25,
+      "step": 2862
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0002715306122448979,
+      "loss": 0.3681,
+      "step": 2863
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.00027151530612244896,
+      "loss": 0.2511,
+      "step": 2864
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0002715,
+      "loss": 0.2114,
+      "step": 2865
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.000271484693877551,
+      "loss": 0.2201,
+      "step": 2866
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.000271469387755102,
+      "loss": 0.2753,
+      "step": 2867
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0002714540816326531,
+      "loss": 0.2398,
+      "step": 2868
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00027143877551020405,
+      "loss": 0.2689,
+      "step": 2869
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0002714234693877551,
+      "loss": 0.2637,
+      "step": 2870
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0002714081632653061,
+      "loss": 0.1823,
+      "step": 2871
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00027139285714285715,
+      "loss": 0.2631,
+      "step": 2872
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0002713775510204081,
+      "loss": 0.2928,
+      "step": 2873
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00027136224489795915,
+      "loss": 0.232,
+      "step": 2874
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0002713469387755102,
+      "loss": 0.3113,
+      "step": 2875
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0002713316326530612,
+      "loss": 0.2809,
+      "step": 2876
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00027131632653061224,
+      "loss": 0.2073,
+      "step": 2877
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0002713010204081632,
+      "loss": 0.2519,
+      "step": 2878
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.00027128571428571425,
+      "loss": 0.3668,
+      "step": 2879
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0002712704081632653,
+      "loss": 0.3088,
+      "step": 2880
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0002712551020408163,
+      "loss": 0.2377,
+      "step": 2881
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.00027123979591836734,
+      "loss": 0.3648,
+      "step": 2882
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0002712244897959183,
+      "loss": 0.4043,
+      "step": 2883
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0002712091836734694,
+      "loss": 0.2588,
+      "step": 2884
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0002711938775510204,
+      "loss": 0.3861,
+      "step": 2885
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0002711785714285714,
+      "loss": 0.3108,
+      "step": 2886
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0002711632653061224,
+      "loss": 0.3811,
+      "step": 2887
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.00027114795918367347,
+      "loss": 0.3881,
+      "step": 2888
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.00027113265306122445,
+      "loss": 0.3065,
+      "step": 2889
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.0002711173469387755,
+      "loss": 0.2553,
+      "step": 2890
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.0002711020408163265,
+      "loss": 0.1951,
+      "step": 2891
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00027108673469387754,
+      "loss": 0.2865,
+      "step": 2892
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00027107142857142857,
+      "loss": 0.2944,
+      "step": 2893
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00027105612244897955,
+      "loss": 0.2484,
+      "step": 2894
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.0002710408163265306,
+      "loss": 0.3157,
+      "step": 2895
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.0002710255102040816,
+      "loss": 0.2458,
+      "step": 2896
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00027101020408163264,
+      "loss": 0.2615,
+      "step": 2897
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00027099489795918367,
+      "loss": 0.3837,
+      "step": 2898
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00027097959183673464,
+      "loss": 0.1897,
+      "step": 2899
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0002709642857142857,
+      "loss": 0.271,
+      "step": 2900
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0002709489795918367,
+      "loss": 0.2393,
+      "step": 2901
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.00027093367346938774,
+      "loss": 0.2661,
+      "step": 2902
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0002709183673469387,
+      "loss": 0.3043,
+      "step": 2903
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0002709030612244898,
+      "loss": 0.442,
+      "step": 2904
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0002708877551020408,
+      "loss": 0.3064,
+      "step": 2905
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0002708724489795918,
+      "loss": 0.2367,
+      "step": 2906
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.00027085714285714283,
+      "loss": 0.1935,
+      "step": 2907
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.00027084183673469387,
+      "loss": 0.2372,
+      "step": 2908
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0002708265306122449,
+      "loss": 0.1719,
+      "step": 2909
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00027081122448979587,
+      "loss": 0.3514,
+      "step": 2910
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00027079591836734696,
+      "loss": 0.3381,
+      "step": 2911
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00027078061224489793,
+      "loss": 0.3309,
+      "step": 2912
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00027076530612244896,
+      "loss": 0.2646,
+      "step": 2913
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00027074999999999994,
+      "loss": 0.2425,
+      "step": 2914
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.000270734693877551,
+      "loss": 0.2694,
+      "step": 2915
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.000270719387755102,
+      "loss": 0.2921,
+      "step": 2916
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00027070408163265303,
+      "loss": 0.4145,
+      "step": 2917
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00027068877551020406,
+      "loss": 0.257,
+      "step": 2918
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.0002706734693877551,
+      "loss": 0.333,
+      "step": 2919
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.0002706581632653061,
+      "loss": 0.2328,
+      "step": 2920
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0002706428571428571,
+      "loss": 0.2237,
+      "step": 2921
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00027062755102040813,
+      "loss": 0.1878,
+      "step": 2922
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00027061224489795916,
+      "loss": 0.2796,
+      "step": 2923
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0002705969387755102,
+      "loss": 0.2984,
+      "step": 2924
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0002705816326530612,
+      "loss": 0.3439,
+      "step": 2925
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0002705663265306122,
+      "loss": 0.3545,
+      "step": 2926
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00027055102040816323,
+      "loss": 0.2655,
+      "step": 2927
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00027053571428571426,
+      "loss": 0.2166,
+      "step": 2928
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0002705204081632653,
+      "loss": 0.2761,
+      "step": 2929
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00027050510204081627,
+      "loss": 0.2373,
+      "step": 2930
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.00027048979591836735,
+      "loss": 0.2647,
+      "step": 2931
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.00027047448979591833,
+      "loss": 0.2786,
+      "step": 2932
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.00027045918367346936,
+      "loss": 0.2254,
+      "step": 2933
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0002704438775510204,
+      "loss": 0.231,
+      "step": 2934
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0002704285714285714,
+      "loss": 0.2542,
+      "step": 2935
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.00027041326530612245,
+      "loss": 0.246,
+      "step": 2936
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0002703979591836734,
+      "loss": 0.2107,
+      "step": 2937
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.00027038265306122446,
+      "loss": 0.2775,
+      "step": 2938
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0002703673469387755,
+      "loss": 0.2449,
+      "step": 2939
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0002703520408163265,
+      "loss": 0.2289,
+      "step": 2940
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.00027033673469387755,
+      "loss": 0.2456,
+      "step": 2941
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002703214285714285,
+      "loss": 0.3688,
+      "step": 2942
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.00027030612244897956,
+      "loss": 0.3045,
+      "step": 2943
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002702908163265306,
+      "loss": 0.252,
+      "step": 2944
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002702755102040816,
+      "loss": 0.2037,
+      "step": 2945
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002702602040816326,
+      "loss": 0.2371,
+      "step": 2946
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002702448979591837,
+      "loss": 0.3216,
+      "step": 2947
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.00027022959183673465,
+      "loss": 0.2463,
+      "step": 2948
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002702142857142857,
+      "loss": 0.3132,
+      "step": 2949
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002701989795918367,
+      "loss": 0.2806,
+      "step": 2950
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00027018367346938775,
+      "loss": 0.2759,
+      "step": 2951
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0002701683673469388,
+      "loss": 0.2383,
+      "step": 2952
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00027015306122448975,
+      "loss": 0.2775,
+      "step": 2953
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0002701377551020408,
+      "loss": 0.2641,
+      "step": 2954
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0002701224489795918,
+      "loss": 0.3148,
+      "step": 2955
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00027010714285714284,
+      "loss": 0.304,
+      "step": 2956
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0002700918367346938,
+      "loss": 0.2842,
+      "step": 2957
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0002700765306122449,
+      "loss": 0.2679,
+      "step": 2958
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0002700612244897959,
+      "loss": 0.1691,
+      "step": 2959
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0002700459183673469,
+      "loss": 0.2918,
+      "step": 2960
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00027003061224489794,
+      "loss": 0.2642,
+      "step": 2961
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.000270015306122449,
+      "loss": 0.204,
+      "step": 2962
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00027,
+      "loss": 0.2226,
+      "step": 2963
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.000269984693877551,
+      "loss": 0.1977,
+      "step": 2964
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.000269969387755102,
+      "loss": 0.3079,
+      "step": 2965
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00026995408163265304,
+      "loss": 0.2276,
+      "step": 2966
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00026993877551020407,
+      "loss": 0.3256,
+      "step": 2967
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.0002699234693877551,
+      "loss": 0.3665,
+      "step": 2968
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.0002699081632653061,
+      "loss": 0.3265,
+      "step": 2969
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.0002698928571428571,
+      "loss": 0.316,
+      "step": 2970
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00026987755102040814,
+      "loss": 0.2755,
+      "step": 2971
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00026986224489795917,
+      "loss": 0.3691,
+      "step": 2972
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00026984693877551015,
+      "loss": 0.4106,
+      "step": 2973
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00026983163265306123,
+      "loss": 0.2717,
+      "step": 2974
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0002698163265306122,
+      "loss": 0.2657,
+      "step": 2975
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00026980102040816324,
+      "loss": 0.2042,
+      "step": 2976
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00026978571428571427,
+      "loss": 0.2873,
+      "step": 2977
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0002697704081632653,
+      "loss": 0.2001,
+      "step": 2978
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00026975510204081633,
+      "loss": 0.3963,
+      "step": 2979
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0002697397959183673,
+      "loss": 0.3443,
+      "step": 2980
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00026972448979591834,
+      "loss": 0.295,
+      "step": 2981
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00026970918367346937,
+      "loss": 0.253,
+      "step": 2982
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0002696938775510204,
+      "loss": 0.2513,
+      "step": 2983
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0002696785714285714,
+      "loss": 0.2741,
+      "step": 2984
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0002696632653061224,
+      "loss": 0.2062,
+      "step": 2985
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00026964795918367343,
+      "loss": 0.2872,
+      "step": 2986
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00026963265306122447,
+      "loss": 0.2188,
+      "step": 2987
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0002696173469387755,
+      "loss": 0.1899,
+      "step": 2988
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00026960204081632647,
+      "loss": 0.1833,
+      "step": 2989
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00026958673469387756,
+      "loss": 0.2323,
+      "step": 2990
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00026957142857142853,
+      "loss": 0.2305,
+      "step": 2991
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00026955612244897956,
+      "loss": 0.3247,
+      "step": 2992
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0002695408163265306,
+      "loss": 0.2266,
+      "step": 2993
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0002695255102040816,
+      "loss": 0.235,
+      "step": 2994
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.00026951020408163266,
+      "loss": 0.2352,
+      "step": 2995
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.00026949489795918363,
+      "loss": 0.228,
+      "step": 2996
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.00026947959183673466,
+      "loss": 0.212,
+      "step": 2997
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0002694642857142857,
+      "loss": 0.4138,
+      "step": 2998
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0002694489795918367,
+      "loss": 0.2887,
+      "step": 2999
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0002694336734693877,
+      "loss": 0.3004,
+      "step": 3000
+    },
+    {
+      "epoch": 2.91,
+      "eval_cer": 0.2686480130689488,
+      "eval_loss": 0.8898212313652039,
+      "eval_runtime": 227.3011,
+      "eval_samples_per_second": 38.777,
+      "eval_steps_per_second": 4.848,
+      "eval_wer": 0.6640191587296356,
+      "step": 3000
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0002694183673469388,
+      "loss": 0.3509,
+      "step": 3001
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.00026940306122448976,
+      "loss": 0.2237,
+      "step": 3002
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.0002693877551020408,
+      "loss": 0.3123,
+      "step": 3003
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.0002693724489795918,
+      "loss": 0.4434,
+      "step": 3004
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00026935714285714285,
+      "loss": 0.3015,
+      "step": 3005
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.0002693418367346939,
+      "loss": 0.2573,
+      "step": 3006
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00026932653061224486,
+      "loss": 0.2534,
+      "step": 3007
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.0002693112244897959,
+      "loss": 0.3183,
+      "step": 3008
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.0002692959183673469,
+      "loss": 0.2367,
+      "step": 3009
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00026928061224489795,
+      "loss": 0.3292,
+      "step": 3010
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00026926530612244893,
+      "loss": 0.3376,
+      "step": 3011
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00026924999999999996,
+      "loss": 0.4141,
+      "step": 3012
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.000269234693877551,
+      "loss": 0.3061,
+      "step": 3013
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.000269219387755102,
+      "loss": 0.2576,
+      "step": 3014
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.00026920408163265305,
+      "loss": 0.2084,
+      "step": 3015
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.000269188775510204,
+      "loss": 0.2778,
+      "step": 3016
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.0002691734693877551,
+      "loss": 0.4082,
+      "step": 3017
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.0002691581632653061,
+      "loss": 0.31,
+      "step": 3018
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.0002691428571428571,
+      "loss": 0.2534,
+      "step": 3019
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.00026912755102040815,
+      "loss": 0.3233,
+      "step": 3020
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.0002691122448979592,
+      "loss": 0.2282,
+      "step": 3021
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.0002690969387755102,
+      "loss": 0.2118,
+      "step": 3022
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.0002690816326530612,
+      "loss": 0.355,
+      "step": 3023
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0002690663265306122,
+      "loss": 0.4106,
+      "step": 3024
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00026905102040816325,
+      "loss": 0.287,
+      "step": 3025
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0002690357142857143,
+      "loss": 0.2821,
+      "step": 3026
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00026902040816326525,
+      "loss": 0.2375,
+      "step": 3027
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0002690051020408163,
+      "loss": 0.2424,
+      "step": 3028
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0002689897959183673,
+      "loss": 0.2533,
+      "step": 3029
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00026897448979591835,
+      "loss": 0.3005,
+      "step": 3030
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0002689591836734694,
+      "loss": 0.2513,
+      "step": 3031
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00026894387755102035,
+      "loss": 0.2257,
+      "step": 3032
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00026892857142857144,
+      "loss": 0.1802,
+      "step": 3033
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0002689132653061224,
+      "loss": 0.2289,
+      "step": 3034
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.00026889795918367344,
+      "loss": 0.369,
+      "step": 3035
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0002688826530612245,
+      "loss": 0.2506,
+      "step": 3036
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0002688673469387755,
+      "loss": 0.2915,
+      "step": 3037
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0002688520408163265,
+      "loss": 0.2771,
+      "step": 3038
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0002688367346938775,
+      "loss": 0.1931,
+      "step": 3039
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.00026882142857142854,
+      "loss": 0.3098,
+      "step": 3040
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0002688061224489796,
+      "loss": 0.3728,
+      "step": 3041
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0002687908163265306,
+      "loss": 0.3852,
+      "step": 3042
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0002687755102040816,
+      "loss": 0.2362,
+      "step": 3043
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.0002687602040816326,
+      "loss": 0.3105,
+      "step": 3044
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00026874489795918364,
+      "loss": 0.2911,
+      "step": 3045
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00026872959183673467,
+      "loss": 0.3016,
+      "step": 3046
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.0002687142857142857,
+      "loss": 0.3164,
+      "step": 3047
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00026869897959183673,
+      "loss": 0.456,
+      "step": 3048
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00026868367346938776,
+      "loss": 0.2958,
+      "step": 3049
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00026866836734693874,
+      "loss": 0.319,
+      "step": 3050
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00026865306122448977,
+      "loss": 0.3412,
+      "step": 3051
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.0002686377551020408,
+      "loss": 0.1854,
+      "step": 3052
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00026862244897959183,
+      "loss": 0.2104,
+      "step": 3053
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.0002686071428571428,
+      "loss": 0.213,
+      "step": 3054
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00026859183673469384,
+      "loss": 0.2934,
+      "step": 3055
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00026857653061224487,
+      "loss": 0.2742,
+      "step": 3056
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.0002685612244897959,
+      "loss": 0.3358,
+      "step": 3057
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00026854591836734693,
+      "loss": 0.2903,
+      "step": 3058
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.0002685306122448979,
+      "loss": 0.2327,
+      "step": 3059
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.000268515306122449,
+      "loss": 0.2771,
+      "step": 3060
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00026849999999999997,
+      "loss": 0.3533,
+      "step": 3061
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.000268484693877551,
+      "loss": 0.3253,
+      "step": 3062
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00026846938775510203,
+      "loss": 0.2537,
+      "step": 3063
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00026845408163265306,
+      "loss": 0.2942,
+      "step": 3064
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0002684387755102041,
+      "loss": 0.2217,
+      "step": 3065
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00026842346938775507,
+      "loss": 0.2172,
+      "step": 3066
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0002684081632653061,
+      "loss": 0.3086,
+      "step": 3067
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0002683928571428571,
+      "loss": 0.3294,
+      "step": 3068
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00026837755102040816,
+      "loss": 0.2557,
+      "step": 3069
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00026836224489795913,
+      "loss": 0.2129,
+      "step": 3070
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00026834693877551016,
+      "loss": 0.251,
+      "step": 3071
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0002683316326530612,
+      "loss": 0.2071,
+      "step": 3072
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0002683163265306122,
+      "loss": 0.4302,
+      "step": 3073
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00026830102040816326,
+      "loss": 0.2658,
+      "step": 3074
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.00026828571428571423,
+      "loss": 0.2255,
+      "step": 3075
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0002682704081632653,
+      "loss": 0.2342,
+      "step": 3076
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0002682551020408163,
+      "loss": 0.3285,
+      "step": 3077
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0002682397959183673,
+      "loss": 0.1978,
+      "step": 3078
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.00026822448979591835,
+      "loss": 0.3604,
+      "step": 3079
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0002682091836734694,
+      "loss": 0.1957,
+      "step": 3080
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.00026819387755102036,
+      "loss": 0.2897,
+      "step": 3081
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0002681785714285714,
+      "loss": 0.2407,
+      "step": 3082
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0002681632653061224,
+      "loss": 0.2997,
+      "step": 3083
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.00026814795918367345,
+      "loss": 0.2549,
+      "step": 3084
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0002681326530612245,
+      "loss": 0.2973,
+      "step": 3085
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00026811734693877546,
+      "loss": 0.2752,
+      "step": 3086
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0002681020408163265,
+      "loss": 0.3083,
+      "step": 3087
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0002680867346938775,
+      "loss": 0.1808,
+      "step": 3088
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00026807142857142855,
+      "loss": 0.2588,
+      "step": 3089
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0002680561224489796,
+      "loss": 0.178,
+      "step": 3090
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0002680408163265306,
+      "loss": 0.3877,
+      "step": 3091
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00026802551020408164,
+      "loss": 0.1601,
+      "step": 3092
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0002680102040816326,
+      "loss": 0.2034,
+      "step": 3093
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00026799489795918365,
+      "loss": 0.2667,
+      "step": 3094
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0002679795918367347,
+      "loss": 0.1718,
+      "step": 3095
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0002679642857142857,
+      "loss": 0.2242,
+      "step": 3096
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0002679489795918367,
+      "loss": 0.2671,
+      "step": 3097
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0002679336734693877,
+      "loss": 0.2172,
+      "step": 3098
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.00026791836734693875,
+      "loss": 0.154,
+      "step": 3099
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0002679030612244898,
+      "loss": 0.2578,
+      "step": 3100
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0002678877551020408,
+      "loss": 0.1515,
+      "step": 3101
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0002678724489795918,
+      "loss": 0.1797,
+      "step": 3102
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.00026785714285714287,
+      "loss": 0.2653,
+      "step": 3103
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.00026784183673469385,
+      "loss": 0.1654,
+      "step": 3104
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0002678265306122449,
+      "loss": 0.2088,
+      "step": 3105
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0002678112244897959,
+      "loss": 0.1443,
+      "step": 3106
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00026779591836734694,
+      "loss": 0.2486,
+      "step": 3107
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0002677806122448979,
+      "loss": 0.1956,
+      "step": 3108
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00026776530612244895,
+      "loss": 0.2669,
+      "step": 3109
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00026775,
+      "loss": 0.2171,
+      "step": 3110
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.000267734693877551,
+      "loss": 0.2769,
+      "step": 3111
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00026771938775510204,
+      "loss": 0.1657,
+      "step": 3112
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.000267704081632653,
+      "loss": 0.227,
+      "step": 3113
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00026768877551020404,
+      "loss": 0.1697,
+      "step": 3114
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0002676734693877551,
+      "loss": 0.1727,
+      "step": 3115
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002676581632653061,
+      "loss": 0.3093,
+      "step": 3116
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00026764285714285714,
+      "loss": 0.1788,
+      "step": 3117
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002676275510204081,
+      "loss": 0.2142,
+      "step": 3118
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002676122448979592,
+      "loss": 0.1719,
+      "step": 3119
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002675969387755102,
+      "loss": 0.1499,
+      "step": 3120
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002675816326530612,
+      "loss": 0.2576,
+      "step": 3121
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00026756632653061223,
+      "loss": 0.4757,
+      "step": 3122
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00026755102040816326,
+      "loss": 0.3219,
+      "step": 3123
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00026753571428571424,
+      "loss": 0.2388,
+      "step": 3124
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00026752040816326527,
+      "loss": 0.217,
+      "step": 3125
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002675051020408163,
+      "loss": 0.1552,
+      "step": 3126
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00026748979591836733,
+      "loss": 0.1645,
+      "step": 3127
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00026747448979591836,
+      "loss": 0.385,
+      "step": 3128
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00026745918367346934,
+      "loss": 0.2707,
+      "step": 3129
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00026744387755102037,
+      "loss": 0.1716,
+      "step": 3130
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.0002674285714285714,
+      "loss": 0.2267,
+      "step": 3131
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00026741326530612243,
+      "loss": 0.1993,
+      "step": 3132
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00026739795918367346,
+      "loss": 0.1801,
+      "step": 3133
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00026738265306122444,
+      "loss": 0.4217,
+      "step": 3134
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00026736734693877547,
+      "loss": 0.2153,
+      "step": 3135
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.0002673520408163265,
+      "loss": 0.1948,
+      "step": 3136
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00026733673469387753,
+      "loss": 0.1752,
+      "step": 3137
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00026732142857142856,
+      "loss": 0.2017,
+      "step": 3138
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0002673061224489796,
+      "loss": 0.1666,
+      "step": 3139
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00026729081632653057,
+      "loss": 0.3038,
+      "step": 3140
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0002672755102040816,
+      "loss": 0.2682,
+      "step": 3141
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00026726020408163263,
+      "loss": 0.3526,
+      "step": 3142
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00026724489795918366,
+      "loss": 0.2027,
+      "step": 3143
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0002672295918367347,
+      "loss": 0.1625,
+      "step": 3144
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00026721428571428567,
+      "loss": 0.1173,
+      "step": 3145
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00026719897959183675,
+      "loss": 0.2579,
+      "step": 3146
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002671836734693877,
+      "loss": 0.2412,
+      "step": 3147
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.00026716836734693876,
+      "loss": 0.1538,
+      "step": 3148
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002671530612244898,
+      "loss": 0.2574,
+      "step": 3149
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002671377551020408,
+      "loss": 0.216,
+      "step": 3150
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002671224489795918,
+      "loss": 0.2771,
+      "step": 3151
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002671071428571428,
+      "loss": 0.2545,
+      "step": 3152
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.00026709183673469386,
+      "loss": 0.44,
+      "step": 3153
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002670765306122449,
+      "loss": 0.3127,
+      "step": 3154
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002670612244897959,
+      "loss": 0.1548,
+      "step": 3155
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002670459183673469,
+      "loss": 0.2841,
+      "step": 3156
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0002670306122448979,
+      "loss": 0.2759,
+      "step": 3157
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00026701530612244895,
+      "loss": 0.1516,
+      "step": 3158
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.000267,
+      "loss": 0.3802,
+      "step": 3159
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.000266984693877551,
+      "loss": 0.3632,
+      "step": 3160
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.000266969387755102,
+      "loss": 0.2661,
+      "step": 3161
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0002669540816326531,
+      "loss": 0.1891,
+      "step": 3162
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00026693877551020405,
+      "loss": 0.1471,
+      "step": 3163
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0002669234693877551,
+      "loss": 0.1602,
+      "step": 3164
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00026690816326530606,
+      "loss": 0.2129,
+      "step": 3165
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00026689285714285714,
+      "loss": 0.3297,
+      "step": 3166
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0002668775510204081,
+      "loss": 0.3187,
+      "step": 3167
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.00026686224489795915,
+      "loss": 0.2215,
+      "step": 3168
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002668469387755102,
+      "loss": 0.2614,
+      "step": 3169
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002668316326530612,
+      "loss": 0.2018,
+      "step": 3170
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.00026681632653061224,
+      "loss": 0.1778,
+      "step": 3171
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002668010204081632,
+      "loss": 0.3856,
+      "step": 3172
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.00026678571428571425,
+      "loss": 0.1733,
+      "step": 3173
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002667704081632653,
+      "loss": 0.1521,
+      "step": 3174
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002667551020408163,
+      "loss": 0.2425,
+      "step": 3175
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.00026673979591836734,
+      "loss": 0.1842,
+      "step": 3176
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002667244897959183,
+      "loss": 0.2529,
+      "step": 3177
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00026670918367346935,
+      "loss": 0.2525,
+      "step": 3178
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002666938775510204,
+      "loss": 0.2803,
+      "step": 3179
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002666785714285714,
+      "loss": 0.256,
+      "step": 3180
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00026666326530612244,
+      "loss": 0.2691,
+      "step": 3181
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00026664795918367347,
+      "loss": 0.1968,
+      "step": 3182
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00026663265306122445,
+      "loss": 0.2538,
+      "step": 3183
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002666173469387755,
+      "loss": 0.2408,
+      "step": 3184
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002666020408163265,
+      "loss": 0.3157,
+      "step": 3185
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00026658673469387754,
+      "loss": 0.2173,
+      "step": 3186
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00026657142857142857,
+      "loss": 0.1963,
+      "step": 3187
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00026655612244897955,
+      "loss": 0.1986,
+      "step": 3188
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00026654081632653063,
+      "loss": 0.2002,
+      "step": 3189
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002665255102040816,
+      "loss": 0.2721,
+      "step": 3190
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00026651020408163264,
+      "loss": 0.3201,
+      "step": 3191
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002664948979591836,
+      "loss": 0.2481,
+      "step": 3192
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002664795918367347,
+      "loss": 0.303,
+      "step": 3193
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002664642857142857,
+      "loss": 0.1718,
+      "step": 3194
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002664489795918367,
+      "loss": 0.1214,
+      "step": 3195
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00026643367346938774,
+      "loss": 0.1598,
+      "step": 3196
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00026641836734693877,
+      "loss": 0.3881,
+      "step": 3197
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002664030612244898,
+      "loss": 0.2675,
+      "step": 3198
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002663877551020408,
+      "loss": 0.1801,
+      "step": 3199
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002663724489795918,
+      "loss": 0.1789,
+      "step": 3200
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00026635714285714283,
+      "loss": 0.1933,
+      "step": 3201
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00026634183673469386,
+      "loss": 0.1931,
+      "step": 3202
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002663265306122449,
+      "loss": 0.339,
+      "step": 3203
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00026631122448979587,
+      "loss": 0.2766,
+      "step": 3204
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002662959183673469,
+      "loss": 0.1639,
+      "step": 3205
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00026628061224489793,
+      "loss": 0.1824,
+      "step": 3206
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00026626530612244896,
+      "loss": 0.1284,
+      "step": 3207
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00026624999999999994,
+      "loss": 0.2272,
+      "step": 3208
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.000266234693877551,
+      "loss": 0.2321,
+      "step": 3209
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.000266219387755102,
+      "loss": 0.2163,
+      "step": 3210
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.00026620408163265303,
+      "loss": 0.2915,
+      "step": 3211
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.00026618877551020406,
+      "loss": 0.216,
+      "step": 3212
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002661734693877551,
+      "loss": 0.1528,
+      "step": 3213
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002661581632653061,
+      "loss": 0.2382,
+      "step": 3214
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002661428571428571,
+      "loss": 0.2084,
+      "step": 3215
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.00026612755102040813,
+      "loss": 0.2435,
+      "step": 3216
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.00026611224489795916,
+      "loss": 0.1784,
+      "step": 3217
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002660969387755102,
+      "loss": 0.3075,
+      "step": 3218
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002660816326530612,
+      "loss": 0.1967,
+      "step": 3219
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002660663265306122,
+      "loss": 0.2177,
+      "step": 3220
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.00026605102040816323,
+      "loss": 0.1732,
+      "step": 3221
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.00026603571428571426,
+      "loss": 0.3564,
+      "step": 3222
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002660204081632653,
+      "loss": 0.2625,
+      "step": 3223
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.00026600510204081627,
+      "loss": 0.2678,
+      "step": 3224
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.00026598979591836735,
+      "loss": 0.2614,
+      "step": 3225
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002659744897959183,
+      "loss": 0.1959,
+      "step": 3226
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.00026595918367346936,
+      "loss": 0.1525,
+      "step": 3227
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002659438775510204,
+      "loss": 0.3349,
+      "step": 3228
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002659285714285714,
+      "loss": 0.248,
+      "step": 3229
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00026591326530612245,
+      "loss": 0.1641,
+      "step": 3230
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002658979591836734,
+      "loss": 0.196,
+      "step": 3231
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00026588265306122446,
+      "loss": 0.1663,
+      "step": 3232
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002658673469387755,
+      "loss": 0.1581,
+      "step": 3233
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002658520408163265,
+      "loss": 0.2923,
+      "step": 3234
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002658367346938775,
+      "loss": 0.2691,
+      "step": 3235
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002658214285714286,
+      "loss": 0.2242,
+      "step": 3236
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00026580612244897955,
+      "loss": 0.1905,
+      "step": 3237
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002657908163265306,
+      "loss": 0.1959,
+      "step": 3238
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002657755102040816,
+      "loss": 0.2874,
+      "step": 3239
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00026576020408163265,
+      "loss": 0.349,
+      "step": 3240
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0002657448979591837,
+      "loss": 0.3363,
+      "step": 3241
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00026572959183673465,
+      "loss": 0.3012,
+      "step": 3242
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0002657142857142857,
+      "loss": 0.2761,
+      "step": 3243
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0002656989795918367,
+      "loss": 0.1926,
+      "step": 3244
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00026568367346938774,
+      "loss": 0.3064,
+      "step": 3245
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0002656683673469388,
+      "loss": 0.1867,
+      "step": 3246
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00026565306122448975,
+      "loss": 0.2857,
+      "step": 3247
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0002656377551020408,
+      "loss": 0.2518,
+      "step": 3248
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0002656224489795918,
+      "loss": 0.1954,
+      "step": 3249
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00026560714285714284,
+      "loss": 0.2278,
+      "step": 3250
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002655918367346938,
+      "loss": 0.3231,
+      "step": 3251
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002655765306122449,
+      "loss": 0.3233,
+      "step": 3252
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002655612244897959,
+      "loss": 0.3382,
+      "step": 3253
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002655459183673469,
+      "loss": 0.3814,
+      "step": 3254
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00026553061224489794,
+      "loss": 0.2057,
+      "step": 3255
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00026551530612244897,
+      "loss": 0.1505,
+      "step": 3256
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002655,
+      "loss": 0.1753,
+      "step": 3257
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.000265484693877551,
+      "loss": 0.1623,
+      "step": 3258
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.000265469387755102,
+      "loss": 0.3493,
+      "step": 3259
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00026545408163265304,
+      "loss": 0.3959,
+      "step": 3260
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00026543877551020407,
+      "loss": 0.2005,
+      "step": 3261
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00026542346938775505,
+      "loss": 0.2015,
+      "step": 3262
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002654081632653061,
+      "loss": 0.2645,
+      "step": 3263
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002653928571428571,
+      "loss": 0.1563,
+      "step": 3264
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00026537755102040814,
+      "loss": 0.2698,
+      "step": 3265
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00026536224489795917,
+      "loss": 0.2107,
+      "step": 3266
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00026534693877551015,
+      "loss": 0.1546,
+      "step": 3267
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00026533163265306123,
+      "loss": 0.1785,
+      "step": 3268
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002653163265306122,
+      "loss": 0.1612,
+      "step": 3269
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00026530102040816324,
+      "loss": 0.1711,
+      "step": 3270
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00026528571428571427,
+      "loss": 0.1736,
+      "step": 3271
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0002652704081632653,
+      "loss": 0.2847,
+      "step": 3272
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00026525510204081633,
+      "loss": 0.2659,
+      "step": 3273
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0002652397959183673,
+      "loss": 0.1631,
+      "step": 3274
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00026522448979591834,
+      "loss": 0.2168,
+      "step": 3275
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00026520918367346937,
+      "loss": 0.1967,
+      "step": 3276
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0002651938775510204,
+      "loss": 0.2195,
+      "step": 3277
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0002651785714285714,
+      "loss": 0.2734,
+      "step": 3278
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00026516326530612246,
+      "loss": 0.2028,
+      "step": 3279
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00026514795918367343,
+      "loss": 0.2013,
+      "step": 3280
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00026513265306122446,
+      "loss": 0.2646,
+      "step": 3281
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002651173469387755,
+      "loss": 0.1388,
+      "step": 3282
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002651020408163265,
+      "loss": 0.2464,
+      "step": 3283
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00026508673469387756,
+      "loss": 0.2027,
+      "step": 3284
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00026507142857142853,
+      "loss": 0.3516,
+      "step": 3285
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00026505612244897956,
+      "loss": 0.1754,
+      "step": 3286
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002650408163265306,
+      "loss": 0.2038,
+      "step": 3287
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002650255102040816,
+      "loss": 0.1886,
+      "step": 3288
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002650102040816326,
+      "loss": 0.1927,
+      "step": 3289
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00026499489795918363,
+      "loss": 0.2415,
+      "step": 3290
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00026497959183673466,
+      "loss": 0.3244,
+      "step": 3291
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002649642857142857,
+      "loss": 0.2328,
+      "step": 3292
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002649489795918367,
+      "loss": 0.2285,
+      "step": 3293
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002649336734693877,
+      "loss": 0.2089,
+      "step": 3294
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002649183673469388,
+      "loss": 0.1394,
+      "step": 3295
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00026490306122448976,
+      "loss": 0.2648,
+      "step": 3296
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002648877551020408,
+      "loss": 0.3248,
+      "step": 3297
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002648724489795918,
+      "loss": 0.2358,
+      "step": 3298
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00026485714285714285,
+      "loss": 0.1639,
+      "step": 3299
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002648418367346939,
+      "loss": 0.158,
+      "step": 3300
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00026482653061224486,
+      "loss": 0.1254,
+      "step": 3301
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0002648112244897959,
+      "loss": 0.1297,
+      "step": 3302
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0002647959183673469,
+      "loss": 0.1852,
+      "step": 3303
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00026478061224489795,
+      "loss": 0.2318,
+      "step": 3304
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0002647653061224489,
+      "loss": 0.1769,
+      "step": 3305
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00026474999999999996,
+      "loss": 0.2089,
+      "step": 3306
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.000264734693877551,
+      "loss": 0.1728,
+      "step": 3307
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.000264719387755102,
+      "loss": 0.2509,
+      "step": 3308
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00026470408163265305,
+      "loss": 0.2237,
+      "step": 3309
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.000264688775510204,
+      "loss": 0.3109,
+      "step": 3310
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0002646734693877551,
+      "loss": 0.2123,
+      "step": 3311
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002646581632653061,
+      "loss": 0.1912,
+      "step": 3312
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002646428571428571,
+      "loss": 0.1941,
+      "step": 3313
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00026462755102040815,
+      "loss": 0.1451,
+      "step": 3314
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002646122448979592,
+      "loss": 0.2744,
+      "step": 3315
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00026459693877551015,
+      "loss": 0.3455,
+      "step": 3316
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002645816326530612,
+      "loss": 0.2401,
+      "step": 3317
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002645663265306122,
+      "loss": 0.1588,
+      "step": 3318
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00026455102040816325,
+      "loss": 0.166,
+      "step": 3319
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002645357142857143,
+      "loss": 0.1739,
+      "step": 3320
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00026452040816326525,
+      "loss": 0.1935,
+      "step": 3321
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00026450510204081634,
+      "loss": 0.3454,
+      "step": 3322
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002644897959183673,
+      "loss": 0.2582,
+      "step": 3323
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00026447448979591834,
+      "loss": 0.1817,
+      "step": 3324
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002644591836734694,
+      "loss": 0.2149,
+      "step": 3325
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002644438775510204,
+      "loss": 0.2253,
+      "step": 3326
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00026442857142857144,
+      "loss": 0.251,
+      "step": 3327
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002644132653061224,
+      "loss": 0.2052,
+      "step": 3328
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00026439795918367344,
+      "loss": 0.2707,
+      "step": 3329
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002643826530612245,
+      "loss": 0.2641,
+      "step": 3330
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002643673469387755,
+      "loss": 0.158,
+      "step": 3331
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002643520408163265,
+      "loss": 0.1914,
+      "step": 3332
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0002643367346938775,
+      "loss": 0.2176,
+      "step": 3333
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00026432142857142854,
+      "loss": 0.2214,
+      "step": 3334
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00026430612244897957,
+      "loss": 0.2449,
+      "step": 3335
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0002642908163265306,
+      "loss": 0.22,
+      "step": 3336
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0002642755102040816,
+      "loss": 0.1828,
+      "step": 3337
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00026426020408163266,
+      "loss": 0.2373,
+      "step": 3338
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00026424489795918364,
+      "loss": 0.127,
+      "step": 3339
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00026422959183673467,
+      "loss": 0.3287,
+      "step": 3340
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0002642142857142857,
+      "loss": 0.2931,
+      "step": 3341
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00026419897959183673,
+      "loss": 0.1954,
+      "step": 3342
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00026418367346938776,
+      "loss": 0.1546,
+      "step": 3343
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00026416836734693874,
+      "loss": 0.2779,
+      "step": 3344
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00026415306122448977,
+      "loss": 0.1447,
+      "step": 3345
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.0002641377551020408,
+      "loss": 0.2645,
+      "step": 3346
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00026412244897959183,
+      "loss": 0.3598,
+      "step": 3347
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.0002641071428571428,
+      "loss": 0.1496,
+      "step": 3348
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00026409183673469384,
+      "loss": 0.1611,
+      "step": 3349
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00026407653061224487,
+      "loss": 0.1967,
+      "step": 3350
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.0002640612244897959,
+      "loss": 0.231,
+      "step": 3351
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00026404591836734693,
+      "loss": 0.2351,
+      "step": 3352
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0002640306122448979,
+      "loss": 0.3733,
+      "step": 3353
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.000264015306122449,
+      "loss": 0.2575,
+      "step": 3354
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00026399999999999997,
+      "loss": 0.1737,
+      "step": 3355
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.000263984693877551,
+      "loss": 0.3189,
+      "step": 3356
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00026396938775510203,
+      "loss": 0.2518,
+      "step": 3357
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00026395408163265306,
+      "loss": 0.1367,
+      "step": 3358
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00026393877551020403,
+      "loss": 0.357,
+      "step": 3359
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00026392346938775506,
+      "loss": 0.2932,
+      "step": 3360
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0002639081632653061,
+      "loss": 0.2225,
+      "step": 3361
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0002638928571428571,
+      "loss": 0.1891,
+      "step": 3362
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.00026387755102040816,
+      "loss": 0.2683,
+      "step": 3363
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.00026386224489795913,
+      "loss": 0.1872,
+      "step": 3364
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002638469387755102,
+      "loss": 0.2088,
+      "step": 3365
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002638316326530612,
+      "loss": 0.2752,
+      "step": 3366
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002638163265306122,
+      "loss": 0.2276,
+      "step": 3367
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.00026380102040816325,
+      "loss": 0.1541,
+      "step": 3368
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002637857142857143,
+      "loss": 0.2056,
+      "step": 3369
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002637704081632653,
+      "loss": 0.2358,
+      "step": 3370
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002637551020408163,
+      "loss": 0.1859,
+      "step": 3371
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002637397959183673,
+      "loss": 0.2089,
+      "step": 3372
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.00026372448979591835,
+      "loss": 0.1791,
+      "step": 3373
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002637091836734694,
+      "loss": 0.1792,
+      "step": 3374
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00026369387755102036,
+      "loss": 0.2271,
+      "step": 3375
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002636785714285714,
+      "loss": 0.2473,
+      "step": 3376
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002636632653061224,
+      "loss": 0.1576,
+      "step": 3377
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00026364795918367345,
+      "loss": 0.3387,
+      "step": 3378
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002636326530612245,
+      "loss": 0.2371,
+      "step": 3379
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00026361734693877546,
+      "loss": 0.2845,
+      "step": 3380
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00026360204081632654,
+      "loss": 0.1382,
+      "step": 3381
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002635867346938775,
+      "loss": 0.218,
+      "step": 3382
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00026357142857142855,
+      "loss": 0.2316,
+      "step": 3383
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002635561224489796,
+      "loss": 0.2039,
+      "step": 3384
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002635408163265306,
+      "loss": 0.229,
+      "step": 3385
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002635255102040816,
+      "loss": 0.1765,
+      "step": 3386
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002635102040816326,
+      "loss": 0.1582,
+      "step": 3387
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00026349489795918365,
+      "loss": 0.2128,
+      "step": 3388
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002634795918367347,
+      "loss": 0.184,
+      "step": 3389
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002634642857142857,
+      "loss": 0.2006,
+      "step": 3390
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002634489795918367,
+      "loss": 0.2563,
+      "step": 3391
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002634336734693877,
+      "loss": 0.1782,
+      "step": 3392
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00026341836734693875,
+      "loss": 0.2008,
+      "step": 3393
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002634030612244898,
+      "loss": 0.2311,
+      "step": 3394
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002633877551020408,
+      "loss": 0.1706,
+      "step": 3395
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002633724489795918,
+      "loss": 0.1958,
+      "step": 3396
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00026335714285714287,
+      "loss": 0.3429,
+      "step": 3397
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00026334183673469385,
+      "loss": 0.2714,
+      "step": 3398
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002633265306122449,
+      "loss": 0.2257,
+      "step": 3399
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002633112244897959,
+      "loss": 0.1612,
+      "step": 3400
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00026329591836734694,
+      "loss": 0.1624,
+      "step": 3401
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002632806122448979,
+      "loss": 0.2474,
+      "step": 3402
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00026326530612244894,
+      "loss": 0.1763,
+      "step": 3403
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00026325,
+      "loss": 0.1927,
+      "step": 3404
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.000263234693877551,
+      "loss": 0.2006,
+      "step": 3405
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00026321938775510204,
+      "loss": 0.2722,
+      "step": 3406
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.000263204081632653,
+      "loss": 0.2875,
+      "step": 3407
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0002631887755102041,
+      "loss": 0.2576,
+      "step": 3408
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0002631734693877551,
+      "loss": 0.3198,
+      "step": 3409
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0002631581632653061,
+      "loss": 0.3624,
+      "step": 3410
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00026314285714285713,
+      "loss": 0.2096,
+      "step": 3411
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00026312755102040817,
+      "loss": 0.2374,
+      "step": 3412
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00026311224489795914,
+      "loss": 0.2066,
+      "step": 3413
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00026309693877551017,
+      "loss": 0.1298,
+      "step": 3414
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0002630816326530612,
+      "loss": 0.2434,
+      "step": 3415
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00026306632653061223,
+      "loss": 0.1969,
+      "step": 3416
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00026305102040816326,
+      "loss": 0.2867,
+      "step": 3417
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00026303571428571424,
+      "loss": 0.1976,
+      "step": 3418
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00026302040816326527,
+      "loss": 0.1607,
+      "step": 3419
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0002630051020408163,
+      "loss": 0.1246,
+      "step": 3420
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00026298979591836733,
+      "loss": 0.1565,
+      "step": 3421
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00026297448979591836,
+      "loss": 0.3095,
+      "step": 3422
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00026295918367346934,
+      "loss": 0.2052,
+      "step": 3423
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0002629438775510204,
+      "loss": 0.2202,
+      "step": 3424
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0002629285714285714,
+      "loss": 0.182,
+      "step": 3425
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00026291326530612243,
+      "loss": 0.2155,
+      "step": 3426
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00026289795918367346,
+      "loss": 0.1639,
+      "step": 3427
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0002628826530612245,
+      "loss": 0.4101,
+      "step": 3428
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00026286734693877547,
+      "loss": 0.1845,
+      "step": 3429
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0002628520408163265,
+      "loss": 0.1969,
+      "step": 3430
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00026283673469387753,
+      "loss": 0.2004,
+      "step": 3431
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00026282142857142856,
+      "loss": 0.1266,
+      "step": 3432
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0002628061224489796,
+      "loss": 0.1975,
+      "step": 3433
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00026279081632653057,
+      "loss": 0.3693,
+      "step": 3434
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0002627755102040816,
+      "loss": 0.2502,
+      "step": 3435
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00026276020408163263,
+      "loss": 0.2341,
+      "step": 3436
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00026274489795918366,
+      "loss": 0.249,
+      "step": 3437
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0002627295918367347,
+      "loss": 0.3005,
+      "step": 3438
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00026271428571428566,
+      "loss": 0.2508,
+      "step": 3439
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00026269897959183675,
+      "loss": 0.2558,
+      "step": 3440
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0002626836734693877,
+      "loss": 0.2398,
+      "step": 3441
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00026266836734693876,
+      "loss": 0.2207,
+      "step": 3442
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00026265306122448973,
+      "loss": 0.1294,
+      "step": 3443
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0002626377551020408,
+      "loss": 0.1967,
+      "step": 3444
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0002626224489795918,
+      "loss": 0.1557,
+      "step": 3445
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002626071428571428,
+      "loss": 0.2335,
+      "step": 3446
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.00026259183673469385,
+      "loss": 0.361,
+      "step": 3447
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002625765306122449,
+      "loss": 0.1922,
+      "step": 3448
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002625612244897959,
+      "loss": 0.2021,
+      "step": 3449
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002625459183673469,
+      "loss": 0.228,
+      "step": 3450
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002625306122448979,
+      "loss": 0.2448,
+      "step": 3451
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.00026251530612244895,
+      "loss": 0.1683,
+      "step": 3452
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002625,
+      "loss": 0.2726,
+      "step": 3453
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.000262484693877551,
+      "loss": 0.16,
+      "step": 3454
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.00026246938775510204,
+      "loss": 0.2485,
+      "step": 3455
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.000262454081632653,
+      "loss": 0.2041,
+      "step": 3456
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00026243877551020405,
+      "loss": 0.3204,
+      "step": 3457
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002624234693877551,
+      "loss": 0.1368,
+      "step": 3458
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002624081632653061,
+      "loss": 0.2722,
+      "step": 3459
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00026239285714285714,
+      "loss": 0.211,
+      "step": 3460
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002623775510204081,
+      "loss": 0.1697,
+      "step": 3461
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00026236224489795915,
+      "loss": 0.2398,
+      "step": 3462
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002623469387755102,
+      "loss": 0.155,
+      "step": 3463
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002623316326530612,
+      "loss": 0.2124,
+      "step": 3464
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00026231632653061224,
+      "loss": 0.2636,
+      "step": 3465
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002623010204081632,
+      "loss": 0.2327,
+      "step": 3466
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002622857142857143,
+      "loss": 0.2839,
+      "step": 3467
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002622704081632653,
+      "loss": 0.2066,
+      "step": 3468
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002622551020408163,
+      "loss": 0.2964,
+      "step": 3469
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002622397959183673,
+      "loss": 0.1821,
+      "step": 3470
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00026222448979591837,
+      "loss": 0.355,
+      "step": 3471
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00026220918367346935,
+      "loss": 0.2873,
+      "step": 3472
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002621938775510204,
+      "loss": 0.2659,
+      "step": 3473
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002621785714285714,
+      "loss": 0.2468,
+      "step": 3474
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00026216326530612244,
+      "loss": 0.1882,
+      "step": 3475
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00026214795918367347,
+      "loss": 0.1603,
+      "step": 3476
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00026213265306122445,
+      "loss": 0.2767,
+      "step": 3477
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002621173469387755,
+      "loss": 0.2996,
+      "step": 3478
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002621020408163265,
+      "loss": 0.2938,
+      "step": 3479
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00026208673469387754,
+      "loss": 0.1707,
+      "step": 3480
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00026207142857142857,
+      "loss": 0.1673,
+      "step": 3481
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00026205612244897954,
+      "loss": 0.2716,
+      "step": 3482
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002620408163265306,
+      "loss": 0.1181,
+      "step": 3483
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002620255102040816,
+      "loss": 0.2279,
+      "step": 3484
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00026201020408163264,
+      "loss": 0.2143,
+      "step": 3485
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002619948979591836,
+      "loss": 0.2526,
+      "step": 3486
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002619795918367347,
+      "loss": 0.2,
+      "step": 3487
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002619642857142857,
+      "loss": 0.2069,
+      "step": 3488
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002619489795918367,
+      "loss": 0.1897,
+      "step": 3489
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00026193367346938773,
+      "loss": 0.1506,
+      "step": 3490
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00026191836734693877,
+      "loss": 0.2973,
+      "step": 3491
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002619030612244898,
+      "loss": 0.183,
+      "step": 3492
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00026188775510204077,
+      "loss": 0.1619,
+      "step": 3493
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002618724489795918,
+      "loss": 0.1562,
+      "step": 3494
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00026185714285714283,
+      "loss": 0.1573,
+      "step": 3495
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00026184183673469386,
+      "loss": 0.1916,
+      "step": 3496
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00026182653061224484,
+      "loss": 0.3634,
+      "step": 3497
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00026181122448979587,
+      "loss": 0.204,
+      "step": 3498
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.0002617959183673469,
+      "loss": 0.1514,
+      "step": 3499
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00026178061224489793,
+      "loss": 0.282,
+      "step": 3500
+    },
+    {
+      "epoch": 3.4,
+      "eval_cer": 0.257091706317674,
+      "eval_loss": 0.8488596081733704,
+      "eval_runtime": 260.057,
+      "eval_samples_per_second": 33.893,
+      "eval_steps_per_second": 4.238,
+      "eval_wer": 0.6636511441408782,
+      "step": 3500
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00026176530612244896,
+      "loss": 0.2449,
+      "step": 3501
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00026175,
+      "loss": 0.2206,
+      "step": 3502
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.000261734693877551,
+      "loss": 0.2196,
+      "step": 3503
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.000261719387755102,
+      "loss": 0.2735,
+      "step": 3504
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00026170408163265303,
+      "loss": 0.3659,
+      "step": 3505
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00026168877551020406,
+      "loss": 0.2408,
+      "step": 3506
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.0002616734693877551,
+      "loss": 0.2362,
+      "step": 3507
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0002616581632653061,
+      "loss": 0.2849,
+      "step": 3508
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0002616428571428571,
+      "loss": 0.295,
+      "step": 3509
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00026162755102040813,
+      "loss": 0.3517,
+      "step": 3510
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00026161224489795916,
+      "loss": 0.2804,
+      "step": 3511
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0002615969387755102,
+      "loss": 0.3181,
+      "step": 3512
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00026158163265306117,
+      "loss": 0.2665,
+      "step": 3513
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00026156632653061225,
+      "loss": 0.1647,
+      "step": 3514
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00026155102040816323,
+      "loss": 0.1829,
+      "step": 3515
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00026153571428571426,
+      "loss": 0.2883,
+      "step": 3516
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0002615204081632653,
+      "loss": 0.2585,
+      "step": 3517
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002615051020408163,
+      "loss": 0.2287,
+      "step": 3518
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00026148979591836735,
+      "loss": 0.1821,
+      "step": 3519
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002614744897959183,
+      "loss": 0.2672,
+      "step": 3520
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00026145918367346936,
+      "loss": 0.3237,
+      "step": 3521
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002614438775510204,
+      "loss": 0.2651,
+      "step": 3522
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002614285714285714,
+      "loss": 0.2967,
+      "step": 3523
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00026141326530612245,
+      "loss": 0.188,
+      "step": 3524
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002613979591836734,
+      "loss": 0.1947,
+      "step": 3525
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00026138265306122445,
+      "loss": 0.1827,
+      "step": 3526
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002613673469387755,
+      "loss": 0.1683,
+      "step": 3527
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002613520408163265,
+      "loss": 0.3336,
+      "step": 3528
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002613367346938775,
+      "loss": 0.1921,
+      "step": 3529
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002613214285714286,
+      "loss": 0.2171,
+      "step": 3530
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.00026130612244897955,
+      "loss": 0.1847,
+      "step": 3531
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002612908163265306,
+      "loss": 0.167,
+      "step": 3532
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002612755102040816,
+      "loss": 0.229,
+      "step": 3533
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.00026126020408163264,
+      "loss": 0.2447,
+      "step": 3534
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002612448979591837,
+      "loss": 0.3484,
+      "step": 3535
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.00026122959183673465,
+      "loss": 0.176,
+      "step": 3536
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002612142857142857,
+      "loss": 0.1705,
+      "step": 3537
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002611989795918367,
+      "loss": 0.1555,
+      "step": 3538
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.00026118367346938774,
+      "loss": 0.2133,
+      "step": 3539
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002611683673469387,
+      "loss": 0.1248,
+      "step": 3540
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.00026115306122448975,
+      "loss": 0.2047,
+      "step": 3541
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002611377551020408,
+      "loss": 0.2466,
+      "step": 3542
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002611224489795918,
+      "loss": 0.2525,
+      "step": 3543
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.00026110714285714284,
+      "loss": 0.1541,
+      "step": 3544
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.00026109183673469387,
+      "loss": 0.1191,
+      "step": 3545
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002610765306122449,
+      "loss": 0.2363,
+      "step": 3546
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002610612244897959,
+      "loss": 0.2745,
+      "step": 3547
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002610459183673469,
+      "loss": 0.2569,
+      "step": 3548
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.00026103061224489794,
+      "loss": 0.2547,
+      "step": 3549
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.00026101530612244897,
+      "loss": 0.1734,
+      "step": 3550
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.000261,
+      "loss": 0.21,
+      "step": 3551
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.000260984693877551,
+      "loss": 0.3352,
+      "step": 3552
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.000260969387755102,
+      "loss": 0.1907,
+      "step": 3553
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.00026095408163265304,
+      "loss": 0.1755,
+      "step": 3554
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.00026093877551020407,
+      "loss": 0.2331,
+      "step": 3555
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.00026092346938775505,
+      "loss": 0.1889,
+      "step": 3556
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.00026090816326530613,
+      "loss": 0.1612,
+      "step": 3557
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0002608928571428571,
+      "loss": 0.1864,
+      "step": 3558
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00026087755102040814,
+      "loss": 0.3168,
+      "step": 3559
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00026086224489795917,
+      "loss": 0.2385,
+      "step": 3560
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0002608469387755102,
+      "loss": 0.1585,
+      "step": 3561
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00026083163265306123,
+      "loss": 0.199,
+      "step": 3562
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0002608163265306122,
+      "loss": 0.1874,
+      "step": 3563
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00026080102040816324,
+      "loss": 0.1741,
+      "step": 3564
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00026078571428571427,
+      "loss": 0.3714,
+      "step": 3565
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0002607704081632653,
+      "loss": 0.2461,
+      "step": 3566
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0002607551020408163,
+      "loss": 0.3134,
+      "step": 3567
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0002607397959183673,
+      "loss": 0.2197,
+      "step": 3568
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00026072448979591833,
+      "loss": 0.1927,
+      "step": 3569
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00026070918367346937,
+      "loss": 0.2242,
+      "step": 3570
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.0002606938775510204,
+      "loss": 0.1546,
+      "step": 3571
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00026067857142857137,
+      "loss": 0.2836,
+      "step": 3572
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00026066326530612246,
+      "loss": 0.2063,
+      "step": 3573
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00026064795918367343,
+      "loss": 0.1519,
+      "step": 3574
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00026063265306122446,
+      "loss": 0.1349,
+      "step": 3575
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.0002606173469387755,
+      "loss": 0.1393,
+      "step": 3576
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.0002606020408163265,
+      "loss": 0.1545,
+      "step": 3577
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00026058673469387756,
+      "loss": 0.3314,
+      "step": 3578
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00026057142857142853,
+      "loss": 0.3192,
+      "step": 3579
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.00026055612244897956,
+      "loss": 0.209,
+      "step": 3580
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002605408163265306,
+      "loss": 0.1825,
+      "step": 3581
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002605255102040816,
+      "loss": 0.1998,
+      "step": 3582
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002605102040816326,
+      "loss": 0.1676,
+      "step": 3583
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.00026049489795918363,
+      "loss": 0.2639,
+      "step": 3584
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.00026047959183673466,
+      "loss": 0.2297,
+      "step": 3585
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002604642857142857,
+      "loss": 0.1531,
+      "step": 3586
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002604489795918367,
+      "loss": 0.2431,
+      "step": 3587
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002604336734693877,
+      "loss": 0.1747,
+      "step": 3588
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002604183673469388,
+      "loss": 0.1343,
+      "step": 3589
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00026040306122448976,
+      "loss": 0.1991,
+      "step": 3590
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002603877551020408,
+      "loss": 0.311,
+      "step": 3591
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002603724489795918,
+      "loss": 0.1551,
+      "step": 3592
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00026035714285714285,
+      "loss": 0.2534,
+      "step": 3593
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00026034183673469383,
+      "loss": 0.1392,
+      "step": 3594
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00026032653061224486,
+      "loss": 0.1778,
+      "step": 3595
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002603112244897959,
+      "loss": 0.1884,
+      "step": 3596
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002602959183673469,
+      "loss": 0.292,
+      "step": 3597
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00026028061224489795,
+      "loss": 0.261,
+      "step": 3598
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002602653061224489,
+      "loss": 0.2839,
+      "step": 3599
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00026025,
+      "loss": 0.178,
+      "step": 3600
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.000260234693877551,
+      "loss": 0.1806,
+      "step": 3601
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.000260219387755102,
+      "loss": 0.1529,
+      "step": 3602
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00026020408163265305,
+      "loss": 0.3198,
+      "step": 3603
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0002601887755102041,
+      "loss": 0.2697,
+      "step": 3604
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0002601734693877551,
+      "loss": 0.1627,
+      "step": 3605
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0002601581632653061,
+      "loss": 0.2074,
+      "step": 3606
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0002601428571428571,
+      "loss": 0.1948,
+      "step": 3607
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00026012755102040815,
+      "loss": 0.2003,
+      "step": 3608
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0002601122448979592,
+      "loss": 0.237,
+      "step": 3609
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00026009693877551015,
+      "loss": 0.1687,
+      "step": 3610
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002600816326530612,
+      "loss": 0.3856,
+      "step": 3611
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002600663265306122,
+      "loss": 0.1707,
+      "step": 3612
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00026005102040816324,
+      "loss": 0.1763,
+      "step": 3613
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002600357142857143,
+      "loss": 0.1225,
+      "step": 3614
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00026002040816326525,
+      "loss": 0.1812,
+      "step": 3615
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00026000510204081634,
+      "loss": 0.2521,
+      "step": 3616
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002599897959183673,
+      "loss": 0.2067,
+      "step": 3617
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00025997448979591834,
+      "loss": 0.2126,
+      "step": 3618
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002599591836734694,
+      "loss": 0.1314,
+      "step": 3619
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002599438775510204,
+      "loss": 0.2206,
+      "step": 3620
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00025992857142857144,
+      "loss": 0.1611,
+      "step": 3621
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002599132653061224,
+      "loss": 0.3689,
+      "step": 3622
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00025989795918367344,
+      "loss": 0.365,
+      "step": 3623
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00025988265306122447,
+      "loss": 0.1978,
+      "step": 3624
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002598673469387755,
+      "loss": 0.2039,
+      "step": 3625
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002598520408163265,
+      "loss": 0.2526,
+      "step": 3626
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002598367346938775,
+      "loss": 0.2249,
+      "step": 3627
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00025982142857142854,
+      "loss": 0.3692,
+      "step": 3628
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00025980612244897957,
+      "loss": 0.2243,
+      "step": 3629
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002597908163265306,
+      "loss": 0.2396,
+      "step": 3630
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.0002597755102040816,
+      "loss": 0.2396,
+      "step": 3631
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00025976020408163266,
+      "loss": 0.2404,
+      "step": 3632
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00025974489795918364,
+      "loss": 0.218,
+      "step": 3633
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00025972959183673467,
+      "loss": 0.2867,
+      "step": 3634
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.0002597142857142857,
+      "loss": 0.2702,
+      "step": 3635
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00025969897959183673,
+      "loss": 0.2077,
+      "step": 3636
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.0002596836734693877,
+      "loss": 0.2557,
+      "step": 3637
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00025966836734693874,
+      "loss": 0.2596,
+      "step": 3638
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00025965306122448977,
+      "loss": 0.1403,
+      "step": 3639
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.0002596377551020408,
+      "loss": 0.2399,
+      "step": 3640
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00025962244897959183,
+      "loss": 0.2695,
+      "step": 3641
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0002596071428571428,
+      "loss": 0.264,
+      "step": 3642
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0002595918367346939,
+      "loss": 0.2779,
+      "step": 3643
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.00025957653061224487,
+      "loss": 0.1716,
+      "step": 3644
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0002595612244897959,
+      "loss": 0.1858,
+      "step": 3645
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.00025954591836734693,
+      "loss": 0.3326,
+      "step": 3646
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.00025953061224489796,
+      "loss": 0.295,
+      "step": 3647
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.000259515306122449,
+      "loss": 0.2316,
+      "step": 3648
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.00025949999999999997,
+      "loss": 0.3052,
+      "step": 3649
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.000259484693877551,
+      "loss": 0.1603,
+      "step": 3650
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.000259469387755102,
+      "loss": 0.2101,
+      "step": 3651
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.00025945408163265306,
+      "loss": 0.2873,
+      "step": 3652
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.00025943877551020403,
+      "loss": 0.3956,
+      "step": 3653
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.00025942346938775506,
+      "loss": 0.2746,
+      "step": 3654
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002594081632653061,
+      "loss": 0.1834,
+      "step": 3655
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002593928571428571,
+      "loss": 0.1774,
+      "step": 3656
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.00025937755102040816,
+      "loss": 0.1882,
+      "step": 3657
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.00025936224489795913,
+      "loss": 0.2667,
+      "step": 3658
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002593469387755102,
+      "loss": 0.1854,
+      "step": 3659
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002593316326530612,
+      "loss": 0.2708,
+      "step": 3660
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002593163265306122,
+      "loss": 0.2048,
+      "step": 3661
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00025930102040816325,
+      "loss": 0.2904,
+      "step": 3662
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002592857142857143,
+      "loss": 0.2134,
+      "step": 3663
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00025927040816326526,
+      "loss": 0.1671,
+      "step": 3664
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002592551020408163,
+      "loss": 0.32,
+      "step": 3665
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002592397959183673,
+      "loss": 0.2724,
+      "step": 3666
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00025922448979591835,
+      "loss": 0.3034,
+      "step": 3667
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002592091836734694,
+      "loss": 0.1709,
+      "step": 3668
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00025919387755102036,
+      "loss": 0.2189,
+      "step": 3669
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002591785714285714,
+      "loss": 0.2101,
+      "step": 3670
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002591632653061224,
+      "loss": 0.2057,
+      "step": 3671
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.00025914795918367345,
+      "loss": 0.2863,
+      "step": 3672
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002591326530612245,
+      "loss": 0.1996,
+      "step": 3673
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.00025911734693877546,
+      "loss": 0.2111,
+      "step": 3674
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.00025910204081632654,
+      "loss": 0.2295,
+      "step": 3675
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002590867346938775,
+      "loss": 0.203,
+      "step": 3676
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.00025907142857142855,
+      "loss": 0.2238,
+      "step": 3677
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002590561224489796,
+      "loss": 0.3247,
+      "step": 3678
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002590408163265306,
+      "loss": 0.225,
+      "step": 3679
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002590255102040816,
+      "loss": 0.2728,
+      "step": 3680
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002590102040816326,
+      "loss": 0.3107,
+      "step": 3681
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.00025899489795918365,
+      "loss": 0.1748,
+      "step": 3682
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002589795918367347,
+      "loss": 0.2159,
+      "step": 3683
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002589642857142857,
+      "loss": 0.3476,
+      "step": 3684
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002589489795918367,
+      "loss": 0.2904,
+      "step": 3685
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00025893367346938777,
+      "loss": 0.1937,
+      "step": 3686
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00025891836734693875,
+      "loss": 0.2559,
+      "step": 3687
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002589030612244898,
+      "loss": 0.2245,
+      "step": 3688
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002588877551020408,
+      "loss": 0.3072,
+      "step": 3689
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00025887244897959184,
+      "loss": 0.1969,
+      "step": 3690
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002588571428571428,
+      "loss": 0.2508,
+      "step": 3691
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00025884183673469384,
+      "loss": 0.247,
+      "step": 3692
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002588265306122449,
+      "loss": 0.2623,
+      "step": 3693
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002588112244897959,
+      "loss": 0.1771,
+      "step": 3694
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00025879591836734694,
+      "loss": 0.1704,
+      "step": 3695
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002587806122448979,
+      "loss": 0.2785,
+      "step": 3696
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00025876530612244894,
+      "loss": 0.3881,
+      "step": 3697
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00025875,
+      "loss": 0.2584,
+      "step": 3698
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.000258734693877551,
+      "loss": 0.3274,
+      "step": 3699
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00025871938775510204,
+      "loss": 0.1413,
+      "step": 3700
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.000258704081632653,
+      "loss": 0.26,
+      "step": 3701
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002586887755102041,
+      "loss": 0.1688,
+      "step": 3702
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025867346938775507,
+      "loss": 0.3668,
+      "step": 3703
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0002586581632653061,
+      "loss": 0.2591,
+      "step": 3704
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025864285714285713,
+      "loss": 0.2492,
+      "step": 3705
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025862755102040816,
+      "loss": 0.2898,
+      "step": 3706
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025861224489795914,
+      "loss": 0.2448,
+      "step": 3707
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025859693877551017,
+      "loss": 0.2928,
+      "step": 3708
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0002585816326530612,
+      "loss": 0.5054,
+      "step": 3709
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025856632653061223,
+      "loss": 0.2401,
+      "step": 3710
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025855102040816326,
+      "loss": 0.173,
+      "step": 3711
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025853571428571424,
+      "loss": 0.2424,
+      "step": 3712
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00025852040816326527,
+      "loss": 0.1725,
+      "step": 3713
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.0002585051020408163,
+      "loss": 0.1167,
+      "step": 3714
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00025848979591836733,
+      "loss": 0.1745,
+      "step": 3715
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00025847448979591836,
+      "loss": 0.221,
+      "step": 3716
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00025845918367346934,
+      "loss": 0.2859,
+      "step": 3717
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.0002584438775510204,
+      "loss": 0.1574,
+      "step": 3718
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.0002584285714285714,
+      "loss": 0.221,
+      "step": 3719
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00025841326530612243,
+      "loss": 0.2141,
+      "step": 3720
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.0002583979591836734,
+      "loss": 0.2559,
+      "step": 3721
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.0002583826530612245,
+      "loss": 0.3166,
+      "step": 3722
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00025836734693877547,
+      "loss": 0.2922,
+      "step": 3723
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002583520408163265,
+      "loss": 0.2777,
+      "step": 3724
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00025833673469387753,
+      "loss": 0.2837,
+      "step": 3725
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00025832142857142856,
+      "loss": 0.1516,
+      "step": 3726
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002583061224489796,
+      "loss": 0.2367,
+      "step": 3727
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00025829081632653057,
+      "loss": 0.3108,
+      "step": 3728
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00025827551020408165,
+      "loss": 0.3003,
+      "step": 3729
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002582602040816326,
+      "loss": 0.2987,
+      "step": 3730
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00025824489795918366,
+      "loss": 0.2581,
+      "step": 3731
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002582295918367347,
+      "loss": 0.1884,
+      "step": 3732
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002582142857142857,
+      "loss": 0.2209,
+      "step": 3733
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002581989795918367,
+      "loss": 0.3631,
+      "step": 3734
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002581836734693877,
+      "loss": 0.2748,
+      "step": 3735
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00025816836734693876,
+      "loss": 0.2586,
+      "step": 3736
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002581530612244898,
+      "loss": 0.1949,
+      "step": 3737
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002581377551020408,
+      "loss": 0.1548,
+      "step": 3738
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002581224489795918,
+      "loss": 0.2233,
+      "step": 3739
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002581071428571428,
+      "loss": 0.3095,
+      "step": 3740
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00025809183673469385,
+      "loss": 0.2572,
+      "step": 3741
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002580765306122449,
+      "loss": 0.2922,
+      "step": 3742
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002580612244897959,
+      "loss": 0.1755,
+      "step": 3743
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002580459183673469,
+      "loss": 0.1524,
+      "step": 3744
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.000258030612244898,
+      "loss": 0.1553,
+      "step": 3745
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.00025801530612244895,
+      "loss": 0.2283,
+      "step": 3746
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.000258,
+      "loss": 0.2922,
+      "step": 3747
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.00025798469387755096,
+      "loss": 0.3131,
+      "step": 3748
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.00025796938775510204,
+      "loss": 0.256,
+      "step": 3749
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.000257954081632653,
+      "loss": 0.1911,
+      "step": 3750
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.00025793877551020405,
+      "loss": 0.1555,
+      "step": 3751
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0002579234693877551,
+      "loss": 0.1663,
+      "step": 3752
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0002579081632653061,
+      "loss": 0.3563,
+      "step": 3753
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.00025789285714285714,
+      "loss": 0.2617,
+      "step": 3754
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002578775510204081,
+      "loss": 0.1644,
+      "step": 3755
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.00025786224489795915,
+      "loss": 0.2438,
+      "step": 3756
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002578469387755102,
+      "loss": 0.2733,
+      "step": 3757
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002578316326530612,
+      "loss": 0.2343,
+      "step": 3758
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.00025781632653061224,
+      "loss": 0.2555,
+      "step": 3759
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002578010204081632,
+      "loss": 0.2648,
+      "step": 3760
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.00025778571428571425,
+      "loss": 0.2456,
+      "step": 3761
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002577704081632653,
+      "loss": 0.265,
+      "step": 3762
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002577551020408163,
+      "loss": 0.216,
+      "step": 3763
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002577397959183673,
+      "loss": 0.144,
+      "step": 3764
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00025772448979591837,
+      "loss": 0.1993,
+      "step": 3765
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00025770918367346935,
+      "loss": 0.3395,
+      "step": 3766
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002576938775510204,
+      "loss": 0.2608,
+      "step": 3767
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002576785714285714,
+      "loss": 0.2509,
+      "step": 3768
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00025766326530612244,
+      "loss": 0.1692,
+      "step": 3769
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00025764795918367347,
+      "loss": 0.2587,
+      "step": 3770
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00025763265306122444,
+      "loss": 0.2422,
+      "step": 3771
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00025761734693877553,
+      "loss": 0.2615,
+      "step": 3772
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002576020408163265,
+      "loss": 0.243,
+      "step": 3773
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00025758673469387754,
+      "loss": 0.2551,
+      "step": 3774
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002575714285714285,
+      "loss": 0.1381,
+      "step": 3775
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002575561224489796,
+      "loss": 0.2038,
+      "step": 3776
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002575408163265306,
+      "loss": 0.2113,
+      "step": 3777
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002575255102040816,
+      "loss": 0.1575,
+      "step": 3778
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00025751020408163264,
+      "loss": 0.2315,
+      "step": 3779
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00025749489795918367,
+      "loss": 0.238,
+      "step": 3780
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002574795918367347,
+      "loss": 0.1909,
+      "step": 3781
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00025746428571428567,
+      "loss": 0.1289,
+      "step": 3782
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002574489795918367,
+      "loss": 0.2084,
+      "step": 3783
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00025743367346938773,
+      "loss": 0.2816,
+      "step": 3784
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00025741836734693876,
+      "loss": 0.2914,
+      "step": 3785
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002574030612244898,
+      "loss": 0.2745,
+      "step": 3786
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00025738775510204077,
+      "loss": 0.2235,
+      "step": 3787
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002573724489795918,
+      "loss": 0.2113,
+      "step": 3788
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00025735714285714283,
+      "loss": 0.2111,
+      "step": 3789
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00025734183673469386,
+      "loss": 0.3661,
+      "step": 3790
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00025732653061224484,
+      "loss": 0.2727,
+      "step": 3791
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002573112244897959,
+      "loss": 0.2539,
+      "step": 3792
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002572959183673469,
+      "loss": 0.202,
+      "step": 3793
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00025728061224489793,
+      "loss": 0.1512,
+      "step": 3794
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00025726530612244896,
+      "loss": 0.241,
+      "step": 3795
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00025725,
+      "loss": 0.2331,
+      "step": 3796
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.000257234693877551,
+      "loss": 0.1852,
+      "step": 3797
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.000257219387755102,
+      "loss": 0.2545,
+      "step": 3798
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00025720408163265303,
+      "loss": 0.2648,
+      "step": 3799
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00025718877551020406,
+      "loss": 0.3103,
+      "step": 3800
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002571734693877551,
+      "loss": 0.1952,
+      "step": 3801
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002571581632653061,
+      "loss": 0.2075,
+      "step": 3802
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002571428571428571,
+      "loss": 0.2225,
+      "step": 3803
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00025712755102040813,
+      "loss": 0.3144,
+      "step": 3804
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00025711224489795916,
+      "loss": 0.178,
+      "step": 3805
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0002570969387755102,
+      "loss": 0.1961,
+      "step": 3806
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00025708163265306117,
+      "loss": 0.2333,
+      "step": 3807
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00025706632653061225,
+      "loss": 0.2453,
+      "step": 3808
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0002570510204081632,
+      "loss": 0.1896,
+      "step": 3809
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00025703571428571426,
+      "loss": 0.2458,
+      "step": 3810
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0002570204081632653,
+      "loss": 0.1764,
+      "step": 3811
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0002570051020408163,
+      "loss": 0.1936,
+      "step": 3812
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00025698979591836735,
+      "loss": 0.1599,
+      "step": 3813
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0002569744897959183,
+      "loss": 0.1773,
+      "step": 3814
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00025695918367346936,
+      "loss": 0.375,
+      "step": 3815
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0002569438775510204,
+      "loss": 0.2427,
+      "step": 3816
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002569285714285714,
+      "loss": 0.1827,
+      "step": 3817
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002569132653061224,
+      "loss": 0.1493,
+      "step": 3818
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002568979591836735,
+      "loss": 0.2139,
+      "step": 3819
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.00025688265306122445,
+      "loss": 0.1839,
+      "step": 3820
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002568673469387755,
+      "loss": 0.2924,
+      "step": 3821
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002568520408163265,
+      "loss": 0.3904,
+      "step": 3822
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.00025683673469387755,
+      "loss": 0.2322,
+      "step": 3823
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002568214285714286,
+      "loss": 0.245,
+      "step": 3824
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.00025680612244897955,
+      "loss": 0.1636,
+      "step": 3825
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002567908163265306,
+      "loss": 0.1917,
+      "step": 3826
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002567755102040816,
+      "loss": 0.2367,
+      "step": 3827
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00025676020408163264,
+      "loss": 0.1842,
+      "step": 3828
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002567448979591837,
+      "loss": 0.2369,
+      "step": 3829
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00025672959183673465,
+      "loss": 0.2202,
+      "step": 3830
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002567142857142857,
+      "loss": 0.1673,
+      "step": 3831
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002566989795918367,
+      "loss": 0.1793,
+      "step": 3832
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00025668367346938774,
+      "loss": 0.1656,
+      "step": 3833
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002566683673469387,
+      "loss": 0.3911,
+      "step": 3834
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002566530612244898,
+      "loss": 0.2271,
+      "step": 3835
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002566377551020408,
+      "loss": 0.1849,
+      "step": 3836
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.0002566224489795918,
+      "loss": 0.2092,
+      "step": 3837
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00025660714285714284,
+      "loss": 0.1966,
+      "step": 3838
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00025659183673469387,
+      "loss": 0.1964,
+      "step": 3839
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.0002565765306122449,
+      "loss": 0.3351,
+      "step": 3840
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.0002565612244897959,
+      "loss": 0.2695,
+      "step": 3841
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.0002565459183673469,
+      "loss": 0.2069,
+      "step": 3842
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00025653061224489794,
+      "loss": 0.2627,
+      "step": 3843
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00025651530612244897,
+      "loss": 0.1349,
+      "step": 3844
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00025649999999999995,
+      "loss": 0.1861,
+      "step": 3845
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.000256484693877551,
+      "loss": 0.1886,
+      "step": 3846
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.000256469387755102,
+      "loss": 0.3066,
+      "step": 3847
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00025645408163265304,
+      "loss": 0.223,
+      "step": 3848
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00025643877551020407,
+      "loss": 0.2428,
+      "step": 3849
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00025642346938775504,
+      "loss": 0.2644,
+      "step": 3850
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00025640816326530613,
+      "loss": 0.1336,
+      "step": 3851
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.0002563928571428571,
+      "loss": 0.234,
+      "step": 3852
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00025637755102040814,
+      "loss": 0.1772,
+      "step": 3853
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00025636224489795917,
+      "loss": 0.267,
+      "step": 3854
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.0002563469387755102,
+      "loss": 0.2132,
+      "step": 3855
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00025633163265306123,
+      "loss": 0.1656,
+      "step": 3856
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.0002563163265306122,
+      "loss": 0.2138,
+      "step": 3857
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00025630102040816324,
+      "loss": 0.198,
+      "step": 3858
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00025628571428571427,
+      "loss": 0.3281,
+      "step": 3859
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0002562704081632653,
+      "loss": 0.227,
+      "step": 3860
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00025625510204081627,
+      "loss": 0.2243,
+      "step": 3861
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00025623979591836736,
+      "loss": 0.2618,
+      "step": 3862
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00025622448979591833,
+      "loss": 0.2092,
+      "step": 3863
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00025620918367346936,
+      "loss": 0.22,
+      "step": 3864
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0002561938775510204,
+      "loss": 0.3818,
+      "step": 3865
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0002561785714285714,
+      "loss": 0.3288,
+      "step": 3866
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00025616326530612246,
+      "loss": 0.1974,
+      "step": 3867
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00025614795918367343,
+      "loss": 0.2272,
+      "step": 3868
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00025613265306122446,
+      "loss": 0.24,
+      "step": 3869
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002561173469387755,
+      "loss": 0.1543,
+      "step": 3870
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002561020408163265,
+      "loss": 0.2349,
+      "step": 3871
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002560867346938775,
+      "loss": 0.2084,
+      "step": 3872
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00025607142857142853,
+      "loss": 0.2853,
+      "step": 3873
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00025605612244897956,
+      "loss": 0.2128,
+      "step": 3874
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002560408163265306,
+      "loss": 0.1592,
+      "step": 3875
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002560255102040816,
+      "loss": 0.2184,
+      "step": 3876
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002560102040816326,
+      "loss": 0.1736,
+      "step": 3877
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002559948979591837,
+      "loss": 0.2737,
+      "step": 3878
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00025597959183673466,
+      "loss": 0.2238,
+      "step": 3879
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002559642857142857,
+      "loss": 0.199,
+      "step": 3880
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002559489795918367,
+      "loss": 0.1937,
+      "step": 3881
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00025593367346938775,
+      "loss": 0.2349,
+      "step": 3882
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002559183673469388,
+      "loss": 0.1864,
+      "step": 3883
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00025590306122448976,
+      "loss": 0.1337,
+      "step": 3884
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002558877551020408,
+      "loss": 0.3337,
+      "step": 3885
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002558724489795918,
+      "loss": 0.192,
+      "step": 3886
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00025585714285714285,
+      "loss": 0.195,
+      "step": 3887
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002558418367346938,
+      "loss": 0.2136,
+      "step": 3888
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00025582653061224486,
+      "loss": 0.1805,
+      "step": 3889
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.0002558112244897959,
+      "loss": 0.1962,
+      "step": 3890
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.0002557959183673469,
+      "loss": 0.21,
+      "step": 3891
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00025578061224489795,
+      "loss": 0.328,
+      "step": 3892
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.0002557653061224489,
+      "loss": 0.1924,
+      "step": 3893
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00025575,
+      "loss": 0.2274,
+      "step": 3894
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.000255734693877551,
+      "loss": 0.1509,
+      "step": 3895
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.000255719387755102,
+      "loss": 0.1775,
+      "step": 3896
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00025570408163265305,
+      "loss": 0.3961,
+      "step": 3897
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.0002556887755102041,
+      "loss": 0.2822,
+      "step": 3898
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0002556734693877551,
+      "loss": 0.3305,
+      "step": 3899
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0002556581632653061,
+      "loss": 0.1862,
+      "step": 3900
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0002556428571428571,
+      "loss": 0.1781,
+      "step": 3901
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00025562755102040815,
+      "loss": 0.2583,
+      "step": 3902
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0002556122448979592,
+      "loss": 0.2468,
+      "step": 3903
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00025559693877551015,
+      "loss": 0.2847,
+      "step": 3904
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0002555816326530612,
+      "loss": 0.2157,
+      "step": 3905
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0002555663265306122,
+      "loss": 0.2556,
+      "step": 3906
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00025555102040816324,
+      "loss": 0.2326,
+      "step": 3907
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0002555357142857143,
+      "loss": 0.2076,
+      "step": 3908
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002555204081632653,
+      "loss": 0.3056,
+      "step": 3909
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00025550510204081634,
+      "loss": 0.184,
+      "step": 3910
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002554897959183673,
+      "loss": 0.2191,
+      "step": 3911
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00025547448979591834,
+      "loss": 0.2154,
+      "step": 3912
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002554591836734694,
+      "loss": 0.2578,
+      "step": 3913
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002554438775510204,
+      "loss": 0.2172,
+      "step": 3914
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002554285714285714,
+      "loss": 0.2943,
+      "step": 3915
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002554132653061224,
+      "loss": 0.3637,
+      "step": 3916
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00025539795918367344,
+      "loss": 0.2209,
+      "step": 3917
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00025538265306122447,
+      "loss": 0.1571,
+      "step": 3918
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002553673469387755,
+      "loss": 0.2727,
+      "step": 3919
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.0002553520408163265,
+      "loss": 0.1625,
+      "step": 3920
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00025533673469387756,
+      "loss": 0.2729,
+      "step": 3921
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00025532142857142854,
+      "loss": 0.3051,
+      "step": 3922
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00025530612244897957,
+      "loss": 0.2805,
+      "step": 3923
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.0002552908163265306,
+      "loss": 0.152,
+      "step": 3924
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00025527551020408163,
+      "loss": 0.1913,
+      "step": 3925
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00025526020408163266,
+      "loss": 0.2064,
+      "step": 3926
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00025524489795918364,
+      "loss": 0.2474,
+      "step": 3927
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00025522959183673467,
+      "loss": 0.392,
+      "step": 3928
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.0002552142857142857,
+      "loss": 0.2727,
+      "step": 3929
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00025519897959183673,
+      "loss": 0.1658,
+      "step": 3930
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0002551836734693877,
+      "loss": 0.2627,
+      "step": 3931
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00025516836734693874,
+      "loss": 0.1822,
+      "step": 3932
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00025515306122448977,
+      "loss": 0.185,
+      "step": 3933
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0002551377551020408,
+      "loss": 0.3681,
+      "step": 3934
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00025512244897959183,
+      "loss": 0.1998,
+      "step": 3935
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0002551071428571428,
+      "loss": 0.2334,
+      "step": 3936
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0002550918367346939,
+      "loss": 0.2324,
+      "step": 3937
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00025507653061224487,
+      "loss": 0.1584,
+      "step": 3938
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0002550612244897959,
+      "loss": 0.1935,
+      "step": 3939
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.0002550459183673469,
+      "loss": 0.1849,
+      "step": 3940
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00025503061224489796,
+      "loss": 0.2584,
+      "step": 3941
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00025501530612244893,
+      "loss": 0.3167,
+      "step": 3942
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 0.2356,
+      "step": 3943
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.000254984693877551,
+      "loss": 0.2476,
+      "step": 3944
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.000254969387755102,
+      "loss": 0.2624,
+      "step": 3945
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00025495408163265306,
+      "loss": 0.3264,
+      "step": 3946
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00025493877551020403,
+      "loss": 0.3682,
+      "step": 3947
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00025492346938775506,
+      "loss": 0.2323,
+      "step": 3948
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.0002549081632653061,
+      "loss": 0.1991,
+      "step": 3949
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.0002548928571428571,
+      "loss": 0.1715,
+      "step": 3950
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00025487755102040815,
+      "loss": 0.2016,
+      "step": 3951
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00025486224489795913,
+      "loss": 0.2141,
+      "step": 3952
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.0002548469387755102,
+      "loss": 0.2837,
+      "step": 3953
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.0002548316326530612,
+      "loss": 0.1869,
+      "step": 3954
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.0002548163265306122,
+      "loss": 0.2198,
+      "step": 3955
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00025480102040816325,
+      "loss": 0.3192,
+      "step": 3956
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.0002547857142857143,
+      "loss": 0.2714,
+      "step": 3957
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00025477040816326526,
+      "loss": 0.1828,
+      "step": 3958
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.0002547551020408163,
+      "loss": 0.3408,
+      "step": 3959
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.0002547397959183673,
+      "loss": 0.2797,
+      "step": 3960
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00025472448979591835,
+      "loss": 0.2205,
+      "step": 3961
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0002547091836734694,
+      "loss": 0.2279,
+      "step": 3962
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00025469387755102036,
+      "loss": 0.2333,
+      "step": 3963
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00025467857142857144,
+      "loss": 0.1643,
+      "step": 3964
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0002546632653061224,
+      "loss": 0.2691,
+      "step": 3965
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00025464795918367345,
+      "loss": 0.2831,
+      "step": 3966
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0002546326530612245,
+      "loss": 0.176,
+      "step": 3967
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0002546173469387755,
+      "loss": 0.2056,
+      "step": 3968
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0002546020408163265,
+      "loss": 0.1648,
+      "step": 3969
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0002545867346938775,
+      "loss": 0.1796,
+      "step": 3970
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00025457142857142855,
+      "loss": 0.1572,
+      "step": 3971
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0002545561224489796,
+      "loss": 0.2364,
+      "step": 3972
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0002545408163265306,
+      "loss": 0.243,
+      "step": 3973
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0002545255102040816,
+      "loss": 0.2653,
+      "step": 3974
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0002545102040816326,
+      "loss": 0.2566,
+      "step": 3975
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00025449489795918365,
+      "loss": 0.1926,
+      "step": 3976
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0002544795918367347,
+      "loss": 0.1979,
+      "step": 3977
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0002544642857142857,
+      "loss": 0.3407,
+      "step": 3978
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0002544489795918367,
+      "loss": 0.2307,
+      "step": 3979
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00025443367346938777,
+      "loss": 0.2944,
+      "step": 3980
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.00025441836734693875,
+      "loss": 0.2789,
+      "step": 3981
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0002544030612244898,
+      "loss": 0.2261,
+      "step": 3982
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0002543877551020408,
+      "loss": 0.1331,
+      "step": 3983
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.00025437244897959184,
+      "loss": 0.4667,
+      "step": 3984
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0002543571428571428,
+      "loss": 0.2087,
+      "step": 3985
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.00025434183673469384,
+      "loss": 0.1448,
+      "step": 3986
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0002543265306122449,
+      "loss": 0.228,
+      "step": 3987
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0002543112244897959,
+      "loss": 0.1534,
+      "step": 3988
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.00025429591836734694,
+      "loss": 0.2058,
+      "step": 3989
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0002542806122448979,
+      "loss": 0.2884,
+      "step": 3990
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.00025426530612244894,
+      "loss": 0.2778,
+      "step": 3991
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025425,
+      "loss": 0.2268,
+      "step": 3992
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.000254234693877551,
+      "loss": 0.1891,
+      "step": 3993
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025421938775510203,
+      "loss": 0.1625,
+      "step": 3994
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.000254204081632653,
+      "loss": 0.1277,
+      "step": 3995
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025418877551020404,
+      "loss": 0.2786,
+      "step": 3996
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025417346938775507,
+      "loss": 0.2229,
+      "step": 3997
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.0002541581632653061,
+      "loss": 0.2047,
+      "step": 3998
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025414285714285713,
+      "loss": 0.2483,
+      "step": 3999
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025412755102040816,
+      "loss": 0.2489,
+      "step": 4000
+    },
+    {
+      "epoch": 3.88,
+      "eval_cer": 0.2690840508519273,
+      "eval_loss": 0.8955472111701965,
+      "eval_runtime": 227.1874,
+      "eval_samples_per_second": 38.796,
+      "eval_steps_per_second": 4.851,
+      "eval_wer": 0.6743510309901241,
+      "step": 4000
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025411224489795914,
+      "loss": 0.2243,
+      "step": 4001
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00025409693877551017,
+      "loss": 0.225,
+      "step": 4002
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0002540816326530612,
+      "loss": 0.3268,
+      "step": 4003
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00025406632653061223,
+      "loss": 0.389,
+      "step": 4004
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00025405102040816326,
+      "loss": 0.228,
+      "step": 4005
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00025403571428571424,
+      "loss": 0.1548,
+      "step": 4006
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0002540204081632653,
+      "loss": 0.2272,
+      "step": 4007
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0002540051020408163,
+      "loss": 0.1385,
+      "step": 4008
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00025398979591836733,
+      "loss": 0.1797,
+      "step": 4009
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00025397448979591836,
+      "loss": 0.3804,
+      "step": 4010
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0002539591836734694,
+      "loss": 0.2606,
+      "step": 4011
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00025394387755102037,
+      "loss": 0.2261,
+      "step": 4012
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.0002539285714285714,
+      "loss": 0.1672,
+      "step": 4013
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00025391326530612243,
+      "loss": 0.2069,
+      "step": 4014
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00025389795918367346,
+      "loss": 0.2889,
+      "step": 4015
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.0002538826530612245,
+      "loss": 0.2324,
+      "step": 4016
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00025386734693877547,
+      "loss": 0.2354,
+      "step": 4017
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.0002538520408163265,
+      "loss": 0.1602,
+      "step": 4018
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.0002538367346938775,
+      "loss": 0.2952,
+      "step": 4019
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00025382142857142856,
+      "loss": 0.2481,
+      "step": 4020
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.0002538061224489796,
+      "loss": 0.1169,
+      "step": 4021
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00025379081632653056,
+      "loss": 0.2298,
+      "step": 4022
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.00025377551020408165,
+      "loss": 0.2325,
+      "step": 4023
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0002537602040816326,
+      "loss": 0.1879,
+      "step": 4024
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.00025374489795918366,
+      "loss": 0.1795,
+      "step": 4025
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.00025372959183673463,
+      "loss": 0.2403,
+      "step": 4026
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0002537142857142857,
+      "loss": 0.1747,
+      "step": 4027
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0002536989795918367,
+      "loss": 0.3255,
+      "step": 4028
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0002536836734693877,
+      "loss": 0.2363,
+      "step": 4029
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.00025366836734693875,
+      "loss": 0.3009,
+      "step": 4030
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0002536530612244898,
+      "loss": 0.2654,
+      "step": 4031
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0002536377551020408,
+      "loss": 0.1867,
+      "step": 4032
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002536224489795918,
+      "loss": 0.2745,
+      "step": 4033
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002536071428571428,
+      "loss": 0.2253,
+      "step": 4034
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00025359183673469385,
+      "loss": 0.2779,
+      "step": 4035
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002535765306122449,
+      "loss": 0.2334,
+      "step": 4036
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002535612244897959,
+      "loss": 0.2196,
+      "step": 4037
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002535459183673469,
+      "loss": 0.2342,
+      "step": 4038
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002535306122448979,
+      "loss": 0.2463,
+      "step": 4039
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00025351530612244895,
+      "loss": 0.2017,
+      "step": 4040
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002535,
+      "loss": 0.2386,
+      "step": 4041
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00025348469387755096,
+      "loss": 0.3105,
+      "step": 4042
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.00025346938775510204,
+      "loss": 0.2274,
+      "step": 4043
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.000253454081632653,
+      "loss": 0.3495,
+      "step": 4044
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.00025343877551020405,
+      "loss": 0.1709,
+      "step": 4045
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0002534234693877551,
+      "loss": 0.1926,
+      "step": 4046
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0002534081632653061,
+      "loss": 0.4468,
+      "step": 4047
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.00025339285714285714,
+      "loss": 0.2141,
+      "step": 4048
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0002533775510204081,
+      "loss": 0.2371,
+      "step": 4049
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0002533622448979592,
+      "loss": 0.1857,
+      "step": 4050
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0002533469387755102,
+      "loss": 0.1716,
+      "step": 4051
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0002533316326530612,
+      "loss": 0.1812,
+      "step": 4052
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0002533163265306122,
+      "loss": 0.2956,
+      "step": 4053
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00025330102040816327,
+      "loss": 0.1843,
+      "step": 4054
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00025328571428571425,
+      "loss": 0.2596,
+      "step": 4055
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0002532704081632653,
+      "loss": 0.2892,
+      "step": 4056
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0002532551020408163,
+      "loss": 0.1446,
+      "step": 4057
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00025323979591836734,
+      "loss": 0.2019,
+      "step": 4058
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00025322448979591837,
+      "loss": 0.2925,
+      "step": 4059
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00025320918367346935,
+      "loss": 0.2973,
+      "step": 4060
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0002531938775510204,
+      "loss": 0.2934,
+      "step": 4061
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0002531785714285714,
+      "loss": 0.1815,
+      "step": 4062
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00025316326530612244,
+      "loss": 0.1606,
+      "step": 4063
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00025314795918367347,
+      "loss": 0.1781,
+      "step": 4064
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00025313265306122444,
+      "loss": 0.2234,
+      "step": 4065
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0002531173469387755,
+      "loss": 0.3011,
+      "step": 4066
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0002531020408163265,
+      "loss": 0.1468,
+      "step": 4067
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00025308673469387754,
+      "loss": 0.2608,
+      "step": 4068
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0002530714285714285,
+      "loss": 0.261,
+      "step": 4069
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0002530561224489796,
+      "loss": 0.2346,
+      "step": 4070
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0002530408163265306,
+      "loss": 0.2391,
+      "step": 4071
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0002530255102040816,
+      "loss": 0.2956,
+      "step": 4072
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00025301020408163263,
+      "loss": 0.2789,
+      "step": 4073
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00025299489795918366,
+      "loss": 0.1699,
+      "step": 4074
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.0002529795918367347,
+      "loss": 0.1332,
+      "step": 4075
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00025296428571428567,
+      "loss": 0.3549,
+      "step": 4076
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.0002529489795918367,
+      "loss": 0.2417,
+      "step": 4077
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00025293367346938773,
+      "loss": 0.407,
+      "step": 4078
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00025291836734693876,
+      "loss": 0.268,
+      "step": 4079
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.0002529030612244898,
+      "loss": 0.1428,
+      "step": 4080
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00025288775510204077,
+      "loss": 0.2024,
+      "step": 4081
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.0002528724489795918,
+      "loss": 0.2328,
+      "step": 4082
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00025285714285714283,
+      "loss": 0.2197,
+      "step": 4083
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00025284183673469386,
+      "loss": 0.26,
+      "step": 4084
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00025282653061224484,
+      "loss": 0.1924,
+      "step": 4085
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.0002528112244897959,
+      "loss": 0.197,
+      "step": 4086
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.0002527959183673469,
+      "loss": 0.2115,
+      "step": 4087
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00025278061224489793,
+      "loss": 0.2543,
+      "step": 4088
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00025276530612244896,
+      "loss": 0.2042,
+      "step": 4089
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00025275,
+      "loss": 0.1949,
+      "step": 4090
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.000252734693877551,
+      "loss": 0.3027,
+      "step": 4091
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.000252719387755102,
+      "loss": 0.2087,
+      "step": 4092
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00025270408163265303,
+      "loss": 0.3047,
+      "step": 4093
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00025268877551020406,
+      "loss": 0.176,
+      "step": 4094
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0002526734693877551,
+      "loss": 0.1532,
+      "step": 4095
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00025265816326530607,
+      "loss": 0.2851,
+      "step": 4096
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00025264285714285715,
+      "loss": 0.1888,
+      "step": 4097
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0002526275510204081,
+      "loss": 0.3039,
+      "step": 4098
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00025261224489795916,
+      "loss": 0.2327,
+      "step": 4099
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0002525969387755102,
+      "loss": 0.3064,
+      "step": 4100
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0002525816326530612,
+      "loss": 0.2049,
+      "step": 4101
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00025256632653061225,
+      "loss": 0.1697,
+      "step": 4102
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0002525510204081632,
+      "loss": 0.2722,
+      "step": 4103
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00025253571428571426,
+      "loss": 0.2822,
+      "step": 4104
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0002525204081632653,
+      "loss": 0.3013,
+      "step": 4105
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0002525051020408163,
+      "loss": 0.1653,
+      "step": 4106
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.00025248979591836735,
+      "loss": 0.1847,
+      "step": 4107
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0002524744897959183,
+      "loss": 0.2483,
+      "step": 4108
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.00025245918367346935,
+      "loss": 0.2755,
+      "step": 4109
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0002524438775510204,
+      "loss": 0.289,
+      "step": 4110
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0002524285714285714,
+      "loss": 0.2044,
+      "step": 4111
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0002524132653061224,
+      "loss": 0.1821,
+      "step": 4112
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0002523979591836735,
+      "loss": 0.252,
+      "step": 4113
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.00025238265306122445,
+      "loss": 0.1606,
+      "step": 4114
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0002523673469387755,
+      "loss": 0.3255,
+      "step": 4115
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0002523520408163265,
+      "loss": 0.2732,
+      "step": 4116
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.00025233673469387754,
+      "loss": 0.1892,
+      "step": 4117
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0002523214285714286,
+      "loss": 0.2335,
+      "step": 4118
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.00025230612244897955,
+      "loss": 0.1783,
+      "step": 4119
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0002522908163265306,
+      "loss": 0.2991,
+      "step": 4120
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0002522755102040816,
+      "loss": 0.2351,
+      "step": 4121
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.00025226020408163264,
+      "loss": 0.2324,
+      "step": 4122
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0002522448979591836,
+      "loss": 0.139,
+      "step": 4123
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.00025222959183673465,
+      "loss": 0.182,
+      "step": 4124
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0002522142857142857,
+      "loss": 0.1475,
+      "step": 4125
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002521989795918367,
+      "loss": 0.3074,
+      "step": 4126
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.00025218367346938774,
+      "loss": 0.4323,
+      "step": 4127
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002521683673469387,
+      "loss": 0.1546,
+      "step": 4128
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002521530612244898,
+      "loss": 0.1327,
+      "step": 4129
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002521377551020408,
+      "loss": 0.1773,
+      "step": 4130
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002521224489795918,
+      "loss": 0.136,
+      "step": 4131
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.00025210714285714284,
+      "loss": 0.1779,
+      "step": 4132
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.00025209183673469387,
+      "loss": 0.2673,
+      "step": 4133
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002520765306122449,
+      "loss": 0.1977,
+      "step": 4134
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002520612244897959,
+      "loss": 0.1619,
+      "step": 4135
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.0002520459183673469,
+      "loss": 0.2386,
+      "step": 4136
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00025203061224489794,
+      "loss": 0.1858,
+      "step": 4137
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00025201530612244897,
+      "loss": 0.1178,
+      "step": 4138
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00025199999999999995,
+      "loss": 0.2323,
+      "step": 4139
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00025198469387755103,
+      "loss": 0.2038,
+      "step": 4140
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.000251969387755102,
+      "loss": 0.1997,
+      "step": 4141
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00025195408163265304,
+      "loss": 0.1668,
+      "step": 4142
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00025193877551020407,
+      "loss": 0.1376,
+      "step": 4143
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.0002519234693877551,
+      "loss": 0.1116,
+      "step": 4144
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00025190816326530613,
+      "loss": 0.1731,
+      "step": 4145
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.0002518928571428571,
+      "loss": 0.343,
+      "step": 4146
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00025187755102040814,
+      "loss": 0.1485,
+      "step": 4147
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00025186224489795917,
+      "loss": 0.2264,
+      "step": 4148
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.0002518469387755102,
+      "loss": 0.1523,
+      "step": 4149
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.0002518316326530612,
+      "loss": 0.1068,
+      "step": 4150
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.0002518163265306122,
+      "loss": 0.2394,
+      "step": 4151
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00025180102040816323,
+      "loss": 0.3647,
+      "step": 4152
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00025178571428571426,
+      "loss": 0.1581,
+      "step": 4153
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.0002517704081632653,
+      "loss": 0.167,
+      "step": 4154
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00025175510204081627,
+      "loss": 0.1151,
+      "step": 4155
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00025173979591836736,
+      "loss": 0.1329,
+      "step": 4156
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.00025172448979591833,
+      "loss": 0.1246,
+      "step": 4157
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.00025170918367346936,
+      "loss": 0.2067,
+      "step": 4158
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002516938775510204,
+      "loss": 0.2121,
+      "step": 4159
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002516785714285714,
+      "loss": 0.1397,
+      "step": 4160
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.00025166326530612245,
+      "loss": 0.1856,
+      "step": 4161
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.00025164795918367343,
+      "loss": 0.2279,
+      "step": 4162
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.00025163265306122446,
+      "loss": 0.1354,
+      "step": 4163
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002516173469387755,
+      "loss": 0.1423,
+      "step": 4164
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002516020408163265,
+      "loss": 0.3131,
+      "step": 4165
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002515867346938775,
+      "loss": 0.1931,
+      "step": 4166
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.00025157142857142853,
+      "loss": 0.155,
+      "step": 4167
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.00025155612244897956,
+      "loss": 0.2978,
+      "step": 4168
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0002515408163265306,
+      "loss": 0.1664,
+      "step": 4169
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0002515255102040816,
+      "loss": 0.2257,
+      "step": 4170
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0002515102040816326,
+      "loss": 0.1993,
+      "step": 4171
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0002514948979591837,
+      "loss": 0.2697,
+      "step": 4172
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.00025147959183673466,
+      "loss": 0.1682,
+      "step": 4173
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0002514642857142857,
+      "loss": 0.1208,
+      "step": 4174
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0002514489795918367,
+      "loss": 0.1358,
+      "step": 4175
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.00025143367346938775,
+      "loss": 0.212,
+      "step": 4176
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0002514183673469388,
+      "loss": 0.2392,
+      "step": 4177
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00025140306122448976,
+      "loss": 0.1606,
+      "step": 4178
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0002513877551020408,
+      "loss": 0.2154,
+      "step": 4179
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0002513724489795918,
+      "loss": 0.1841,
+      "step": 4180
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00025135714285714285,
+      "loss": 0.1256,
+      "step": 4181
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0002513418367346938,
+      "loss": 0.1531,
+      "step": 4182
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0002513265306122449,
+      "loss": 0.2606,
+      "step": 4183
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0002513112244897959,
+      "loss": 0.2312,
+      "step": 4184
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0002512959183673469,
+      "loss": 0.1808,
+      "step": 4185
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00025128061224489795,
+      "loss": 0.1787,
+      "step": 4186
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.000251265306122449,
+      "loss": 0.1925,
+      "step": 4187
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.00025125,
+      "loss": 0.1089,
+      "step": 4188
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.000251234693877551,
+      "loss": 0.226,
+      "step": 4189
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.000251219387755102,
+      "loss": 0.2945,
+      "step": 4190
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.00025120408163265305,
+      "loss": 0.1514,
+      "step": 4191
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0002511887755102041,
+      "loss": 0.2224,
+      "step": 4192
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.00025117346938775505,
+      "loss": 0.217,
+      "step": 4193
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0002511581632653061,
+      "loss": 0.1447,
+      "step": 4194
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0002511428571428571,
+      "loss": 0.299,
+      "step": 4195
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.00025112755102040814,
+      "loss": 0.3174,
+      "step": 4196
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0002511122448979592,
+      "loss": 0.2018,
+      "step": 4197
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.00025109693877551015,
+      "loss": 0.1069,
+      "step": 4198
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.00025108163265306124,
+      "loss": 0.1517,
+      "step": 4199
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0002510663265306122,
+      "loss": 0.1346,
+      "step": 4200
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.00025105102040816324,
+      "loss": 0.1733,
+      "step": 4201
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0002510357142857143,
+      "loss": 0.2477,
+      "step": 4202
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0002510204081632653,
+      "loss": 0.1497,
+      "step": 4203
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.00025100510204081633,
+      "loss": 0.1881,
+      "step": 4204
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0002509897959183673,
+      "loss": 0.2658,
+      "step": 4205
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.00025097448979591834,
+      "loss": 0.1331,
+      "step": 4206
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.00025095918367346937,
+      "loss": 0.1522,
+      "step": 4207
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0002509438775510204,
+      "loss": 0.2451,
+      "step": 4208
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0002509285714285714,
+      "loss": 0.2153,
+      "step": 4209
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0002509132653061224,
+      "loss": 0.277,
+      "step": 4210
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.00025089795918367344,
+      "loss": 0.3271,
+      "step": 4211
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.00025088265306122447,
+      "loss": 0.2729,
+      "step": 4212
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0002508673469387755,
+      "loss": 0.1558,
+      "step": 4213
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0002508520408163265,
+      "loss": 0.3483,
+      "step": 4214
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.00025083673469387756,
+      "loss": 0.1598,
+      "step": 4215
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.00025082142857142854,
+      "loss": 0.2381,
+      "step": 4216
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.00025080612244897957,
+      "loss": 0.1623,
+      "step": 4217
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.0002507908163265306,
+      "loss": 0.2126,
+      "step": 4218
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.00025077551020408163,
+      "loss": 0.1463,
+      "step": 4219
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.0002507602040816326,
+      "loss": 0.2446,
+      "step": 4220
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.00025074489795918364,
+      "loss": 0.3434,
+      "step": 4221
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.00025072959183673467,
+      "loss": 0.213,
+      "step": 4222
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.0002507142857142857,
+      "loss": 0.1899,
+      "step": 4223
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.00025069897959183673,
+      "loss": 0.2093,
+      "step": 4224
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.0002506836734693877,
+      "loss": 0.2478,
+      "step": 4225
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.0002506683673469388,
+      "loss": 0.1455,
+      "step": 4226
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.00025065306122448977,
+      "loss": 0.1872,
+      "step": 4227
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.0002506377551020408,
+      "loss": 0.3051,
+      "step": 4228
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.00025062244897959183,
+      "loss": 0.1386,
+      "step": 4229
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.00025060714285714286,
+      "loss": 0.1324,
+      "step": 4230
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.0002505918367346939,
+      "loss": 0.1823,
+      "step": 4231
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.00025057653061224486,
+      "loss": 0.1197,
+      "step": 4232
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.0002505612244897959,
+      "loss": 0.2417,
+      "step": 4233
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.0002505459183673469,
+      "loss": 0.1991,
+      "step": 4234
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.00025053061224489796,
+      "loss": 0.1302,
+      "step": 4235
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.00025051530612244893,
+      "loss": 0.1656,
+      "step": 4236
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.00025049999999999996,
+      "loss": 0.1815,
+      "step": 4237
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.000250484693877551,
+      "loss": 0.1583,
+      "step": 4238
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.000250469387755102,
+      "loss": 0.2159,
+      "step": 4239
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.00025045408163265305,
+      "loss": 0.1836,
+      "step": 4240
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.00025043877551020403,
+      "loss": 0.1839,
+      "step": 4241
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0002504234693877551,
+      "loss": 0.1377,
+      "step": 4242
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0002504081632653061,
+      "loss": 0.1709,
+      "step": 4243
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0002503928571428571,
+      "loss": 0.1234,
+      "step": 4244
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.00025037755102040815,
+      "loss": 0.12,
+      "step": 4245
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0002503622448979592,
+      "loss": 0.3733,
+      "step": 4246
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.00025034693877551016,
+      "loss": 0.1526,
+      "step": 4247
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0002503316326530612,
+      "loss": 0.1847,
+      "step": 4248
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.0002503163265306122,
+      "loss": 0.1819,
+      "step": 4249
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00025030102040816325,
+      "loss": 0.1382,
+      "step": 4250
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.0002502857142857143,
+      "loss": 0.189,
+      "step": 4251
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00025027040816326526,
+      "loss": 0.2842,
+      "step": 4252
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.0002502551020408163,
+      "loss": 0.1674,
+      "step": 4253
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.0002502397959183673,
+      "loss": 0.1439,
+      "step": 4254
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00025022448979591835,
+      "loss": 0.1648,
+      "step": 4255
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.0002502091836734694,
+      "loss": 0.2117,
+      "step": 4256
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00025019387755102036,
+      "loss": 0.2078,
+      "step": 4257
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00025017857142857144,
+      "loss": 0.2408,
+      "step": 4258
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.0002501632653061224,
+      "loss": 0.2174,
+      "step": 4259
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.00025014795918367345,
+      "loss": 0.1777,
+      "step": 4260
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002501326530612245,
+      "loss": 0.1918,
+      "step": 4261
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002501173469387755,
+      "loss": 0.2101,
+      "step": 4262
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002501020408163265,
+      "loss": 0.1404,
+      "step": 4263
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002500867346938775,
+      "loss": 0.2201,
+      "step": 4264
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.00025007142857142855,
+      "loss": 0.2496,
+      "step": 4265
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002500561224489796,
+      "loss": 0.2222,
+      "step": 4266
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002500408163265306,
+      "loss": 0.1888,
+      "step": 4267
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002500255102040816,
+      "loss": 0.2088,
+      "step": 4268
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002500102040816326,
+      "loss": 0.1391,
+      "step": 4269
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.00024999489795918365,
+      "loss": 0.2199,
+      "step": 4270
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002499795918367347,
+      "loss": 0.2374,
+      "step": 4271
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002499642857142857,
+      "loss": 0.1971,
+      "step": 4272
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.00024994897959183674,
+      "loss": 0.1494,
+      "step": 4273
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002499336734693877,
+      "loss": 0.1109,
+      "step": 4274
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.00024991836734693874,
+      "loss": 0.1195,
+      "step": 4275
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002499030612244898,
+      "loss": 0.199,
+      "step": 4276
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002498877551020408,
+      "loss": 0.1824,
+      "step": 4277
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.00024987244897959184,
+      "loss": 0.1888,
+      "step": 4278
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002498571428571428,
+      "loss": 0.1929,
+      "step": 4279
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00024984183673469384,
+      "loss": 0.2038,
+      "step": 4280
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.0002498265306122449,
+      "loss": 0.1511,
+      "step": 4281
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.0002498112244897959,
+      "loss": 0.208,
+      "step": 4282
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00024979591836734693,
+      "loss": 0.3551,
+      "step": 4283
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.0002497806122448979,
+      "loss": 0.3133,
+      "step": 4284
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.000249765306122449,
+      "loss": 0.2312,
+      "step": 4285
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00024974999999999997,
+      "loss": 0.1364,
+      "step": 4286
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.000249734693877551,
+      "loss": 0.1558,
+      "step": 4287
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00024971938775510203,
+      "loss": 0.1555,
+      "step": 4288
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00024970408163265306,
+      "loss": 0.2349,
+      "step": 4289
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024968877551020404,
+      "loss": 0.3057,
+      "step": 4290
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024967346938775507,
+      "loss": 0.2296,
+      "step": 4291
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.0002496581632653061,
+      "loss": 0.1311,
+      "step": 4292
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024964285714285713,
+      "loss": 0.2563,
+      "step": 4293
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024962755102040816,
+      "loss": 0.1557,
+      "step": 4294
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024961224489795914,
+      "loss": 0.1703,
+      "step": 4295
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024959693877551017,
+      "loss": 0.1811,
+      "step": 4296
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.0002495816326530612,
+      "loss": 0.1862,
+      "step": 4297
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024956632653061223,
+      "loss": 0.1912,
+      "step": 4298
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024955102040816326,
+      "loss": 0.1318,
+      "step": 4299
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024953571428571424,
+      "loss": 0.1065,
+      "step": 4300
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0002495204081632653,
+      "loss": 0.1537,
+      "step": 4301
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0002495051020408163,
+      "loss": 0.2086,
+      "step": 4302
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00024948979591836733,
+      "loss": 0.1838,
+      "step": 4303
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0002494744897959183,
+      "loss": 0.3242,
+      "step": 4304
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0002494591836734694,
+      "loss": 0.1396,
+      "step": 4305
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00024944387755102037,
+      "loss": 0.1598,
+      "step": 4306
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0002494285714285714,
+      "loss": 0.1648,
+      "step": 4307
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00024941326530612243,
+      "loss": 0.1998,
+      "step": 4308
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00024939795918367346,
+      "loss": 0.2839,
+      "step": 4309
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0002493826530612245,
+      "loss": 0.1547,
+      "step": 4310
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.00024936734693877546,
+      "loss": 0.1193,
+      "step": 4311
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002493520408163265,
+      "loss": 0.1209,
+      "step": 4312
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002493367346938775,
+      "loss": 0.1579,
+      "step": 4313
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.00024932142857142856,
+      "loss": 0.2497,
+      "step": 4314
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002493061224489796,
+      "loss": 0.2127,
+      "step": 4315
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002492908163265306,
+      "loss": 0.2242,
+      "step": 4316
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002492755102040816,
+      "loss": 0.1589,
+      "step": 4317
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002492602040816326,
+      "loss": 0.1112,
+      "step": 4318
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.00024924489795918365,
+      "loss": 0.1674,
+      "step": 4319
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002492295918367347,
+      "loss": 0.2492,
+      "step": 4320
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002492142857142857,
+      "loss": 0.1843,
+      "step": 4321
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002491989795918367,
+      "loss": 0.2005,
+      "step": 4322
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002491836734693877,
+      "loss": 0.2031,
+      "step": 4323
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.00024916836734693875,
+      "loss": 0.2128,
+      "step": 4324
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002491530612244898,
+      "loss": 0.1921,
+      "step": 4325
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002491377551020408,
+      "loss": 0.137,
+      "step": 4326
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002491224489795918,
+      "loss": 0.2238,
+      "step": 4327
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002491071428571429,
+      "loss": 0.2036,
+      "step": 4328
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.00024909183673469385,
+      "loss": 0.1762,
+      "step": 4329
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002490765306122449,
+      "loss": 0.2208,
+      "step": 4330
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.00024906122448979586,
+      "loss": 0.151,
+      "step": 4331
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.00024904591836734694,
+      "loss": 0.1339,
+      "step": 4332
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.0002490306122448979,
+      "loss": 0.2944,
+      "step": 4333
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.00024901530612244895,
+      "loss": 0.2138,
+      "step": 4334
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.000249,
+      "loss": 0.1211,
+      "step": 4335
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.000248984693877551,
+      "loss": 0.1657,
+      "step": 4336
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.00024896938775510204,
+      "loss": 0.2147,
+      "step": 4337
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.000248954081632653,
+      "loss": 0.1207,
+      "step": 4338
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.00024893877551020405,
+      "loss": 0.24,
+      "step": 4339
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.0002489234693877551,
+      "loss": 0.2713,
+      "step": 4340
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.0002489081632653061,
+      "loss": 0.2574,
+      "step": 4341
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.00024889285714285714,
+      "loss": 0.1586,
+      "step": 4342
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002488775510204081,
+      "loss": 0.2074,
+      "step": 4343
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.00024886224489795915,
+      "loss": 0.2048,
+      "step": 4344
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002488469387755102,
+      "loss": 0.2915,
+      "step": 4345
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002488316326530612,
+      "loss": 0.2227,
+      "step": 4346
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002488163265306122,
+      "loss": 0.2278,
+      "step": 4347
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.00024880102040816327,
+      "loss": 0.2105,
+      "step": 4348
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.00024878571428571425,
+      "loss": 0.165,
+      "step": 4349
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002487704081632653,
+      "loss": 0.1895,
+      "step": 4350
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002487551020408163,
+      "loss": 0.1553,
+      "step": 4351
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00024873979591836734,
+      "loss": 0.2863,
+      "step": 4352
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00024872448979591837,
+      "loss": 0.1464,
+      "step": 4353
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00024870918367346934,
+      "loss": 0.1938,
+      "step": 4354
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0002486938775510204,
+      "loss": 0.2008,
+      "step": 4355
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0002486785714285714,
+      "loss": 0.1092,
+      "step": 4356
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00024866326530612244,
+      "loss": 0.1765,
+      "step": 4357
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00024864795918367347,
+      "loss": 0.3121,
+      "step": 4358
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00024863265306122444,
+      "loss": 0.1722,
+      "step": 4359
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0002486173469387755,
+      "loss": 0.1408,
+      "step": 4360
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0002486020408163265,
+      "loss": 0.1941,
+      "step": 4361
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00024858673469387753,
+      "loss": 0.1359,
+      "step": 4362
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00024857142857142857,
+      "loss": 0.1463,
+      "step": 4363
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.0002485561224489796,
+      "loss": 0.2905,
+      "step": 4364
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00024854081632653057,
+      "loss": 0.21,
+      "step": 4365
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.0002485255102040816,
+      "loss": 0.1429,
+      "step": 4366
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00024851020408163263,
+      "loss": 0.2294,
+      "step": 4367
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00024849489795918366,
+      "loss": 0.1356,
+      "step": 4368
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.0002484795918367347,
+      "loss": 0.1396,
+      "step": 4369
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00024846428571428567,
+      "loss": 0.3508,
+      "step": 4370
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.0002484489795918367,
+      "loss": 0.2182,
+      "step": 4371
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00024843367346938773,
+      "loss": 0.2807,
+      "step": 4372
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.00024841836734693876,
+      "loss": 0.162,
+      "step": 4373
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.00024840306122448974,
+      "loss": 0.2207,
+      "step": 4374
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.0002483877551020408,
+      "loss": 0.1092,
+      "step": 4375
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.0002483724489795918,
+      "loss": 0.2694,
+      "step": 4376
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.00024835714285714283,
+      "loss": 0.3698,
+      "step": 4377
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.00024834183673469386,
+      "loss": 0.3205,
+      "step": 4378
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.0002483265306122449,
+      "loss": 0.1353,
+      "step": 4379
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.0002483112244897959,
+      "loss": 0.2322,
+      "step": 4380
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.0002482959183673469,
+      "loss": 0.1335,
+      "step": 4381
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.00024828061224489793,
+      "loss": 0.1636,
+      "step": 4382
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00024826530612244896,
+      "loss": 0.2851,
+      "step": 4383
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00024825,
+      "loss": 0.2502,
+      "step": 4384
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.000248234693877551,
+      "loss": 0.1525,
+      "step": 4385
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.000248219387755102,
+      "loss": 0.2159,
+      "step": 4386
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00024820408163265303,
+      "loss": 0.1527,
+      "step": 4387
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00024818877551020406,
+      "loss": 0.1452,
+      "step": 4388
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.0002481734693877551,
+      "loss": 0.2186,
+      "step": 4389
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00024815816326530606,
+      "loss": 0.2192,
+      "step": 4390
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00024814285714285715,
+      "loss": 0.2008,
+      "step": 4391
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.0002481275510204081,
+      "loss": 0.185,
+      "step": 4392
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.00024811224489795916,
+      "loss": 0.1671,
+      "step": 4393
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002480969387755102,
+      "loss": 0.1623,
+      "step": 4394
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002480816326530612,
+      "loss": 0.203,
+      "step": 4395
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.00024806632653061225,
+      "loss": 0.3008,
+      "step": 4396
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002480510204081632,
+      "loss": 0.182,
+      "step": 4397
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.00024803571428571425,
+      "loss": 0.237,
+      "step": 4398
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002480204081632653,
+      "loss": 0.2419,
+      "step": 4399
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002480051020408163,
+      "loss": 0.1516,
+      "step": 4400
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002479897959183673,
+      "loss": 0.1706,
+      "step": 4401
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002479744897959183,
+      "loss": 0.1764,
+      "step": 4402
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.00024795918367346935,
+      "loss": 0.2485,
+      "step": 4403
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002479438775510204,
+      "loss": 0.2107,
+      "step": 4404
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002479285714285714,
+      "loss": 0.1907,
+      "step": 4405
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00024791326530612244,
+      "loss": 0.1786,
+      "step": 4406
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002478979591836735,
+      "loss": 0.1444,
+      "step": 4407
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00024788265306122445,
+      "loss": 0.2767,
+      "step": 4408
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002478673469387755,
+      "loss": 0.3002,
+      "step": 4409
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002478520408163265,
+      "loss": 0.1693,
+      "step": 4410
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00024783673469387754,
+      "loss": 0.1966,
+      "step": 4411
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002478214285714286,
+      "loss": 0.1992,
+      "step": 4412
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00024780612244897955,
+      "loss": 0.1609,
+      "step": 4413
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0002477908163265306,
+      "loss": 0.217,
+      "step": 4414
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0002477755102040816,
+      "loss": 0.2108,
+      "step": 4415
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.00024776020408163264,
+      "loss": 0.143,
+      "step": 4416
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0002477448979591836,
+      "loss": 0.2618,
+      "step": 4417
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0002477295918367347,
+      "loss": 0.1282,
+      "step": 4418
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0002477142857142857,
+      "loss": 0.1284,
+      "step": 4419
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0002476989795918367,
+      "loss": 0.2986,
+      "step": 4420
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.00024768367346938774,
+      "loss": 0.3232,
+      "step": 4421
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.00024766836734693877,
+      "loss": 0.3591,
+      "step": 4422
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0002476530612244898,
+      "loss": 0.1676,
+      "step": 4423
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.0002476377551020408,
+      "loss": 0.1941,
+      "step": 4424
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.0002476224489795918,
+      "loss": 0.2337,
+      "step": 4425
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00024760714285714284,
+      "loss": 0.3148,
+      "step": 4426
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00024759183673469387,
+      "loss": 0.3915,
+      "step": 4427
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00024757653061224485,
+      "loss": 0.2127,
+      "step": 4428
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.0002475612244897959,
+      "loss": 0.1372,
+      "step": 4429
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.0002475459183673469,
+      "loss": 0.1913,
+      "step": 4430
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00024753061224489794,
+      "loss": 0.1824,
+      "step": 4431
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00024751530612244897,
+      "loss": 0.1656,
+      "step": 4432
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00024749999999999994,
+      "loss": 0.2693,
+      "step": 4433
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00024748469387755103,
+      "loss": 0.2369,
+      "step": 4434
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.000247469387755102,
+      "loss": 0.1451,
+      "step": 4435
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00024745408163265304,
+      "loss": 0.1721,
+      "step": 4436
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00024743877551020407,
+      "loss": 0.1357,
+      "step": 4437
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.0002474234693877551,
+      "loss": 0.1337,
+      "step": 4438
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00024740816326530613,
+      "loss": 0.1978,
+      "step": 4439
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.0002473928571428571,
+      "loss": 0.2183,
+      "step": 4440
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00024737755102040813,
+      "loss": 0.1612,
+      "step": 4441
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00024736224489795917,
+      "loss": 0.1318,
+      "step": 4442
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.0002473469387755102,
+      "loss": 0.18,
+      "step": 4443
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00024733163265306117,
+      "loss": 0.2175,
+      "step": 4444
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0002473163265306122,
+      "loss": 0.1667,
+      "step": 4445
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.00024730102040816323,
+      "loss": 0.2041,
+      "step": 4446
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.00024728571428571426,
+      "loss": 0.1513,
+      "step": 4447
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0002472704081632653,
+      "loss": 0.1846,
+      "step": 4448
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.00024725510204081627,
+      "loss": 0.1589,
+      "step": 4449
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.00024723979591836736,
+      "loss": 0.1596,
+      "step": 4450
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.00024722448979591833,
+      "loss": 0.1482,
+      "step": 4451
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.00024720918367346936,
+      "loss": 0.321,
+      "step": 4452
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0002471938775510204,
+      "loss": 0.1885,
+      "step": 4453
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0002471785714285714,
+      "loss": 0.2662,
+      "step": 4454
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0002471632653061224,
+      "loss": 0.1598,
+      "step": 4455
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00024714795918367343,
+      "loss": 0.1811,
+      "step": 4456
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00024713265306122446,
+      "loss": 0.088,
+      "step": 4457
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0002471173469387755,
+      "loss": 0.3608,
+      "step": 4458
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0002471020408163265,
+      "loss": 0.1401,
+      "step": 4459
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0002470867346938775,
+      "loss": 0.1852,
+      "step": 4460
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0002470714285714286,
+      "loss": 0.1253,
+      "step": 4461
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00024705612244897956,
+      "loss": 0.13,
+      "step": 4462
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0002470408163265306,
+      "loss": 0.1583,
+      "step": 4463
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0002470255102040816,
+      "loss": 0.4244,
+      "step": 4464
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00024701020408163265,
+      "loss": 0.2183,
+      "step": 4465
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0002469948979591837,
+      "loss": 0.2326,
+      "step": 4466
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00024697959183673466,
+      "loss": 0.1858,
+      "step": 4467
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0002469642857142857,
+      "loss": 0.1395,
+      "step": 4468
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0002469489795918367,
+      "loss": 0.1706,
+      "step": 4469
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00024693367346938775,
+      "loss": 0.2417,
+      "step": 4470
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0002469183673469387,
+      "loss": 0.1827,
+      "step": 4471
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00024690306122448976,
+      "loss": 0.1967,
+      "step": 4472
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0002468877551020408,
+      "loss": 0.2788,
+      "step": 4473
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0002468724489795918,
+      "loss": 0.2103,
+      "step": 4474
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00024685714285714285,
+      "loss": 0.1724,
+      "step": 4475
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.0002468418367346938,
+      "loss": 0.1569,
+      "step": 4476
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.0002468265306122449,
+      "loss": 0.4448,
+      "step": 4477
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.0002468112244897959,
+      "loss": 0.197,
+      "step": 4478
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.0002467959183673469,
+      "loss": 0.1559,
+      "step": 4479
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.00024678061224489795,
+      "loss": 0.1535,
+      "step": 4480
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.000246765306122449,
+      "loss": 0.1247,
+      "step": 4481
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.00024675,
+      "loss": 0.1986,
+      "step": 4482
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.000246734693877551,
+      "loss": 0.3513,
+      "step": 4483
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.000246719387755102,
+      "loss": 0.2205,
+      "step": 4484
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.00024670408163265304,
+      "loss": 0.2581,
+      "step": 4485
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0002466887755102041,
+      "loss": 0.1515,
+      "step": 4486
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.00024667346938775505,
+      "loss": 0.2021,
+      "step": 4487
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0002466581632653061,
+      "loss": 0.1201,
+      "step": 4488
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0002466428571428571,
+      "loss": 0.3562,
+      "step": 4489
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.00024662755102040814,
+      "loss": 0.2219,
+      "step": 4490
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0002466122448979592,
+      "loss": 0.1987,
+      "step": 4491
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.00024659693877551015,
+      "loss": 0.2038,
+      "step": 4492
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.00024658163265306124,
+      "loss": 0.1773,
+      "step": 4493
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0002465663265306122,
+      "loss": 0.167,
+      "step": 4494
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.00024655102040816324,
+      "loss": 0.2816,
+      "step": 4495
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00024653571428571427,
+      "loss": 0.2155,
+      "step": 4496
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.0002465204081632653,
+      "loss": 0.1555,
+      "step": 4497
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.0002465051020408163,
+      "loss": 0.1821,
+      "step": 4498
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.0002464897959183673,
+      "loss": 0.1265,
+      "step": 4499
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00024647448979591834,
+      "loss": 0.1706,
+      "step": 4500
+    },
+    {
+      "epoch": 4.37,
+      "eval_cer": 0.2688255613877664,
+      "eval_loss": 0.9190245270729065,
+      "eval_runtime": 226.4899,
+      "eval_samples_per_second": 38.916,
+      "eval_steps_per_second": 4.866,
+      "eval_wer": 0.6788001625855496,
+      "step": 4500
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00024645918367346937,
+      "loss": 0.2646,
+      "step": 4501
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.0002464438775510204,
+      "loss": 0.1723,
+      "step": 4502
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.0002464285714285714,
+      "loss": 0.2505,
+      "step": 4503
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00024641326530612246,
+      "loss": 0.1969,
+      "step": 4504
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00024639795918367344,
+      "loss": 0.2405,
+      "step": 4505
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00024638265306122447,
+      "loss": 0.1561,
+      "step": 4506
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.0002463673469387755,
+      "loss": 0.1759,
+      "step": 4507
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00024635204081632653,
+      "loss": 0.2895,
+      "step": 4508
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00024633673469387756,
+      "loss": 0.2451,
+      "step": 4509
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00024632142857142854,
+      "loss": 0.222,
+      "step": 4510
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00024630612244897957,
+      "loss": 0.171,
+      "step": 4511
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.0002462908163265306,
+      "loss": 0.2287,
+      "step": 4512
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00024627551020408163,
+      "loss": 0.2014,
+      "step": 4513
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.0002462602040816326,
+      "loss": 0.1286,
+      "step": 4514
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00024624489795918364,
+      "loss": 0.2291,
+      "step": 4515
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00024622959183673467,
+      "loss": 0.1564,
+      "step": 4516
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.0002462142857142857,
+      "loss": 0.1806,
+      "step": 4517
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00024619897959183673,
+      "loss": 0.2083,
+      "step": 4518
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.0002461836734693877,
+      "loss": 0.1324,
+      "step": 4519
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.0002461683673469388,
+      "loss": 0.1982,
+      "step": 4520
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00024615306122448977,
+      "loss": 0.2273,
+      "step": 4521
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.0002461377551020408,
+      "loss": 0.2063,
+      "step": 4522
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.0002461224489795918,
+      "loss": 0.2307,
+      "step": 4523
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00024610714285714286,
+      "loss": 0.2151,
+      "step": 4524
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00024609183673469383,
+      "loss": 0.1334,
+      "step": 4525
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00024607653061224486,
+      "loss": 0.183,
+      "step": 4526
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.0002460612244897959,
+      "loss": 0.3283,
+      "step": 4527
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.0002460459183673469,
+      "loss": 0.2353,
+      "step": 4528
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.00024603061224489796,
+      "loss": 0.1216,
+      "step": 4529
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.00024601530612244893,
+      "loss": 0.1625,
+      "step": 4530
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.00024599999999999996,
+      "loss": 0.2303,
+      "step": 4531
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.000245984693877551,
+      "loss": 0.1382,
+      "step": 4532
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.000245969387755102,
+      "loss": 0.5514,
+      "step": 4533
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.00024595408163265305,
+      "loss": 0.2109,
+      "step": 4534
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.00024593877551020403,
+      "loss": 0.1448,
+      "step": 4535
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.0002459234693877551,
+      "loss": 0.1334,
+      "step": 4536
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.0002459081632653061,
+      "loss": 0.141,
+      "step": 4537
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0002458928571428571,
+      "loss": 0.1365,
+      "step": 4538
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.00024587755102040815,
+      "loss": 0.2905,
+      "step": 4539
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0002458622448979592,
+      "loss": 0.2881,
+      "step": 4540
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.00024584693877551016,
+      "loss": 0.1537,
+      "step": 4541
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0002458316326530612,
+      "loss": 0.1788,
+      "step": 4542
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0002458163265306122,
+      "loss": 0.1629,
+      "step": 4543
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.00024580102040816325,
+      "loss": 0.1405,
+      "step": 4544
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0002457857142857143,
+      "loss": 0.172,
+      "step": 4545
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.00024577040816326526,
+      "loss": 0.2016,
+      "step": 4546
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.00024575510204081634,
+      "loss": 0.179,
+      "step": 4547
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0002457397959183673,
+      "loss": 0.2334,
+      "step": 4548
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.00024572448979591835,
+      "loss": 0.123,
+      "step": 4549
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0002457091836734694,
+      "loss": 0.1745,
+      "step": 4550
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0002456938775510204,
+      "loss": 0.2056,
+      "step": 4551
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0002456785714285714,
+      "loss": 0.2032,
+      "step": 4552
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0002456632653061224,
+      "loss": 0.1352,
+      "step": 4553
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.00024564795918367345,
+      "loss": 0.1796,
+      "step": 4554
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0002456326530612245,
+      "loss": 0.1222,
+      "step": 4555
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0002456173469387755,
+      "loss": 0.1445,
+      "step": 4556
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0002456020408163265,
+      "loss": 0.1481,
+      "step": 4557
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0002455867346938775,
+      "loss": 0.2332,
+      "step": 4558
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.00024557142857142855,
+      "loss": 0.2581,
+      "step": 4559
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0002455561224489796,
+      "loss": 0.198,
+      "step": 4560
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0002455408163265306,
+      "loss": 0.2274,
+      "step": 4561
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0002455255102040816,
+      "loss": 0.125,
+      "step": 4562
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.00024551020408163267,
+      "loss": 0.1666,
+      "step": 4563
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.00024549489795918364,
+      "loss": 0.2358,
+      "step": 4564
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0002454795918367347,
+      "loss": 0.2183,
+      "step": 4565
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0002454642857142857,
+      "loss": 0.2173,
+      "step": 4566
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.00024544897959183674,
+      "loss": 0.1859,
+      "step": 4567
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0002454336734693877,
+      "loss": 0.2408,
+      "step": 4568
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00024541836734693874,
+      "loss": 0.0992,
+      "step": 4569
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.0002454030612244898,
+      "loss": 0.1562,
+      "step": 4570
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.0002453877551020408,
+      "loss": 0.1992,
+      "step": 4571
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00024537244897959184,
+      "loss": 0.1934,
+      "step": 4572
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.0002453571428571428,
+      "loss": 0.1122,
+      "step": 4573
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00024534183673469384,
+      "loss": 0.1991,
+      "step": 4574
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00024532653061224487,
+      "loss": 0.1303,
+      "step": 4575
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.0002453112244897959,
+      "loss": 0.2366,
+      "step": 4576
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00024529591836734693,
+      "loss": 0.2081,
+      "step": 4577
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.0002452806122448979,
+      "loss": 0.2543,
+      "step": 4578
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.000245265306122449,
+      "loss": 0.1892,
+      "step": 4579
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.00024524999999999997,
+      "loss": 0.2041,
+      "step": 4580
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.000245234693877551,
+      "loss": 0.2531,
+      "step": 4581
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.000245219387755102,
+      "loss": 0.2179,
+      "step": 4582
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.00024520408163265306,
+      "loss": 0.2591,
+      "step": 4583
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.00024518877551020404,
+      "loss": 0.1526,
+      "step": 4584
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.00024517346938775507,
+      "loss": 0.1619,
+      "step": 4585
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.0002451581632653061,
+      "loss": 0.1783,
+      "step": 4586
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.00024514285714285713,
+      "loss": 0.2053,
+      "step": 4587
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.00024512755102040816,
+      "loss": 0.1148,
+      "step": 4588
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00024511224489795914,
+      "loss": 0.2427,
+      "step": 4589
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.0002450969387755102,
+      "loss": 0.162,
+      "step": 4590
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.0002450816326530612,
+      "loss": 0.2297,
+      "step": 4591
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00024506632653061223,
+      "loss": 0.1825,
+      "step": 4592
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00024505102040816326,
+      "loss": 0.2005,
+      "step": 4593
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.0002450357142857143,
+      "loss": 0.1765,
+      "step": 4594
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00024502040816326527,
+      "loss": 0.2072,
+      "step": 4595
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.0002450051020408163,
+      "loss": 0.1735,
+      "step": 4596
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00024498979591836733,
+      "loss": 0.2259,
+      "step": 4597
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00024497448979591836,
+      "loss": 0.1429,
+      "step": 4598
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.0002449591836734694,
+      "loss": 0.1631,
+      "step": 4599
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00024494387755102037,
+      "loss": 0.1511,
+      "step": 4600
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.0002449285714285714,
+      "loss": 0.1153,
+      "step": 4601
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.0002449132653061224,
+      "loss": 0.2135,
+      "step": 4602
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00024489795918367346,
+      "loss": 0.1558,
+      "step": 4603
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.0002448826530612245,
+      "loss": 0.1463,
+      "step": 4604
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00024486734693877546,
+      "loss": 0.1347,
+      "step": 4605
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00024485204081632655,
+      "loss": 0.1751,
+      "step": 4606
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.0002448367346938775,
+      "loss": 0.1302,
+      "step": 4607
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00024482142857142856,
+      "loss": 0.4522,
+      "step": 4608
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00024480612244897953,
+      "loss": 0.2929,
+      "step": 4609
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0002447908163265306,
+      "loss": 0.1795,
+      "step": 4610
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0002447755102040816,
+      "loss": 0.1899,
+      "step": 4611
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0002447602040816326,
+      "loss": 0.1821,
+      "step": 4612
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.00024474489795918365,
+      "loss": 0.1186,
+      "step": 4613
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0002447295918367347,
+      "loss": 0.1957,
+      "step": 4614
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0002447142857142857,
+      "loss": 0.2296,
+      "step": 4615
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0002446989795918367,
+      "loss": 0.2808,
+      "step": 4616
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0002446836734693877,
+      "loss": 0.2178,
+      "step": 4617
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.00024466836734693875,
+      "loss": 0.2566,
+      "step": 4618
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0002446530612244898,
+      "loss": 0.158,
+      "step": 4619
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002446377551020408,
+      "loss": 0.216,
+      "step": 4620
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002446224489795918,
+      "loss": 0.2267,
+      "step": 4621
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002446071428571428,
+      "loss": 0.2222,
+      "step": 4622
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.00024459183673469385,
+      "loss": 0.2343,
+      "step": 4623
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002445765306122449,
+      "loss": 0.2526,
+      "step": 4624
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.00024456122448979586,
+      "loss": 0.2503,
+      "step": 4625
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.00024454591836734694,
+      "loss": 0.1796,
+      "step": 4626
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002445306122448979,
+      "loss": 0.1999,
+      "step": 4627
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.00024451530612244895,
+      "loss": 0.3079,
+      "step": 4628
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002445,
+      "loss": 0.1439,
+      "step": 4629
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.000244484693877551,
+      "loss": 0.1498,
+      "step": 4630
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00024446938775510204,
+      "loss": 0.1701,
+      "step": 4631
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.000244454081632653,
+      "loss": 0.2191,
+      "step": 4632
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0002444387755102041,
+      "loss": 0.3959,
+      "step": 4633
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0002444234693877551,
+      "loss": 0.2095,
+      "step": 4634
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0002444081632653061,
+      "loss": 0.2825,
+      "step": 4635
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00024439285714285714,
+      "loss": 0.154,
+      "step": 4636
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00024437755102040817,
+      "loss": 0.1684,
+      "step": 4637
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00024436224489795915,
+      "loss": 0.1866,
+      "step": 4638
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0002443469387755102,
+      "loss": 0.3117,
+      "step": 4639
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0002443316326530612,
+      "loss": 0.2129,
+      "step": 4640
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.00024431632653061224,
+      "loss": 0.2161,
+      "step": 4641
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.00024430102040816327,
+      "loss": 0.1293,
+      "step": 4642
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.00024428571428571424,
+      "loss": 0.1969,
+      "step": 4643
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0002442704081632653,
+      "loss": 0.1677,
+      "step": 4644
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0002442551020408163,
+      "loss": 0.2535,
+      "step": 4645
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.00024423979591836734,
+      "loss": 0.2576,
+      "step": 4646
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.00024422448979591837,
+      "loss": 0.2193,
+      "step": 4647
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.00024420918367346934,
+      "loss": 0.1375,
+      "step": 4648
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0002441938775510204,
+      "loss": 0.1459,
+      "step": 4649
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0002441785714285714,
+      "loss": 0.1497,
+      "step": 4650
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.00024416326530612244,
+      "loss": 0.1606,
+      "step": 4651
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.0002441479591836734,
+      "loss": 0.3181,
+      "step": 4652
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.00024413265306122447,
+      "loss": 0.2118,
+      "step": 4653
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.0002441173469387755,
+      "loss": 0.1817,
+      "step": 4654
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.0002441020408163265,
+      "loss": 0.1561,
+      "step": 4655
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.0002440867346938775,
+      "loss": 0.1455,
+      "step": 4656
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.00024407142857142856,
+      "loss": 0.1912,
+      "step": 4657
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.00024405612244897957,
+      "loss": 0.1842,
+      "step": 4658
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.0002440408163265306,
+      "loss": 0.2175,
+      "step": 4659
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.0002440255102040816,
+      "loss": 0.1636,
+      "step": 4660
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024401020408163263,
+      "loss": 0.1898,
+      "step": 4661
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024399489795918366,
+      "loss": 0.2215,
+      "step": 4662
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024397959183673467,
+      "loss": 0.1276,
+      "step": 4663
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024396428571428567,
+      "loss": 0.3034,
+      "step": 4664
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024394897959183673,
+      "loss": 0.2502,
+      "step": 4665
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024393367346938773,
+      "loss": 0.1816,
+      "step": 4666
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024391836734693876,
+      "loss": 0.2081,
+      "step": 4667
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024390306122448976,
+      "loss": 0.1451,
+      "step": 4668
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.0002438877551020408,
+      "loss": 0.1886,
+      "step": 4669
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024387244897959183,
+      "loss": 0.2259,
+      "step": 4670
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00024385714285714283,
+      "loss": 0.2176,
+      "step": 4671
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00024384183673469383,
+      "loss": 0.16,
+      "step": 4672
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.0002438265306122449,
+      "loss": 0.2167,
+      "step": 4673
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.0002438112244897959,
+      "loss": 0.2595,
+      "step": 4674
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.0002437959183673469,
+      "loss": 0.1108,
+      "step": 4675
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00024378061224489793,
+      "loss": 0.2088,
+      "step": 4676
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00024376530612244896,
+      "loss": 0.2992,
+      "step": 4677
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00024375,
+      "loss": 0.1953,
+      "step": 4678
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.000243734693877551,
+      "loss": 0.132,
+      "step": 4679
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00024371938775510202,
+      "loss": 0.2227,
+      "step": 4680
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00024370408163265305,
+      "loss": 0.1493,
+      "step": 4681
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024368877551020406,
+      "loss": 0.1919,
+      "step": 4682
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024367346938775506,
+      "loss": 0.2795,
+      "step": 4683
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024365816326530612,
+      "loss": 0.172,
+      "step": 4684
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024364285714285712,
+      "loss": 0.1455,
+      "step": 4685
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024362755102040815,
+      "loss": 0.1866,
+      "step": 4686
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024361224489795916,
+      "loss": 0.1379,
+      "step": 4687
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024359693877551019,
+      "loss": 0.1742,
+      "step": 4688
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024358163265306122,
+      "loss": 0.1461,
+      "step": 4689
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024356632653061222,
+      "loss": 0.2057,
+      "step": 4690
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00024355102040816322,
+      "loss": 0.2129,
+      "step": 4691
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024353571428571428,
+      "loss": 0.1859,
+      "step": 4692
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024352040816326528,
+      "loss": 0.1631,
+      "step": 4693
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024350510204081631,
+      "loss": 0.1614,
+      "step": 4694
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024348979591836732,
+      "loss": 0.2806,
+      "step": 4695
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024347448979591835,
+      "loss": 0.2201,
+      "step": 4696
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024345918367346938,
+      "loss": 0.1869,
+      "step": 4697
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024344387755102038,
+      "loss": 0.2117,
+      "step": 4698
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024342857142857139,
+      "loss": 0.2495,
+      "step": 4699
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024341326530612244,
+      "loss": 0.165,
+      "step": 4700
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00024339795918367345,
+      "loss": 0.1424,
+      "step": 4701
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00024338265306122445,
+      "loss": 0.3047,
+      "step": 4702
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00024336734693877548,
+      "loss": 0.2435,
+      "step": 4703
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.0002433520408163265,
+      "loss": 0.2129,
+      "step": 4704
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00024333673469387754,
+      "loss": 0.2392,
+      "step": 4705
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00024332142857142855,
+      "loss": 0.1759,
+      "step": 4706
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00024330612244897955,
+      "loss": 0.1813,
+      "step": 4707
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.0002432908163265306,
+      "loss": 0.203,
+      "step": 4708
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.0002432755102040816,
+      "loss": 0.3297,
+      "step": 4709
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00024326020408163261,
+      "loss": 0.2783,
+      "step": 4710
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00024324489795918364,
+      "loss": 0.1727,
+      "step": 4711
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00024322959183673468,
+      "loss": 0.133,
+      "step": 4712
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.0002432142857142857,
+      "loss": 0.1804,
+      "step": 4713
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.0002431989795918367,
+      "loss": 0.2395,
+      "step": 4714
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.0002431836734693877,
+      "loss": 0.2018,
+      "step": 4715
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00024316836734693877,
+      "loss": 0.2312,
+      "step": 4716
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00024315306122448977,
+      "loss": 0.2409,
+      "step": 4717
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00024313775510204078,
+      "loss": 0.1753,
+      "step": 4718
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.0002431224489795918,
+      "loss": 0.1354,
+      "step": 4719
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00024310714285714284,
+      "loss": 0.3762,
+      "step": 4720
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00024309183673469387,
+      "loss": 0.2454,
+      "step": 4721
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00024307653061224487,
+      "loss": 0.1662,
+      "step": 4722
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00024306122448979588,
+      "loss": 0.209,
+      "step": 4723
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00024304591836734693,
+      "loss": 0.1764,
+      "step": 4724
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00024303061224489794,
+      "loss": 0.2187,
+      "step": 4725
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00024301530612244894,
+      "loss": 0.2204,
+      "step": 4726
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.000243,
+      "loss": 0.244,
+      "step": 4727
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.000242984693877551,
+      "loss": 0.1992,
+      "step": 4728
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00024296938775510203,
+      "loss": 0.2519,
+      "step": 4729
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00024295408163265304,
+      "loss": 0.2086,
+      "step": 4730
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00024293877551020407,
+      "loss": 0.1392,
+      "step": 4731
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.0002429234693877551,
+      "loss": 0.1523,
+      "step": 4732
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.0002429081632653061,
+      "loss": 0.2518,
+      "step": 4733
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.0002428928571428571,
+      "loss": 0.1576,
+      "step": 4734
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00024287755102040816,
+      "loss": 0.2313,
+      "step": 4735
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00024286224489795916,
+      "loss": 0.1955,
+      "step": 4736
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00024284693877551017,
+      "loss": 0.2594,
+      "step": 4737
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.0002428316326530612,
+      "loss": 0.1193,
+      "step": 4738
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00024281632653061223,
+      "loss": 0.294,
+      "step": 4739
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00024280102040816326,
+      "loss": 0.3482,
+      "step": 4740
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00024278571428571426,
+      "loss": 0.1457,
+      "step": 4741
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00024277040816326527,
+      "loss": 0.239,
+      "step": 4742
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00024275510204081632,
+      "loss": 0.1466,
+      "step": 4743
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.00024273979591836733,
+      "loss": 0.1696,
+      "step": 4744
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.00024272448979591833,
+      "loss": 0.2364,
+      "step": 4745
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.00024270918367346936,
+      "loss": 0.2044,
+      "step": 4746
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0002426938775510204,
+      "loss": 0.1692,
+      "step": 4747
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.00024267857142857142,
+      "loss": 0.1667,
+      "step": 4748
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.00024266326530612243,
+      "loss": 0.2004,
+      "step": 4749
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.00024264795918367343,
+      "loss": 0.1848,
+      "step": 4750
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0002426326530612245,
+      "loss": 0.2069,
+      "step": 4751
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0002426173469387755,
+      "loss": 0.3438,
+      "step": 4752
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0002426020408163265,
+      "loss": 0.2061,
+      "step": 4753
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00024258673469387752,
+      "loss": 0.2108,
+      "step": 4754
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00024257142857142855,
+      "loss": 0.1358,
+      "step": 4755
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00024255612244897959,
+      "loss": 0.1726,
+      "step": 4756
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0002425408163265306,
+      "loss": 0.1642,
+      "step": 4757
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0002425255102040816,
+      "loss": 0.2533,
+      "step": 4758
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00024251020408163265,
+      "loss": 0.1744,
+      "step": 4759
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00024249489795918365,
+      "loss": 0.2351,
+      "step": 4760
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00024247959183673466,
+      "loss": 0.2597,
+      "step": 4761
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0002424642857142857,
+      "loss": 0.1378,
+      "step": 4762
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00024244897959183672,
+      "loss": 0.1865,
+      "step": 4763
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024243367346938772,
+      "loss": 0.3067,
+      "step": 4764
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024241836734693875,
+      "loss": 0.3579,
+      "step": 4765
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024240306122448976,
+      "loss": 0.1417,
+      "step": 4766
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.0002423877551020408,
+      "loss": 0.163,
+      "step": 4767
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024237244897959182,
+      "loss": 0.1698,
+      "step": 4768
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024235714285714282,
+      "loss": 0.1477,
+      "step": 4769
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024234183673469388,
+      "loss": 0.2465,
+      "step": 4770
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024232653061224488,
+      "loss": 0.2955,
+      "step": 4771
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024231122448979588,
+      "loss": 0.1779,
+      "step": 4772
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024229591836734691,
+      "loss": 0.172,
+      "step": 4773
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00024228061224489795,
+      "loss": 0.15,
+      "step": 4774
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024226530612244898,
+      "loss": 0.1729,
+      "step": 4775
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024224999999999998,
+      "loss": 0.1497,
+      "step": 4776
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024223469387755098,
+      "loss": 0.2047,
+      "step": 4777
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024221938775510204,
+      "loss": 0.1687,
+      "step": 4778
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024220408163265304,
+      "loss": 0.2008,
+      "step": 4779
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024218877551020405,
+      "loss": 0.2211,
+      "step": 4780
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024217346938775508,
+      "loss": 0.1507,
+      "step": 4781
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.0002421581632653061,
+      "loss": 0.1634,
+      "step": 4782
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024214285714285714,
+      "loss": 0.3945,
+      "step": 4783
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00024212755102040814,
+      "loss": 0.2785,
+      "step": 4784
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.00024211224489795915,
+      "loss": 0.1788,
+      "step": 4785
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0002420969387755102,
+      "loss": 0.165,
+      "step": 4786
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0002420816326530612,
+      "loss": 0.2157,
+      "step": 4787
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0002420663265306122,
+      "loss": 0.1334,
+      "step": 4788
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.00024205102040816324,
+      "loss": 0.3475,
+      "step": 4789
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.00024203571428571427,
+      "loss": 0.2352,
+      "step": 4790
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0002420204081632653,
+      "loss": 0.2123,
+      "step": 4791
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0002420051020408163,
+      "loss": 0.2403,
+      "step": 4792
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0002419897959183673,
+      "loss": 0.1648,
+      "step": 4793
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.00024197448979591837,
+      "loss": 0.2314,
+      "step": 4794
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024195918367346937,
+      "loss": 0.1715,
+      "step": 4795
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024194387755102037,
+      "loss": 0.1741,
+      "step": 4796
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.0002419285714285714,
+      "loss": 0.2353,
+      "step": 4797
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024191326530612243,
+      "loss": 0.2436,
+      "step": 4798
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024189795918367344,
+      "loss": 0.2,
+      "step": 4799
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024188265306122447,
+      "loss": 0.179,
+      "step": 4800
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024186734693877547,
+      "loss": 0.2123,
+      "step": 4801
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024185204081632653,
+      "loss": 0.392,
+      "step": 4802
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024183673469387753,
+      "loss": 0.194,
+      "step": 4803
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00024182142857142854,
+      "loss": 0.1575,
+      "step": 4804
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00024180612244897957,
+      "loss": 0.2214,
+      "step": 4805
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0002417908163265306,
+      "loss": 0.1894,
+      "step": 4806
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0002417755102040816,
+      "loss": 0.166,
+      "step": 4807
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00024176020408163263,
+      "loss": 0.2948,
+      "step": 4808
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00024174489795918364,
+      "loss": 0.2522,
+      "step": 4809
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0002417295918367347,
+      "loss": 0.1749,
+      "step": 4810
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0002417142857142857,
+      "loss": 0.2948,
+      "step": 4811
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0002416989795918367,
+      "loss": 0.1672,
+      "step": 4812
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00024168367346938773,
+      "loss": 0.1993,
+      "step": 4813
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00024166836734693876,
+      "loss": 0.2676,
+      "step": 4814
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00024165306122448976,
+      "loss": 0.3003,
+      "step": 4815
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0002416377551020408,
+      "loss": 0.1435,
+      "step": 4816
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00024162244897959183,
+      "loss": 0.2031,
+      "step": 4817
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00024160714285714286,
+      "loss": 0.2471,
+      "step": 4818
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00024159183673469386,
+      "loss": 0.1246,
+      "step": 4819
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00024157653061224486,
+      "loss": 0.4699,
+      "step": 4820
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00024156122448979592,
+      "loss": 0.3831,
+      "step": 4821
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00024154591836734692,
+      "loss": 0.1606,
+      "step": 4822
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00024153061224489793,
+      "loss": 0.2403,
+      "step": 4823
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00024151530612244896,
+      "loss": 0.1843,
+      "step": 4824
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0002415,
+      "loss": 0.1134,
+      "step": 4825
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00024148469387755102,
+      "loss": 0.1376,
+      "step": 4826
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00024146938775510202,
+      "loss": 0.1793,
+      "step": 4827
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00024145408163265303,
+      "loss": 0.2239,
+      "step": 4828
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00024143877551020408,
+      "loss": 0.1718,
+      "step": 4829
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.0002414234693877551,
+      "loss": 0.1616,
+      "step": 4830
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.0002414081632653061,
+      "loss": 0.103,
+      "step": 4831
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00024139285714285712,
+      "loss": 0.2872,
+      "step": 4832
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00024137755102040815,
+      "loss": 0.2328,
+      "step": 4833
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00024136224489795915,
+      "loss": 0.2049,
+      "step": 4834
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00024134693877551019,
+      "loss": 0.2438,
+      "step": 4835
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.0002413316326530612,
+      "loss": 0.1546,
+      "step": 4836
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00024131632653061225,
+      "loss": 0.2063,
+      "step": 4837
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00024130102040816325,
+      "loss": 0.2503,
+      "step": 4838
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00024128571428571425,
+      "loss": 0.2219,
+      "step": 4839
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00024127040816326528,
+      "loss": 0.2916,
+      "step": 4840
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00024125510204081631,
+      "loss": 0.1835,
+      "step": 4841
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00024123979591836732,
+      "loss": 0.2063,
+      "step": 4842
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00024122448979591835,
+      "loss": 0.1695,
+      "step": 4843
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00024120918367346935,
+      "loss": 0.1801,
+      "step": 4844
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.0002411938775510204,
+      "loss": 0.2408,
+      "step": 4845
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.0002411785714285714,
+      "loss": 0.2332,
+      "step": 4846
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024116326530612242,
+      "loss": 0.2767,
+      "step": 4847
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024114795918367345,
+      "loss": 0.1136,
+      "step": 4848
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024113265306122448,
+      "loss": 0.2691,
+      "step": 4849
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024111734693877548,
+      "loss": 0.1519,
+      "step": 4850
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0002411020408163265,
+      "loss": 0.1738,
+      "step": 4851
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024108673469387751,
+      "loss": 0.2097,
+      "step": 4852
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024107142857142857,
+      "loss": 0.1938,
+      "step": 4853
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024105612244897958,
+      "loss": 0.1769,
+      "step": 4854
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024104081632653058,
+      "loss": 0.2199,
+      "step": 4855
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00024102551020408158,
+      "loss": 0.1293,
+      "step": 4856
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00024101020408163264,
+      "loss": 0.3035,
+      "step": 4857
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00024099489795918364,
+      "loss": 0.2153,
+      "step": 4858
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00024097959183673467,
+      "loss": 0.2297,
+      "step": 4859
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0002409642857142857,
+      "loss": 0.2059,
+      "step": 4860
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0002409489795918367,
+      "loss": 0.2209,
+      "step": 4861
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00024093367346938774,
+      "loss": 0.149,
+      "step": 4862
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00024091836734693874,
+      "loss": 0.2031,
+      "step": 4863
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0002409030612244898,
+      "loss": 0.4279,
+      "step": 4864
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0002408877551020408,
+      "loss": 0.1487,
+      "step": 4865
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0002408724489795918,
+      "loss": 0.1517,
+      "step": 4866
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00024085714285714284,
+      "loss": 0.1355,
+      "step": 4867
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00024084183673469387,
+      "loss": 0.1079,
+      "step": 4868
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00024082653061224487,
+      "loss": 0.1612,
+      "step": 4869
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.0002408112244897959,
+      "loss": 0.1329,
+      "step": 4870
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.0002407959183673469,
+      "loss": 0.257,
+      "step": 4871
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00024078061224489796,
+      "loss": 0.1791,
+      "step": 4872
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00024076530612244897,
+      "loss": 0.2192,
+      "step": 4873
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00024074999999999997,
+      "loss": 0.2002,
+      "step": 4874
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.000240734693877551,
+      "loss": 0.1158,
+      "step": 4875
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00024071938775510203,
+      "loss": 0.1402,
+      "step": 4876
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00024070408163265303,
+      "loss": 0.2425,
+      "step": 4877
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00024068877551020407,
+      "loss": 0.1824,
+      "step": 4878
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00024067346938775507,
+      "loss": 0.1664,
+      "step": 4879
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00024065816326530613,
+      "loss": 0.1458,
+      "step": 4880
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00024064285714285713,
+      "loss": 0.1958,
+      "step": 4881
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00024062755102040813,
+      "loss": 0.1819,
+      "step": 4882
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00024061224489795914,
+      "loss": 0.3185,
+      "step": 4883
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.0002405969387755102,
+      "loss": 0.241,
+      "step": 4884
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.0002405816326530612,
+      "loss": 0.1652,
+      "step": 4885
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00024056632653061223,
+      "loss": 0.2506,
+      "step": 4886
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00024055102040816323,
+      "loss": 0.2022,
+      "step": 4887
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0002405357142857143,
+      "loss": 0.2013,
+      "step": 4888
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0002405204081632653,
+      "loss": 0.2911,
+      "step": 4889
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0002405051020408163,
+      "loss": 0.1835,
+      "step": 4890
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0002404897959183673,
+      "loss": 0.1843,
+      "step": 4891
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.00024047448979591836,
+      "loss": 0.1392,
+      "step": 4892
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.00024045918367346936,
+      "loss": 0.2657,
+      "step": 4893
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0002404438775510204,
+      "loss": 0.1842,
+      "step": 4894
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0002404285714285714,
+      "loss": 0.2351,
+      "step": 4895
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.00024041326530612243,
+      "loss": 0.2677,
+      "step": 4896
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.00024039795918367346,
+      "loss": 0.1761,
+      "step": 4897
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024038265306122446,
+      "loss": 0.1759,
+      "step": 4898
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024036734693877546,
+      "loss": 0.127,
+      "step": 4899
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024035204081632652,
+      "loss": 0.121,
+      "step": 4900
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024033673469387752,
+      "loss": 0.1711,
+      "step": 4901
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024032142857142855,
+      "loss": 0.2884,
+      "step": 4902
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024030612244897956,
+      "loss": 0.1653,
+      "step": 4903
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.0002402908163265306,
+      "loss": 0.2142,
+      "step": 4904
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024027551020408162,
+      "loss": 0.1982,
+      "step": 4905
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024026020408163262,
+      "loss": 0.1591,
+      "step": 4906
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00024024489795918368,
+      "loss": 0.2432,
+      "step": 4907
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024022959183673468,
+      "loss": 0.3221,
+      "step": 4908
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.0002402142857142857,
+      "loss": 0.1967,
+      "step": 4909
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024019897959183672,
+      "loss": 0.2015,
+      "step": 4910
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024018367346938775,
+      "loss": 0.2417,
+      "step": 4911
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024016836734693875,
+      "loss": 0.1897,
+      "step": 4912
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024015306122448978,
+      "loss": 0.2237,
+      "step": 4913
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024013775510204079,
+      "loss": 0.1882,
+      "step": 4914
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024012244897959184,
+      "loss": 0.1721,
+      "step": 4915
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024010714285714285,
+      "loss": 0.186,
+      "step": 4916
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024009183673469385,
+      "loss": 0.2576,
+      "step": 4917
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00024007653061224485,
+      "loss": 0.128,
+      "step": 4918
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.0002400612244897959,
+      "loss": 0.1324,
+      "step": 4919
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00024004591836734691,
+      "loss": 0.1712,
+      "step": 4920
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00024003061224489794,
+      "loss": 0.2951,
+      "step": 4921
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00024001530612244895,
+      "loss": 0.1954,
+      "step": 4922
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 0.1631,
+      "step": 4923
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.000239984693877551,
+      "loss": 0.1576,
+      "step": 4924
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.000239969387755102,
+      "loss": 0.1941,
+      "step": 4925
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00023995408163265302,
+      "loss": 0.1828,
+      "step": 4926
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00023993877551020407,
+      "loss": 0.3567,
+      "step": 4927
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00023992346938775508,
+      "loss": 0.1997,
+      "step": 4928
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.0002399081632653061,
+      "loss": 0.2158,
+      "step": 4929
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.0002398928571428571,
+      "loss": 0.1674,
+      "step": 4930
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00023987755102040814,
+      "loss": 0.2625,
+      "step": 4931
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00023986224489795917,
+      "loss": 0.179,
+      "step": 4932
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00023984693877551018,
+      "loss": 0.3593,
+      "step": 4933
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00023983163265306118,
+      "loss": 0.2291,
+      "step": 4934
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00023981632653061224,
+      "loss": 0.158,
+      "step": 4935
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00023980102040816324,
+      "loss": 0.2294,
+      "step": 4936
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00023978571428571427,
+      "loss": 0.2093,
+      "step": 4937
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00023977040816326527,
+      "loss": 0.1282,
+      "step": 4938
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0002397551020408163,
+      "loss": 0.2521,
+      "step": 4939
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.00023973979591836734,
+      "loss": 0.2029,
+      "step": 4940
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.00023972448979591834,
+      "loss": 0.1743,
+      "step": 4941
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.00023970918367346934,
+      "loss": 0.1727,
+      "step": 4942
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0002396938775510204,
+      "loss": 0.1243,
+      "step": 4943
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0002396785714285714,
+      "loss": 0.1127,
+      "step": 4944
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0002396785714285714,
+      "loss": 0.1251,
+      "step": 4945
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.00023966326530612243,
+      "loss": 0.2151,
+      "step": 4946
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.00023964795918367344,
+      "loss": 0.1883,
+      "step": 4947
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.00023963265306122447,
+      "loss": 0.1623,
+      "step": 4948
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0002396173469387755,
+      "loss": 0.1836,
+      "step": 4949
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.0002396020408163265,
+      "loss": 0.1902,
+      "step": 4950
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.0002395867346938775,
+      "loss": 0.1915,
+      "step": 4951
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00023957142857142856,
+      "loss": 0.2355,
+      "step": 4952
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00023955612244897957,
+      "loss": 0.1618,
+      "step": 4953
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00023954081632653057,
+      "loss": 0.221,
+      "step": 4954
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00023952551020408163,
+      "loss": 0.1838,
+      "step": 4955
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00023951020408163263,
+      "loss": 0.1654,
+      "step": 4956
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00023949489795918366,
+      "loss": 0.2163,
+      "step": 4957
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00023947959183673467,
+      "loss": 0.2516,
+      "step": 4958
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.0002394642857142857,
+      "loss": 0.2666,
+      "step": 4959
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00023944897959183673,
+      "loss": 0.1425,
+      "step": 4960
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00023943367346938773,
+      "loss": 0.3203,
+      "step": 4961
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00023941836734693873,
+      "loss": 0.2069,
+      "step": 4962
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0002394030612244898,
+      "loss": 0.1821,
+      "step": 4963
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0002393877551020408,
+      "loss": 0.4887,
+      "step": 4964
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00023937244897959182,
+      "loss": 0.2872,
+      "step": 4965
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00023935714285714283,
+      "loss": 0.2564,
+      "step": 4966
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00023934183673469386,
+      "loss": 0.1556,
+      "step": 4967
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0002393265306122449,
+      "loss": 0.1881,
+      "step": 4968
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0002393112244897959,
+      "loss": 0.1689,
+      "step": 4969
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.0002392959183673469,
+      "loss": 0.1554,
+      "step": 4970
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023928061224489795,
+      "loss": 0.2481,
+      "step": 4971
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023926530612244896,
+      "loss": 0.1843,
+      "step": 4972
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023925,
+      "loss": 0.1801,
+      "step": 4973
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.000239234693877551,
+      "loss": 0.1969,
+      "step": 4974
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023921938775510202,
+      "loss": 0.1479,
+      "step": 4975
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023920408163265305,
+      "loss": 0.2219,
+      "step": 4976
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023918877551020406,
+      "loss": 0.3248,
+      "step": 4977
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023917346938775506,
+      "loss": 0.2265,
+      "step": 4978
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023915816326530612,
+      "loss": 0.138,
+      "step": 4979
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023914285714285712,
+      "loss": 0.2202,
+      "step": 4980
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023912755102040812,
+      "loss": 0.1257,
+      "step": 4981
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023911224489795915,
+      "loss": 0.2529,
+      "step": 4982
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023909693877551018,
+      "loss": 0.2049,
+      "step": 4983
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023908163265306122,
+      "loss": 0.1857,
+      "step": 4984
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023906632653061222,
+      "loss": 0.1735,
+      "step": 4985
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023905102040816322,
+      "loss": 0.1899,
+      "step": 4986
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023903571428571428,
+      "loss": 0.1673,
+      "step": 4987
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023902040816326528,
+      "loss": 0.1966,
+      "step": 4988
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.0002390051020408163,
+      "loss": 0.267,
+      "step": 4989
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00023898979591836732,
+      "loss": 0.2516,
+      "step": 4990
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00023897448979591835,
+      "loss": 0.2035,
+      "step": 4991
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00023895918367346938,
+      "loss": 0.1568,
+      "step": 4992
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00023894387755102038,
+      "loss": 0.2364,
+      "step": 4993
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00023892857142857139,
+      "loss": 0.2089,
+      "step": 4994
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00023891326530612244,
+      "loss": 0.2452,
+      "step": 4995
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00023889795918367345,
+      "loss": 0.3298,
+      "step": 4996
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00023888265306122445,
+      "loss": 0.2573,
+      "step": 4997
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.0002388673469387755,
+      "loss": 0.2566,
+      "step": 4998
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.0002388520408163265,
+      "loss": 0.1714,
+      "step": 4999
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00023883673469387754,
+      "loss": 0.3336,
+      "step": 5000
+    },
+    {
+      "epoch": 4.85,
+      "eval_cer": 0.2572022388494869,
+      "eval_loss": 0.8915189504623413,
+      "eval_runtime": 228.5293,
+      "eval_samples_per_second": 38.568,
+      "eval_steps_per_second": 4.822,
+      "eval_wer": 0.6594272155027519,
+      "step": 5000
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.00023882142857142854,
+      "loss": 0.1418,
+      "step": 5001
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.00023880612244897958,
+      "loss": 0.2306,
+      "step": 5002
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0002387908163265306,
+      "loss": 0.2443,
+      "step": 5003
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0002387755102040816,
+      "loss": 0.2139,
+      "step": 5004
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0002387602040816326,
+      "loss": 0.1748,
+      "step": 5005
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.00023874489795918367,
+      "loss": 0.2511,
+      "step": 5006
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.00023872959183673467,
+      "loss": 0.2181,
+      "step": 5007
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0002387142857142857,
+      "loss": 0.2938,
+      "step": 5008
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0002386989795918367,
+      "loss": 0.261,
+      "step": 5009
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.00023868367346938774,
+      "loss": 0.1489,
+      "step": 5010
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023866836734693877,
+      "loss": 0.1362,
+      "step": 5011
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023865306122448977,
+      "loss": 0.2081,
+      "step": 5012
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023863775510204078,
+      "loss": 0.2801,
+      "step": 5013
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023862244897959183,
+      "loss": 0.3144,
+      "step": 5014
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023860714285714284,
+      "loss": 0.1699,
+      "step": 5015
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023859183673469384,
+      "loss": 0.2291,
+      "step": 5016
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023857653061224487,
+      "loss": 0.2203,
+      "step": 5017
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.0002385612244897959,
+      "loss": 0.1778,
+      "step": 5018
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023854591836734693,
+      "loss": 0.1687,
+      "step": 5019
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023853061224489794,
+      "loss": 0.2355,
+      "step": 5020
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023851530612244894,
+      "loss": 0.1674,
+      "step": 5021
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0002385,
+      "loss": 0.2148,
+      "step": 5022
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.000238484693877551,
+      "loss": 0.1974,
+      "step": 5023
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.000238469387755102,
+      "loss": 0.1838,
+      "step": 5024
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.00023845408163265303,
+      "loss": 0.1867,
+      "step": 5025
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.00023843877551020406,
+      "loss": 0.2032,
+      "step": 5026
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0002384234693877551,
+      "loss": 0.2589,
+      "step": 5027
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0002384081632653061,
+      "loss": 0.2969,
+      "step": 5028
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0002383928571428571,
+      "loss": 0.2286,
+      "step": 5029
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.00023837755102040816,
+      "loss": 0.1335,
+      "step": 5030
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.00023836224489795916,
+      "loss": 0.185,
+      "step": 5031
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023834693877551017,
+      "loss": 0.1698,
+      "step": 5032
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.0002383316326530612,
+      "loss": 0.2097,
+      "step": 5033
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023831632653061223,
+      "loss": 0.238,
+      "step": 5034
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023830102040816326,
+      "loss": 0.1456,
+      "step": 5035
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023828571428571426,
+      "loss": 0.137,
+      "step": 5036
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023827040816326527,
+      "loss": 0.1644,
+      "step": 5037
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023825510204081632,
+      "loss": 0.1994,
+      "step": 5038
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023823979591836733,
+      "loss": 0.2342,
+      "step": 5039
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023822448979591833,
+      "loss": 0.2064,
+      "step": 5040
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00023820918367346936,
+      "loss": 0.237,
+      "step": 5041
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0002381938775510204,
+      "loss": 0.1401,
+      "step": 5042
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0002381785714285714,
+      "loss": 0.2139,
+      "step": 5043
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00023816326530612242,
+      "loss": 0.1499,
+      "step": 5044
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00023814795918367346,
+      "loss": 0.2804,
+      "step": 5045
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00023813265306122449,
+      "loss": 0.2482,
+      "step": 5046
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0002381173469387755,
+      "loss": 0.2112,
+      "step": 5047
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0002381020408163265,
+      "loss": 0.1585,
+      "step": 5048
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00023808673469387755,
+      "loss": 0.1838,
+      "step": 5049
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00023807142857142855,
+      "loss": 0.2139,
+      "step": 5050
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00023805612244897956,
+      "loss": 0.2264,
+      "step": 5051
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0002380408163265306,
+      "loss": 0.2232,
+      "step": 5052
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023802551020408162,
+      "loss": 0.2204,
+      "step": 5053
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023801020408163265,
+      "loss": 0.2204,
+      "step": 5054
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023799489795918365,
+      "loss": 0.2455,
+      "step": 5055
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023797959183673466,
+      "loss": 0.1562,
+      "step": 5056
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.0002379642857142857,
+      "loss": 0.261,
+      "step": 5057
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023794897959183672,
+      "loss": 0.3657,
+      "step": 5058
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023793367346938772,
+      "loss": 0.2999,
+      "step": 5059
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023791836734693875,
+      "loss": 0.1527,
+      "step": 5060
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023790306122448978,
+      "loss": 0.1157,
+      "step": 5061
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.0002378877551020408,
+      "loss": 0.1369,
+      "step": 5062
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023787244897959182,
+      "loss": 0.1298,
+      "step": 5063
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023785714285714282,
+      "loss": 0.3211,
+      "step": 5064
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023784183673469388,
+      "loss": 0.2611,
+      "step": 5065
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023782653061224488,
+      "loss": 0.2474,
+      "step": 5066
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023781122448979588,
+      "loss": 0.1146,
+      "step": 5067
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023779591836734691,
+      "loss": 0.1804,
+      "step": 5068
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023778061224489794,
+      "loss": 0.1652,
+      "step": 5069
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023776530612244898,
+      "loss": 0.2135,
+      "step": 5070
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023774999999999998,
+      "loss": 0.3233,
+      "step": 5071
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023773469387755098,
+      "loss": 0.2317,
+      "step": 5072
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00023771938775510204,
+      "loss": 0.1938,
+      "step": 5073
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00023770408163265304,
+      "loss": 0.2087,
+      "step": 5074
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00023768877551020405,
+      "loss": 0.1482,
+      "step": 5075
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00023767346938775508,
+      "loss": 0.1653,
+      "step": 5076
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0002376581632653061,
+      "loss": 0.1438,
+      "step": 5077
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0002376428571428571,
+      "loss": 0.3308,
+      "step": 5078
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00023762755102040814,
+      "loss": 0.2182,
+      "step": 5079
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00023761224489795914,
+      "loss": 0.1248,
+      "step": 5080
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0002375969387755102,
+      "loss": 0.1667,
+      "step": 5081
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0002375816326530612,
+      "loss": 0.1996,
+      "step": 5082
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0002375663265306122,
+      "loss": 0.2036,
+      "step": 5083
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023755102040816324,
+      "loss": 0.1879,
+      "step": 5084
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023753571428571427,
+      "loss": 0.219,
+      "step": 5085
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023752040816326527,
+      "loss": 0.1353,
+      "step": 5086
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0002375051020408163,
+      "loss": 0.1467,
+      "step": 5087
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023748979591836734,
+      "loss": 0.2923,
+      "step": 5088
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023747448979591837,
+      "loss": 0.1784,
+      "step": 5089
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023745918367346937,
+      "loss": 0.2833,
+      "step": 5090
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023744387755102037,
+      "loss": 0.1805,
+      "step": 5091
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023742857142857143,
+      "loss": 0.2056,
+      "step": 5092
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.00023741326530612243,
+      "loss": 0.1416,
+      "step": 5093
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00023739795918367344,
+      "loss": 0.1871,
+      "step": 5094
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00023738265306122447,
+      "loss": 0.1753,
+      "step": 5095
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0002373673469387755,
+      "loss": 0.2201,
+      "step": 5096
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00023735204081632653,
+      "loss": 0.1909,
+      "step": 5097
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00023733673469387753,
+      "loss": 0.2288,
+      "step": 5098
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00023732142857142854,
+      "loss": 0.1379,
+      "step": 5099
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0002373061224489796,
+      "loss": 0.18,
+      "step": 5100
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0002372908163265306,
+      "loss": 0.1995,
+      "step": 5101
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0002372755102040816,
+      "loss": 0.3136,
+      "step": 5102
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00023726020408163263,
+      "loss": 0.2488,
+      "step": 5103
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00023724489795918366,
+      "loss": 0.1616,
+      "step": 5104
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00023722959183673466,
+      "loss": 0.1826,
+      "step": 5105
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.0002372142857142857,
+      "loss": 0.2493,
+      "step": 5106
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.0002371989795918367,
+      "loss": 0.1333,
+      "step": 5107
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00023718367346938776,
+      "loss": 0.2008,
+      "step": 5108
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00023716836734693876,
+      "loss": 0.2609,
+      "step": 5109
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00023715306122448976,
+      "loss": 0.2295,
+      "step": 5110
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.0002371377551020408,
+      "loss": 0.1733,
+      "step": 5111
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00023712244897959182,
+      "loss": 0.1179,
+      "step": 5112
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00023710714285714283,
+      "loss": 0.2131,
+      "step": 5113
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023709183673469386,
+      "loss": 0.2275,
+      "step": 5114
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023707653061224486,
+      "loss": 0.1442,
+      "step": 5115
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023706122448979592,
+      "loss": 0.2445,
+      "step": 5116
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023704591836734692,
+      "loss": 0.2236,
+      "step": 5117
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023703061224489793,
+      "loss": 0.163,
+      "step": 5118
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023701530612244896,
+      "loss": 0.1231,
+      "step": 5119
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.000237,
+      "loss": 0.2973,
+      "step": 5120
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.000236984693877551,
+      "loss": 0.2579,
+      "step": 5121
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023696938775510202,
+      "loss": 0.1458,
+      "step": 5122
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023695408163265302,
+      "loss": 0.1419,
+      "step": 5123
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00023693877551020408,
+      "loss": 0.1538,
+      "step": 5124
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00023692346938775509,
+      "loss": 0.1403,
+      "step": 5125
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.0002369081632653061,
+      "loss": 0.1807,
+      "step": 5126
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00023689285714285712,
+      "loss": 0.2474,
+      "step": 5127
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00023687755102040815,
+      "loss": 0.2701,
+      "step": 5128
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00023686224489795915,
+      "loss": 0.1693,
+      "step": 5129
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00023684693877551018,
+      "loss": 0.1488,
+      "step": 5130
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.0002368316326530612,
+      "loss": 0.1674,
+      "step": 5131
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00023681632653061225,
+      "loss": 0.3278,
+      "step": 5132
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00023680102040816325,
+      "loss": 0.2672,
+      "step": 5133
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00023678571428571425,
+      "loss": 0.1991,
+      "step": 5134
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.0002367704081632653,
+      "loss": 0.2018,
+      "step": 5135
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.0002367551020408163,
+      "loss": 0.2032,
+      "step": 5136
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00023673979591836732,
+      "loss": 0.176,
+      "step": 5137
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00023672448979591835,
+      "loss": 0.1161,
+      "step": 5138
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00023670918367346938,
+      "loss": 0.1737,
+      "step": 5139
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00023669387755102038,
+      "loss": 0.348,
+      "step": 5140
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.0002366785714285714,
+      "loss": 0.3212,
+      "step": 5141
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00023666326530612242,
+      "loss": 0.2265,
+      "step": 5142
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00023664795918367347,
+      "loss": 0.1468,
+      "step": 5143
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00023663265306122448,
+      "loss": 0.1375,
+      "step": 5144
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023661734693877548,
+      "loss": 0.2123,
+      "step": 5145
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.0002366020408163265,
+      "loss": 0.1991,
+      "step": 5146
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023658673469387754,
+      "loss": 0.3094,
+      "step": 5147
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023657142857142854,
+      "loss": 0.1916,
+      "step": 5148
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023655612244897958,
+      "loss": 0.1953,
+      "step": 5149
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023654081632653058,
+      "loss": 0.1685,
+      "step": 5150
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023652551020408164,
+      "loss": 0.1682,
+      "step": 5151
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023651020408163264,
+      "loss": 0.218,
+      "step": 5152
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023649489795918364,
+      "loss": 0.1788,
+      "step": 5153
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023647959183673467,
+      "loss": 0.0864,
+      "step": 5154
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.0002364642857142857,
+      "loss": 0.1385,
+      "step": 5155
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002364489795918367,
+      "loss": 0.1034,
+      "step": 5156
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.00023643367346938774,
+      "loss": 0.2907,
+      "step": 5157
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.00023641836734693874,
+      "loss": 0.1545,
+      "step": 5158
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002364030612244898,
+      "loss": 0.1552,
+      "step": 5159
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002363877551020408,
+      "loss": 0.1458,
+      "step": 5160
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002363724489795918,
+      "loss": 0.1833,
+      "step": 5161
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002363571428571428,
+      "loss": 0.1905,
+      "step": 5162
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.00023634183673469387,
+      "loss": 0.1784,
+      "step": 5163
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.00023632653061224487,
+      "loss": 0.1975,
+      "step": 5164
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002363112244897959,
+      "loss": 0.1851,
+      "step": 5165
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.0002362959183673469,
+      "loss": 0.1683,
+      "step": 5166
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00023628061224489796,
+      "loss": 0.0747,
+      "step": 5167
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00023626530612244897,
+      "loss": 0.1862,
+      "step": 5168
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00023624999999999997,
+      "loss": 0.1721,
+      "step": 5169
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00023623469387755097,
+      "loss": 0.3111,
+      "step": 5170
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00023621938775510203,
+      "loss": 0.2047,
+      "step": 5171
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00023620408163265303,
+      "loss": 0.1635,
+      "step": 5172
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00023618877551020406,
+      "loss": 0.1198,
+      "step": 5173
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00023617346938775507,
+      "loss": 0.1375,
+      "step": 5174
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.0002361581632653061,
+      "loss": 0.3102,
+      "step": 5175
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00023614285714285713,
+      "loss": 0.1549,
+      "step": 5176
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00023612755102040813,
+      "loss": 0.2255,
+      "step": 5177
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00023611224489795914,
+      "loss": 0.1282,
+      "step": 5178
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.0002360969387755102,
+      "loss": 0.1283,
+      "step": 5179
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.0002360816326530612,
+      "loss": 0.1044,
+      "step": 5180
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00023606632653061223,
+      "loss": 0.1599,
+      "step": 5181
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00023605102040816326,
+      "loss": 0.4048,
+      "step": 5182
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00023603571428571426,
+      "loss": 0.2185,
+      "step": 5183
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.0002360204081632653,
+      "loss": 0.1831,
+      "step": 5184
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.0002360051020408163,
+      "loss": 0.1692,
+      "step": 5185
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00023598979591836735,
+      "loss": 0.1046,
+      "step": 5186
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023597448979591836,
+      "loss": 0.2083,
+      "step": 5187
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023595918367346936,
+      "loss": 0.2234,
+      "step": 5188
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.0002359438775510204,
+      "loss": 0.1832,
+      "step": 5189
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023592857142857142,
+      "loss": 0.2292,
+      "step": 5190
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023591326530612242,
+      "loss": 0.1746,
+      "step": 5191
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023589795918367345,
+      "loss": 0.1543,
+      "step": 5192
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023588265306122446,
+      "loss": 0.1808,
+      "step": 5193
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023586734693877552,
+      "loss": 0.1773,
+      "step": 5194
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023585204081632652,
+      "loss": 0.1545,
+      "step": 5195
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023583673469387752,
+      "loss": 0.267,
+      "step": 5196
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00023582142857142853,
+      "loss": 0.1546,
+      "step": 5197
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00023580612244897958,
+      "loss": 0.1563,
+      "step": 5198
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0002357908163265306,
+      "loss": 0.2361,
+      "step": 5199
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00023577551020408162,
+      "loss": 0.1362,
+      "step": 5200
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00023576020408163262,
+      "loss": 0.1897,
+      "step": 5201
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00023574489795918365,
+      "loss": 0.3007,
+      "step": 5202
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00023572959183673468,
+      "loss": 0.2322,
+      "step": 5203
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00023571428571428569,
+      "loss": 0.1287,
+      "step": 5204
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0002356989795918367,
+      "loss": 0.1094,
+      "step": 5205
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00023568367346938775,
+      "loss": 0.1496,
+      "step": 5206
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00023566836734693875,
+      "loss": 0.2587,
+      "step": 5207
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00023565306122448978,
+      "loss": 0.1811,
+      "step": 5208
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00023563775510204078,
+      "loss": 0.1607,
+      "step": 5209
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00023562244897959181,
+      "loss": 0.1633,
+      "step": 5210
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00023560714285714285,
+      "loss": 0.1638,
+      "step": 5211
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00023559183673469385,
+      "loss": 0.1825,
+      "step": 5212
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00023557653061224485,
+      "loss": 0.2234,
+      "step": 5213
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0002355612244897959,
+      "loss": 0.2548,
+      "step": 5214
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0002355459183673469,
+      "loss": 0.1255,
+      "step": 5215
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00023553061224489794,
+      "loss": 0.1262,
+      "step": 5216
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023551530612244895,
+      "loss": 0.1407,
+      "step": 5217
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023549999999999998,
+      "loss": 0.097,
+      "step": 5218
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.000235484693877551,
+      "loss": 0.1965,
+      "step": 5219
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.000235469387755102,
+      "loss": 0.2005,
+      "step": 5220
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023545408163265302,
+      "loss": 0.1965,
+      "step": 5221
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023543877551020407,
+      "loss": 0.1461,
+      "step": 5222
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023542346938775508,
+      "loss": 0.1238,
+      "step": 5223
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023540816326530608,
+      "loss": 0.1535,
+      "step": 5224
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023539285714285714,
+      "loss": 0.1835,
+      "step": 5225
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023537755102040814,
+      "loss": 0.2416,
+      "step": 5226
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00023536224489795917,
+      "loss": 0.2325,
+      "step": 5227
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00023534693877551018,
+      "loss": 0.1828,
+      "step": 5228
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00023533163265306123,
+      "loss": 0.1607,
+      "step": 5229
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00023531632653061224,
+      "loss": 0.1457,
+      "step": 5230
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00023530102040816324,
+      "loss": 0.1358,
+      "step": 5231
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00023528571428571424,
+      "loss": 0.433,
+      "step": 5232
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.0002352704081632653,
+      "loss": 0.1156,
+      "step": 5233
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.0002352551020408163,
+      "loss": 0.152,
+      "step": 5234
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00023523979591836733,
+      "loss": 0.1142,
+      "step": 5235
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00023522448979591834,
+      "loss": 0.1029,
+      "step": 5236
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00023520918367346937,
+      "loss": 0.1047,
+      "step": 5237
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0002351938775510204,
+      "loss": 0.2023,
+      "step": 5238
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0002351785714285714,
+      "loss": 0.1353,
+      "step": 5239
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0002351632653061224,
+      "loss": 0.1963,
+      "step": 5240
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.00023514795918367346,
+      "loss": 0.2039,
+      "step": 5241
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.00023513265306122447,
+      "loss": 0.1555,
+      "step": 5242
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0002351173469387755,
+      "loss": 0.2373,
+      "step": 5243
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0002351020408163265,
+      "loss": 0.2331,
+      "step": 5244
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.00023508673469387753,
+      "loss": 0.1457,
+      "step": 5245
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.00023507142857142856,
+      "loss": 0.2744,
+      "step": 5246
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.00023505612244897957,
+      "loss": 0.1476,
+      "step": 5247
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00023504081632653057,
+      "loss": 0.1909,
+      "step": 5248
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00023502551020408163,
+      "loss": 0.2042,
+      "step": 5249
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00023501020408163263,
+      "loss": 0.2068,
+      "step": 5250
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00023499489795918366,
+      "loss": 0.2278,
+      "step": 5251
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00023497959183673466,
+      "loss": 0.2101,
+      "step": 5252
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.0002349642857142857,
+      "loss": 0.0988,
+      "step": 5253
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00023494897959183673,
+      "loss": 0.0933,
+      "step": 5254
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00023493367346938773,
+      "loss": 0.1931,
+      "step": 5255
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00023491836734693873,
+      "loss": 0.357,
+      "step": 5256
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.0002349030612244898,
+      "loss": 0.1913,
+      "step": 5257
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.0002348877551020408,
+      "loss": 0.2233,
+      "step": 5258
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.0002348724489795918,
+      "loss": 0.1492,
+      "step": 5259
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00023485714285714283,
+      "loss": 0.1538,
+      "step": 5260
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00023484183673469386,
+      "loss": 0.1646,
+      "step": 5261
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.0002348265306122449,
+      "loss": 0.1811,
+      "step": 5262
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.0002348112244897959,
+      "loss": 0.2652,
+      "step": 5263
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.0002347959183673469,
+      "loss": 0.154,
+      "step": 5264
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00023478061224489795,
+      "loss": 0.2442,
+      "step": 5265
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00023476530612244896,
+      "loss": 0.0989,
+      "step": 5266
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00023474999999999996,
+      "loss": 0.1421,
+      "step": 5267
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.000234734693877551,
+      "loss": 0.1468,
+      "step": 5268
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023471938775510202,
+      "loss": 0.342,
+      "step": 5269
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023470408163265305,
+      "loss": 0.2395,
+      "step": 5270
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023468877551020405,
+      "loss": 0.2288,
+      "step": 5271
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023467346938775509,
+      "loss": 0.0947,
+      "step": 5272
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023465816326530612,
+      "loss": 0.2089,
+      "step": 5273
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023464285714285712,
+      "loss": 0.1728,
+      "step": 5274
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023462755102040812,
+      "loss": 0.1814,
+      "step": 5275
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023461224489795918,
+      "loss": 0.152,
+      "step": 5276
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023459693877551018,
+      "loss": 0.1702,
+      "step": 5277
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00023458163265306121,
+      "loss": 0.155,
+      "step": 5278
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023456632653061222,
+      "loss": 0.1331,
+      "step": 5279
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023455102040816325,
+      "loss": 0.1104,
+      "step": 5280
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023453571428571428,
+      "loss": 0.1426,
+      "step": 5281
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023452040816326528,
+      "loss": 0.3587,
+      "step": 5282
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023450510204081629,
+      "loss": 0.1283,
+      "step": 5283
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023448979591836734,
+      "loss": 0.2176,
+      "step": 5284
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023447448979591835,
+      "loss": 0.145,
+      "step": 5285
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023445918367346938,
+      "loss": 0.0865,
+      "step": 5286
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023444387755102038,
+      "loss": 0.1601,
+      "step": 5287
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.0002344285714285714,
+      "loss": 0.2898,
+      "step": 5288
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00023441326530612244,
+      "loss": 0.1742,
+      "step": 5289
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.00023439795918367345,
+      "loss": 0.2023,
+      "step": 5290
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.00023438265306122445,
+      "loss": 0.1876,
+      "step": 5291
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002343673469387755,
+      "loss": 0.1662,
+      "step": 5292
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002343520408163265,
+      "loss": 0.1278,
+      "step": 5293
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002343367346938775,
+      "loss": 0.2373,
+      "step": 5294
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.00023432142857142854,
+      "loss": 0.1577,
+      "step": 5295
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.00023430612244897957,
+      "loss": 0.1785,
+      "step": 5296
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002342908163265306,
+      "loss": 0.1329,
+      "step": 5297
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002342755102040816,
+      "loss": 0.1367,
+      "step": 5298
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002342602040816326,
+      "loss": 0.1797,
+      "step": 5299
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023424489795918367,
+      "loss": 0.1982,
+      "step": 5300
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023422959183673467,
+      "loss": 0.1949,
+      "step": 5301
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023421428571428568,
+      "loss": 0.1855,
+      "step": 5302
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.0002341989795918367,
+      "loss": 0.1178,
+      "step": 5303
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023418367346938774,
+      "loss": 0.226,
+      "step": 5304
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023416836734693877,
+      "loss": 0.1346,
+      "step": 5305
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023415306122448977,
+      "loss": 0.1431,
+      "step": 5306
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023413775510204078,
+      "loss": 0.1598,
+      "step": 5307
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023412244897959183,
+      "loss": 0.1975,
+      "step": 5308
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00023410714285714284,
+      "loss": 0.182,
+      "step": 5309
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00023409183673469384,
+      "loss": 0.1576,
+      "step": 5310
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00023407653061224487,
+      "loss": 0.2391,
+      "step": 5311
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.0002340612244897959,
+      "loss": 0.135,
+      "step": 5312
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00023404591836734693,
+      "loss": 0.2614,
+      "step": 5313
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00023403061224489793,
+      "loss": 0.2071,
+      "step": 5314
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00023401530612244897,
+      "loss": 0.1798,
+      "step": 5315
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.000234,
+      "loss": 0.2209,
+      "step": 5316
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.000233984693877551,
+      "loss": 0.1616,
+      "step": 5317
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.000233969387755102,
+      "loss": 0.1947,
+      "step": 5318
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00023395408163265306,
+      "loss": 0.2206,
+      "step": 5319
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023393877551020406,
+      "loss": 0.1845,
+      "step": 5320
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023392346938775507,
+      "loss": 0.1715,
+      "step": 5321
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.0002339081632653061,
+      "loss": 0.1121,
+      "step": 5322
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023389285714285713,
+      "loss": 0.1556,
+      "step": 5323
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023387755102040816,
+      "loss": 0.1096,
+      "step": 5324
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023386224489795916,
+      "loss": 0.1359,
+      "step": 5325
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023384693877551017,
+      "loss": 0.2977,
+      "step": 5326
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023383163265306122,
+      "loss": 0.1344,
+      "step": 5327
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023381632653061223,
+      "loss": 0.1165,
+      "step": 5328
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023380102040816323,
+      "loss": 0.1189,
+      "step": 5329
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00023378571428571426,
+      "loss": 0.1414,
+      "step": 5330
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.0002337704081632653,
+      "loss": 0.1676,
+      "step": 5331
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00023375510204081632,
+      "loss": 0.2194,
+      "step": 5332
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00023373979591836733,
+      "loss": 0.226,
+      "step": 5333
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00023372448979591833,
+      "loss": 0.1381,
+      "step": 5334
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00023370918367346939,
+      "loss": 0.1567,
+      "step": 5335
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.0002336938775510204,
+      "loss": 0.1512,
+      "step": 5336
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.0002336785714285714,
+      "loss": 0.1581,
+      "step": 5337
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00023366326530612242,
+      "loss": 0.2575,
+      "step": 5338
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00023364795918367345,
+      "loss": 0.1734,
+      "step": 5339
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00023363265306122448,
+      "loss": 0.1682,
+      "step": 5340
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.0002336173469387755,
+      "loss": 0.1246,
+      "step": 5341
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.0002336020408163265,
+      "loss": 0.1928,
+      "step": 5342
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00023358673469387755,
+      "loss": 0.1404,
+      "step": 5343
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00023357142857142855,
+      "loss": 0.2995,
+      "step": 5344
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00023355612244897956,
+      "loss": 0.2023,
+      "step": 5345
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.0002335408163265306,
+      "loss": 0.1312,
+      "step": 5346
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00023352551020408162,
+      "loss": 0.1589,
+      "step": 5347
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00023351020408163265,
+      "loss": 0.1906,
+      "step": 5348
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00023349489795918365,
+      "loss": 0.1573,
+      "step": 5349
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00023347959183673465,
+      "loss": 0.2425,
+      "step": 5350
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.0002334642857142857,
+      "loss": 0.433,
+      "step": 5351
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023344897959183672,
+      "loss": 0.2045,
+      "step": 5352
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023343367346938772,
+      "loss": 0.1646,
+      "step": 5353
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023341836734693875,
+      "loss": 0.1587,
+      "step": 5354
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023340306122448978,
+      "loss": 0.1577,
+      "step": 5355
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023338775510204078,
+      "loss": 0.2287,
+      "step": 5356
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023337244897959181,
+      "loss": 0.2209,
+      "step": 5357
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023335714285714282,
+      "loss": 0.2315,
+      "step": 5358
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023334183673469388,
+      "loss": 0.1911,
+      "step": 5359
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023332653061224488,
+      "loss": 0.1448,
+      "step": 5360
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00023331122448979588,
+      "loss": 0.158,
+      "step": 5361
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.00023329591836734694,
+      "loss": 0.1673,
+      "step": 5362
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.00023328061224489794,
+      "loss": 0.2299,
+      "step": 5363
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.00023326530612244895,
+      "loss": 0.3454,
+      "step": 5364
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.00023324999999999998,
+      "loss": 0.1682,
+      "step": 5365
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.000233234693877551,
+      "loss": 0.1535,
+      "step": 5366
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.00023321938775510204,
+      "loss": 0.1036,
+      "step": 5367
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.00023320408163265304,
+      "loss": 0.1061,
+      "step": 5368
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.00023318877551020405,
+      "loss": 0.2937,
+      "step": 5369
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.0002331734693877551,
+      "loss": 0.2488,
+      "step": 5370
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.0002331581632653061,
+      "loss": 0.1762,
+      "step": 5371
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002331428571428571,
+      "loss": 0.2288,
+      "step": 5372
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.00023312755102040814,
+      "loss": 0.1225,
+      "step": 5373
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.00023311224489795917,
+      "loss": 0.1009,
+      "step": 5374
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002330969387755102,
+      "loss": 0.1299,
+      "step": 5375
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002330816326530612,
+      "loss": 0.2557,
+      "step": 5376
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002330663265306122,
+      "loss": 0.2402,
+      "step": 5377
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.00023305102040816327,
+      "loss": 0.1842,
+      "step": 5378
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.00023303571428571427,
+      "loss": 0.124,
+      "step": 5379
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.00023302040816326527,
+      "loss": 0.1442,
+      "step": 5380
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002330051020408163,
+      "loss": 0.0977,
+      "step": 5381
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023298979591836733,
+      "loss": 0.2283,
+      "step": 5382
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023297448979591834,
+      "loss": 0.1386,
+      "step": 5383
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023295918367346937,
+      "loss": 0.1921,
+      "step": 5384
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023294387755102037,
+      "loss": 0.1478,
+      "step": 5385
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023292857142857143,
+      "loss": 0.092,
+      "step": 5386
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023291326530612243,
+      "loss": 0.1659,
+      "step": 5387
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023289795918367344,
+      "loss": 0.1449,
+      "step": 5388
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023288265306122447,
+      "loss": 0.1779,
+      "step": 5389
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.0002328673469387755,
+      "loss": 0.1369,
+      "step": 5390
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.0002328520408163265,
+      "loss": 0.2633,
+      "step": 5391
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00023283673469387753,
+      "loss": 0.0944,
+      "step": 5392
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.00023282142857142853,
+      "loss": 0.1408,
+      "step": 5393
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0002328061224489796,
+      "loss": 0.1952,
+      "step": 5394
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0002327908163265306,
+      "loss": 0.1772,
+      "step": 5395
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0002327755102040816,
+      "loss": 0.132,
+      "step": 5396
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.00023276020408163263,
+      "loss": 0.1516,
+      "step": 5397
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.00023274489795918366,
+      "loss": 0.1148,
+      "step": 5398
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.00023272959183673466,
+      "loss": 0.1045,
+      "step": 5399
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0002327142857142857,
+      "loss": 0.1606,
+      "step": 5400
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0002326989795918367,
+      "loss": 0.2515,
+      "step": 5401
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.00023268367346938776,
+      "loss": 0.126,
+      "step": 5402
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023266836734693876,
+      "loss": 0.1219,
+      "step": 5403
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023265306122448976,
+      "loss": 0.2443,
+      "step": 5404
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023263775510204082,
+      "loss": 0.1322,
+      "step": 5405
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023262244897959182,
+      "loss": 0.1688,
+      "step": 5406
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023260714285714283,
+      "loss": 0.1818,
+      "step": 5407
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023259183673469386,
+      "loss": 0.1416,
+      "step": 5408
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.0002325765306122449,
+      "loss": 0.1136,
+      "step": 5409
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023256122448979592,
+      "loss": 0.2992,
+      "step": 5410
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023254591836734692,
+      "loss": 0.1673,
+      "step": 5411
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00023253061224489793,
+      "loss": 0.1448,
+      "step": 5412
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00023251530612244898,
+      "loss": 0.1375,
+      "step": 5413
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00023249999999999999,
+      "loss": 0.1461,
+      "step": 5414
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.000232484693877551,
+      "loss": 0.1073,
+      "step": 5415
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00023246938775510202,
+      "loss": 0.1676,
+      "step": 5416
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00023245408163265305,
+      "loss": 0.2043,
+      "step": 5417
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00023243877551020405,
+      "loss": 0.1274,
+      "step": 5418
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00023242346938775508,
+      "loss": 0.2683,
+      "step": 5419
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.0002324081632653061,
+      "loss": 0.2681,
+      "step": 5420
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00023239285714285715,
+      "loss": 0.1433,
+      "step": 5421
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00023237755102040815,
+      "loss": 0.1599,
+      "step": 5422
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023236224489795915,
+      "loss": 0.2249,
+      "step": 5423
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023234693877551018,
+      "loss": 0.0951,
+      "step": 5424
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023233163265306121,
+      "loss": 0.2294,
+      "step": 5425
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023231632653061222,
+      "loss": 0.1917,
+      "step": 5426
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023230102040816325,
+      "loss": 0.1279,
+      "step": 5427
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023228571428571425,
+      "loss": 0.1566,
+      "step": 5428
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0002322704081632653,
+      "loss": 0.1018,
+      "step": 5429
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0002322551020408163,
+      "loss": 0.2332,
+      "step": 5430
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023223979591836732,
+      "loss": 0.1554,
+      "step": 5431
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023222448979591835,
+      "loss": 0.1797,
+      "step": 5432
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00023220918367346938,
+      "loss": 0.2519,
+      "step": 5433
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023219387755102038,
+      "loss": 0.1324,
+      "step": 5434
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.0002321785714285714,
+      "loss": 0.1722,
+      "step": 5435
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023216326530612241,
+      "loss": 0.2099,
+      "step": 5436
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023214795918367347,
+      "loss": 0.1459,
+      "step": 5437
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023213265306122448,
+      "loss": 0.2482,
+      "step": 5438
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023211734693877548,
+      "loss": 0.1759,
+      "step": 5439
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023210204081632648,
+      "loss": 0.1811,
+      "step": 5440
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023208673469387754,
+      "loss": 0.1621,
+      "step": 5441
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023207142857142854,
+      "loss": 0.1152,
+      "step": 5442
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00023205612244897957,
+      "loss": 0.1179,
+      "step": 5443
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00023204081632653058,
+      "loss": 0.2135,
+      "step": 5444
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.0002320255102040816,
+      "loss": 0.1656,
+      "step": 5445
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00023201020408163264,
+      "loss": 0.1623,
+      "step": 5446
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00023199489795918364,
+      "loss": 0.1214,
+      "step": 5447
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00023197959183673465,
+      "loss": 0.1134,
+      "step": 5448
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.0002319642857142857,
+      "loss": 0.1711,
+      "step": 5449
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.0002319489795918367,
+      "loss": 0.2132,
+      "step": 5450
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00023193367346938774,
+      "loss": 0.1594,
+      "step": 5451
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00023191836734693877,
+      "loss": 0.142,
+      "step": 5452
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00023190306122448977,
+      "loss": 0.2208,
+      "step": 5453
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.0002318877551020408,
+      "loss": 0.1185,
+      "step": 5454
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.0002318724489795918,
+      "loss": 0.127,
+      "step": 5455
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00023185714285714286,
+      "loss": 0.1032,
+      "step": 5456
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00023184183673469387,
+      "loss": 0.2598,
+      "step": 5457
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00023182653061224487,
+      "loss": 0.1515,
+      "step": 5458
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.0002318112244897959,
+      "loss": 0.1339,
+      "step": 5459
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00023179591836734693,
+      "loss": 0.1882,
+      "step": 5460
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00023178061224489793,
+      "loss": 0.1472,
+      "step": 5461
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00023176530612244896,
+      "loss": 0.2046,
+      "step": 5462
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00023174999999999997,
+      "loss": 0.4023,
+      "step": 5463
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00023173469387755103,
+      "loss": 0.1721,
+      "step": 5464
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00023171938775510203,
+      "loss": 0.1687,
+      "step": 5465
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00023170408163265303,
+      "loss": 0.1099,
+      "step": 5466
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00023168877551020406,
+      "loss": 0.1524,
+      "step": 5467
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0002316734693877551,
+      "loss": 0.1722,
+      "step": 5468
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0002316581632653061,
+      "loss": 0.391,
+      "step": 5469
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00023164285714285713,
+      "loss": 0.2288,
+      "step": 5470
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00023162755102040813,
+      "loss": 0.1916,
+      "step": 5471
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0002316122448979592,
+      "loss": 0.232,
+      "step": 5472
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0002315969387755102,
+      "loss": 0.1483,
+      "step": 5473
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0002315816326530612,
+      "loss": 0.0978,
+      "step": 5474
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.0002315663265306122,
+      "loss": 0.2173,
+      "step": 5475
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00023155102040816326,
+      "loss": 0.2029,
+      "step": 5476
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00023153571428571426,
+      "loss": 0.2161,
+      "step": 5477
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.0002315204081632653,
+      "loss": 0.1769,
+      "step": 5478
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.0002315051020408163,
+      "loss": 0.1787,
+      "step": 5479
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00023148979591836732,
+      "loss": 0.1115,
+      "step": 5480
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00023147448979591836,
+      "loss": 0.15,
+      "step": 5481
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00023145918367346936,
+      "loss": 0.2521,
+      "step": 5482
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00023144387755102036,
+      "loss": 0.1682,
+      "step": 5483
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00023142857142857142,
+      "loss": 0.1727,
+      "step": 5484
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023141326530612242,
+      "loss": 0.1654,
+      "step": 5485
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023139795918367345,
+      "loss": 0.1686,
+      "step": 5486
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023138265306122446,
+      "loss": 0.1249,
+      "step": 5487
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.0002313673469387755,
+      "loss": 0.2046,
+      "step": 5488
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023135204081632652,
+      "loss": 0.2016,
+      "step": 5489
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023133673469387752,
+      "loss": 0.2332,
+      "step": 5490
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023132142857142853,
+      "loss": 0.0786,
+      "step": 5491
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023130612244897958,
+      "loss": 0.1908,
+      "step": 5492
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023129081632653059,
+      "loss": 0.0781,
+      "step": 5493
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023127551020408162,
+      "loss": 0.2064,
+      "step": 5494
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00023126020408163262,
+      "loss": 0.2755,
+      "step": 5495
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023124489795918365,
+      "loss": 0.2272,
+      "step": 5496
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023122959183673468,
+      "loss": 0.2014,
+      "step": 5497
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023121428571428568,
+      "loss": 0.134,
+      "step": 5498
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023119897959183674,
+      "loss": 0.11,
+      "step": 5499
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023118367346938775,
+      "loss": 0.1426,
+      "step": 5500
+    },
+    {
+      "epoch": 5.34,
+      "eval_cer": 0.26859753364497124,
+      "eval_loss": 0.9501057863235474,
+      "eval_runtime": 234.91,
+      "eval_samples_per_second": 37.521,
+      "eval_steps_per_second": 4.691,
+      "eval_wer": 0.6783662349361193,
+      "step": 5500
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023116836734693875,
+      "loss": 0.2486,
+      "step": 5501
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023115306122448975,
+      "loss": 0.1969,
+      "step": 5502
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.0002311377551020408,
+      "loss": 0.3211,
+      "step": 5503
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023112244897959181,
+      "loss": 0.1591,
+      "step": 5504
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00023110714285714284,
+      "loss": 0.1406,
+      "step": 5505
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.00023109183673469385,
+      "loss": 0.2264,
+      "step": 5506
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0002310765306122449,
+      "loss": 0.2698,
+      "step": 5507
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0002310612244897959,
+      "loss": 0.1371,
+      "step": 5508
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0002310459183673469,
+      "loss": 0.1516,
+      "step": 5509
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.00023103061224489792,
+      "loss": 0.1611,
+      "step": 5510
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.00023101530612244897,
+      "loss": 0.09,
+      "step": 5511
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.00023099999999999998,
+      "loss": 0.2135,
+      "step": 5512
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.000230984693877551,
+      "loss": 0.253,
+      "step": 5513
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.000230969387755102,
+      "loss": 0.164,
+      "step": 5514
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.00023095408163265304,
+      "loss": 0.1935,
+      "step": 5515
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023093877551020407,
+      "loss": 0.1048,
+      "step": 5516
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023092346938775508,
+      "loss": 0.1681,
+      "step": 5517
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023090816326530608,
+      "loss": 0.1683,
+      "step": 5518
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023089285714285714,
+      "loss": 0.1885,
+      "step": 5519
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023087755102040814,
+      "loss": 0.2299,
+      "step": 5520
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023086224489795917,
+      "loss": 0.155,
+      "step": 5521
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023084693877551017,
+      "loss": 0.1155,
+      "step": 5522
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.0002308316326530612,
+      "loss": 0.153,
+      "step": 5523
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023081632653061224,
+      "loss": 0.0919,
+      "step": 5524
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00023080102040816324,
+      "loss": 0.1715,
+      "step": 5525
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00023078571428571424,
+      "loss": 0.1708,
+      "step": 5526
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0002307704081632653,
+      "loss": 0.2546,
+      "step": 5527
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0002307551020408163,
+      "loss": 0.1554,
+      "step": 5528
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00023073979591836733,
+      "loss": 0.2333,
+      "step": 5529
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00023072448979591834,
+      "loss": 0.0918,
+      "step": 5530
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00023070918367346937,
+      "loss": 0.1638,
+      "step": 5531
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0002306938775510204,
+      "loss": 0.21,
+      "step": 5532
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0002306785714285714,
+      "loss": 0.1872,
+      "step": 5533
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0002306632653061224,
+      "loss": 0.2313,
+      "step": 5534
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00023064795918367346,
+      "loss": 0.1521,
+      "step": 5535
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00023063265306122447,
+      "loss": 0.0917,
+      "step": 5536
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00023061734693877547,
+      "loss": 0.1669,
+      "step": 5537
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.0002306020408163265,
+      "loss": 0.465,
+      "step": 5538
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00023058673469387753,
+      "loss": 0.2557,
+      "step": 5539
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00023057142857142856,
+      "loss": 0.192,
+      "step": 5540
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00023055612244897956,
+      "loss": 0.1965,
+      "step": 5541
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.0002305408163265306,
+      "loss": 0.1742,
+      "step": 5542
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00023052551020408163,
+      "loss": 0.1573,
+      "step": 5543
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00023051020408163263,
+      "loss": 0.1224,
+      "step": 5544
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00023049489795918363,
+      "loss": 0.176,
+      "step": 5545
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.0002304795918367347,
+      "loss": 0.168,
+      "step": 5546
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0002304642857142857,
+      "loss": 0.1351,
+      "step": 5547
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00023044897959183672,
+      "loss": 0.1744,
+      "step": 5548
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00023043367346938773,
+      "loss": 0.1553,
+      "step": 5549
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00023041836734693876,
+      "loss": 0.1973,
+      "step": 5550
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0002304030612244898,
+      "loss": 0.2331,
+      "step": 5551
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0002303877551020408,
+      "loss": 0.2036,
+      "step": 5552
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0002303724489795918,
+      "loss": 0.1164,
+      "step": 5553
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00023035714285714285,
+      "loss": 0.2081,
+      "step": 5554
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00023034183673469386,
+      "loss": 0.1045,
+      "step": 5555
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0002303265306122449,
+      "loss": 0.1281,
+      "step": 5556
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.0002303112244897959,
+      "loss": 0.231,
+      "step": 5557
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023029591836734692,
+      "loss": 0.158,
+      "step": 5558
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023028061224489795,
+      "loss": 0.2272,
+      "step": 5559
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023026530612244896,
+      "loss": 0.1163,
+      "step": 5560
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023024999999999996,
+      "loss": 0.0995,
+      "step": 5561
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023023469387755102,
+      "loss": 0.1067,
+      "step": 5562
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023021938775510202,
+      "loss": 0.2137,
+      "step": 5563
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023020408163265302,
+      "loss": 0.1863,
+      "step": 5564
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023018877551020405,
+      "loss": 0.1847,
+      "step": 5565
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023017346938775508,
+      "loss": 0.2718,
+      "step": 5566
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00023015816326530611,
+      "loss": 0.1038,
+      "step": 5567
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023014285714285712,
+      "loss": 0.1657,
+      "step": 5568
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023012755102040812,
+      "loss": 0.3976,
+      "step": 5569
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023011224489795918,
+      "loss": 0.2039,
+      "step": 5570
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023009693877551018,
+      "loss": 0.1758,
+      "step": 5571
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023008163265306119,
+      "loss": 0.1298,
+      "step": 5572
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023006632653061222,
+      "loss": 0.1363,
+      "step": 5573
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023005102040816325,
+      "loss": 0.1442,
+      "step": 5574
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023003571428571428,
+      "loss": 0.1605,
+      "step": 5575
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023002040816326528,
+      "loss": 0.2235,
+      "step": 5576
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00023000510204081628,
+      "loss": 0.2257,
+      "step": 5577
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00022998979591836734,
+      "loss": 0.1656,
+      "step": 5578
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00022997448979591835,
+      "loss": 0.1515,
+      "step": 5579
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00022995918367346935,
+      "loss": 0.2237,
+      "step": 5580
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00022994387755102038,
+      "loss": 0.2032,
+      "step": 5581
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.0002299285714285714,
+      "loss": 0.1514,
+      "step": 5582
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00022991326530612244,
+      "loss": 0.1942,
+      "step": 5583
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00022989795918367344,
+      "loss": 0.122,
+      "step": 5584
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00022988265306122445,
+      "loss": 0.1462,
+      "step": 5585
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.0002298673469387755,
+      "loss": 0.1253,
+      "step": 5586
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.0002298520408163265,
+      "loss": 0.1464,
+      "step": 5587
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.0002298367346938775,
+      "loss": 0.2842,
+      "step": 5588
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022982142857142857,
+      "loss": 0.207,
+      "step": 5589
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022980612244897957,
+      "loss": 0.2021,
+      "step": 5590
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.0002297908163265306,
+      "loss": 0.1124,
+      "step": 5591
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.0002297755102040816,
+      "loss": 0.1935,
+      "step": 5592
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022976020408163264,
+      "loss": 0.1684,
+      "step": 5593
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022974489795918367,
+      "loss": 0.2148,
+      "step": 5594
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022972959183673467,
+      "loss": 0.1894,
+      "step": 5595
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022971428571428568,
+      "loss": 0.1477,
+      "step": 5596
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022969897959183673,
+      "loss": 0.3286,
+      "step": 5597
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022968367346938774,
+      "loss": 0.2232,
+      "step": 5598
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00022966836734693874,
+      "loss": 0.1197,
+      "step": 5599
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00022965306122448977,
+      "loss": 0.0951,
+      "step": 5600
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.0002296377551020408,
+      "loss": 0.2308,
+      "step": 5601
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00022962244897959183,
+      "loss": 0.3105,
+      "step": 5602
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00022960714285714284,
+      "loss": 0.1572,
+      "step": 5603
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00022959183673469384,
+      "loss": 0.1236,
+      "step": 5604
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.0002295765306122449,
+      "loss": 0.0908,
+      "step": 5605
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.0002295612244897959,
+      "loss": 0.1793,
+      "step": 5606
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.0002295459183673469,
+      "loss": 0.1965,
+      "step": 5607
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00022953061224489793,
+      "loss": 0.2017,
+      "step": 5608
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00022951530612244896,
+      "loss": 0.2138,
+      "step": 5609
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.0002295,
+      "loss": 0.1985,
+      "step": 5610
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.000229484693877551,
+      "loss": 0.0967,
+      "step": 5611
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.000229469387755102,
+      "loss": 0.1392,
+      "step": 5612
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00022945408163265306,
+      "loss": 0.2338,
+      "step": 5613
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00022943877551020406,
+      "loss": 0.2162,
+      "step": 5614
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00022942346938775507,
+      "loss": 0.2054,
+      "step": 5615
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.0002294081632653061,
+      "loss": 0.1042,
+      "step": 5616
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00022939285714285713,
+      "loss": 0.1265,
+      "step": 5617
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00022937755102040816,
+      "loss": 0.2016,
+      "step": 5618
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022936224489795916,
+      "loss": 0.1339,
+      "step": 5619
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022934693877551016,
+      "loss": 0.2175,
+      "step": 5620
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022933163265306122,
+      "loss": 0.1398,
+      "step": 5621
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022931632653061223,
+      "loss": 0.1801,
+      "step": 5622
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022930102040816323,
+      "loss": 0.1494,
+      "step": 5623
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022928571428571426,
+      "loss": 0.1129,
+      "step": 5624
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.0002292704081632653,
+      "loss": 0.2372,
+      "step": 5625
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022925510204081632,
+      "loss": 0.1653,
+      "step": 5626
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022923979591836732,
+      "loss": 0.2006,
+      "step": 5627
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022922448979591833,
+      "loss": 0.1495,
+      "step": 5628
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.00022920918367346939,
+      "loss": 0.212,
+      "step": 5629
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0002291938775510204,
+      "loss": 0.1465,
+      "step": 5630
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0002291785714285714,
+      "loss": 0.2103,
+      "step": 5631
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.00022916326530612245,
+      "loss": 0.2274,
+      "step": 5632
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.00022914795918367345,
+      "loss": 0.1941,
+      "step": 5633
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.00022913265306122446,
+      "loss": 0.1693,
+      "step": 5634
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0002291173469387755,
+      "loss": 0.1222,
+      "step": 5635
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.00022910204081632652,
+      "loss": 0.1092,
+      "step": 5636
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.00022908673469387755,
+      "loss": 0.1559,
+      "step": 5637
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.00022907142857142855,
+      "loss": 0.2748,
+      "step": 5638
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.00022905612244897956,
+      "loss": 0.1385,
+      "step": 5639
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.0002290408163265306,
+      "loss": 0.108,
+      "step": 5640
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022902551020408162,
+      "loss": 0.1099,
+      "step": 5641
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022901020408163262,
+      "loss": 0.1513,
+      "step": 5642
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022899489795918365,
+      "loss": 0.1515,
+      "step": 5643
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022897959183673468,
+      "loss": 0.1818,
+      "step": 5644
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.0002289642857142857,
+      "loss": 0.1922,
+      "step": 5645
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022894897959183671,
+      "loss": 0.1717,
+      "step": 5646
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022893367346938772,
+      "loss": 0.1727,
+      "step": 5647
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022891836734693878,
+      "loss": 0.109,
+      "step": 5648
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022890306122448978,
+      "loss": 0.1454,
+      "step": 5649
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022888775510204078,
+      "loss": 0.2409,
+      "step": 5650
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.0002288724489795918,
+      "loss": 0.2515,
+      "step": 5651
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022885714285714284,
+      "loss": 0.1316,
+      "step": 5652
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022884183673469387,
+      "loss": 0.1548,
+      "step": 5653
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022882653061224488,
+      "loss": 0.1133,
+      "step": 5654
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022881122448979588,
+      "loss": 0.1725,
+      "step": 5655
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022879591836734694,
+      "loss": 0.1553,
+      "step": 5656
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022878061224489794,
+      "loss": 0.2123,
+      "step": 5657
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022876530612244895,
+      "loss": 0.2083,
+      "step": 5658
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00022874999999999998,
+      "loss": 0.1627,
+      "step": 5659
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.000228734693877551,
+      "loss": 0.1505,
+      "step": 5660
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.000228719387755102,
+      "loss": 0.143,
+      "step": 5661
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.00022870408163265304,
+      "loss": 0.1877,
+      "step": 5662
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.00022868877551020404,
+      "loss": 0.1884,
+      "step": 5663
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0002286734693877551,
+      "loss": 0.1873,
+      "step": 5664
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0002286581632653061,
+      "loss": 0.1202,
+      "step": 5665
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0002286428571428571,
+      "loss": 0.1313,
+      "step": 5666
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.00022862755102040814,
+      "loss": 0.1677,
+      "step": 5667
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.00022861224489795917,
+      "loss": 0.1586,
+      "step": 5668
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.00022859693877551017,
+      "loss": 0.2123,
+      "step": 5669
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0002285816326530612,
+      "loss": 0.1652,
+      "step": 5670
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.0002285663265306122,
+      "loss": 0.1633,
+      "step": 5671
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00022855102040816327,
+      "loss": 0.2198,
+      "step": 5672
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00022853571428571427,
+      "loss": 0.1546,
+      "step": 5673
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00022852040816326527,
+      "loss": 0.097,
+      "step": 5674
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.0002285051020408163,
+      "loss": 0.3194,
+      "step": 5675
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00022848979591836733,
+      "loss": 0.188,
+      "step": 5676
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00022847448979591834,
+      "loss": 0.1419,
+      "step": 5677
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00022845918367346937,
+      "loss": 0.1529,
+      "step": 5678
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.0002284438775510204,
+      "loss": 0.1362,
+      "step": 5679
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00022842857142857143,
+      "loss": 0.1109,
+      "step": 5680
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.00022841326530612243,
+      "loss": 0.1812,
+      "step": 5681
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.00022839795918367344,
+      "loss": 0.2879,
+      "step": 5682
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0002283826530612245,
+      "loss": 0.1212,
+      "step": 5683
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0002283673469387755,
+      "loss": 0.1218,
+      "step": 5684
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0002283520408163265,
+      "loss": 0.1619,
+      "step": 5685
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.00022833673469387753,
+      "loss": 0.186,
+      "step": 5686
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.00022832142857142856,
+      "loss": 0.1985,
+      "step": 5687
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0002283061224489796,
+      "loss": 0.3336,
+      "step": 5688
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0002282908163265306,
+      "loss": 0.1836,
+      "step": 5689
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0002282755102040816,
+      "loss": 0.1969,
+      "step": 5690
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022826020408163266,
+      "loss": 0.1893,
+      "step": 5691
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022824489795918366,
+      "loss": 0.1751,
+      "step": 5692
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022822959183673466,
+      "loss": 0.1471,
+      "step": 5693
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.0002282142857142857,
+      "loss": 0.2655,
+      "step": 5694
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022819897959183672,
+      "loss": 0.3141,
+      "step": 5695
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022818367346938773,
+      "loss": 0.1607,
+      "step": 5696
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022816836734693876,
+      "loss": 0.1481,
+      "step": 5697
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022815306122448976,
+      "loss": 0.1875,
+      "step": 5698
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022813775510204082,
+      "loss": 0.1474,
+      "step": 5699
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022812244897959182,
+      "loss": 0.2276,
+      "step": 5700
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00022810714285714283,
+      "loss": 0.1624,
+      "step": 5701
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00022809183673469386,
+      "loss": 0.1824,
+      "step": 5702
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.0002280765306122449,
+      "loss": 0.1791,
+      "step": 5703
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.0002280612244897959,
+      "loss": 0.2045,
+      "step": 5704
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00022804591836734692,
+      "loss": 0.1057,
+      "step": 5705
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00022803061224489792,
+      "loss": 0.138,
+      "step": 5706
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00022801530612244898,
+      "loss": 0.1725,
+      "step": 5707
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 0.1386,
+      "step": 5708
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.000227984693877551,
+      "loss": 0.1441,
+      "step": 5709
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00022796938775510202,
+      "loss": 0.2081,
+      "step": 5710
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00022795408163265305,
+      "loss": 0.1606,
+      "step": 5711
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00022793877551020405,
+      "loss": 0.1328,
+      "step": 5712
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00022792346938775508,
+      "loss": 0.2191,
+      "step": 5713
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0002279081632653061,
+      "loss": 0.1318,
+      "step": 5714
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00022789285714285715,
+      "loss": 0.1573,
+      "step": 5715
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00022787755102040815,
+      "loss": 0.1158,
+      "step": 5716
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00022786224489795915,
+      "loss": 0.1465,
+      "step": 5717
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00022784693877551016,
+      "loss": 0.1613,
+      "step": 5718
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0002278316326530612,
+      "loss": 0.29,
+      "step": 5719
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00022781632653061222,
+      "loss": 0.2512,
+      "step": 5720
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00022780102040816325,
+      "loss": 0.191,
+      "step": 5721
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022778571428571425,
+      "loss": 0.1916,
+      "step": 5722
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022777040816326528,
+      "loss": 0.255,
+      "step": 5723
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.0002277551020408163,
+      "loss": 0.1324,
+      "step": 5724
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022773979591836731,
+      "loss": 0.2212,
+      "step": 5725
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022772448979591837,
+      "loss": 0.186,
+      "step": 5726
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022770918367346938,
+      "loss": 0.1161,
+      "step": 5727
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022769387755102038,
+      "loss": 0.1269,
+      "step": 5728
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.0002276785714285714,
+      "loss": 0.1371,
+      "step": 5729
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022766326530612244,
+      "loss": 0.1779,
+      "step": 5730
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022764795918367344,
+      "loss": 0.1221,
+      "step": 5731
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022763265306122447,
+      "loss": 0.2052,
+      "step": 5732
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022761734693877548,
+      "loss": 0.2008,
+      "step": 5733
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022760204081632654,
+      "loss": 0.1071,
+      "step": 5734
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022758673469387754,
+      "loss": 0.1691,
+      "step": 5735
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022757142857142854,
+      "loss": 0.1196,
+      "step": 5736
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022755612244897957,
+      "loss": 0.1418,
+      "step": 5737
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.0002275408163265306,
+      "loss": 0.163,
+      "step": 5738
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.0002275255102040816,
+      "loss": 0.1812,
+      "step": 5739
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022751020408163264,
+      "loss": 0.1106,
+      "step": 5740
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022749489795918364,
+      "loss": 0.2201,
+      "step": 5741
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.0002274795918367347,
+      "loss": 0.1536,
+      "step": 5742
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.0002274642857142857,
+      "loss": 0.1299,
+      "step": 5743
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.0002274489795918367,
+      "loss": 0.2209,
+      "step": 5744
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00022743367346938774,
+      "loss": 0.2158,
+      "step": 5745
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00022741836734693877,
+      "loss": 0.1048,
+      "step": 5746
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00022740306122448977,
+      "loss": 0.1341,
+      "step": 5747
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.0002273877551020408,
+      "loss": 0.1557,
+      "step": 5748
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.0002273724489795918,
+      "loss": 0.1229,
+      "step": 5749
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00022735714285714286,
+      "loss": 0.1459,
+      "step": 5750
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00022734183673469387,
+      "loss": 0.2336,
+      "step": 5751
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00022732653061224487,
+      "loss": 0.2194,
+      "step": 5752
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022731122448979587,
+      "loss": 0.1117,
+      "step": 5753
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022729591836734693,
+      "loss": 0.1261,
+      "step": 5754
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022728061224489793,
+      "loss": 0.1621,
+      "step": 5755
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022726530612244896,
+      "loss": 0.2484,
+      "step": 5756
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022724999999999997,
+      "loss": 0.2596,
+      "step": 5757
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.000227234693877551,
+      "loss": 0.2314,
+      "step": 5758
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022721938775510203,
+      "loss": 0.16,
+      "step": 5759
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022720408163265303,
+      "loss": 0.1675,
+      "step": 5760
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022718877551020404,
+      "loss": 0.1859,
+      "step": 5761
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.0002271734693877551,
+      "loss": 0.1638,
+      "step": 5762
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0002271581632653061,
+      "loss": 0.2738,
+      "step": 5763
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.00022714285714285713,
+      "loss": 0.2275,
+      "step": 5764
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.00022712755102040813,
+      "loss": 0.2484,
+      "step": 5765
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.00022711224489795916,
+      "loss": 0.1544,
+      "step": 5766
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0002270969387755102,
+      "loss": 0.1684,
+      "step": 5767
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0002270816326530612,
+      "loss": 0.1055,
+      "step": 5768
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.00022706632653061225,
+      "loss": 0.2481,
+      "step": 5769
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.00022705102040816326,
+      "loss": 0.219,
+      "step": 5770
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.00022703571428571426,
+      "loss": 0.1531,
+      "step": 5771
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0002270204081632653,
+      "loss": 0.1496,
+      "step": 5772
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.00022700510204081632,
+      "loss": 0.1535,
+      "step": 5773
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022698979591836732,
+      "loss": 0.1552,
+      "step": 5774
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022697448979591835,
+      "loss": 0.1779,
+      "step": 5775
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022695918367346936,
+      "loss": 0.1999,
+      "step": 5776
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022694387755102042,
+      "loss": 0.1097,
+      "step": 5777
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022692857142857142,
+      "loss": 0.127,
+      "step": 5778
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022691326530612242,
+      "loss": 0.165,
+      "step": 5779
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022689795918367343,
+      "loss": 0.1283,
+      "step": 5780
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022688265306122448,
+      "loss": 0.274,
+      "step": 5781
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0002268673469387755,
+      "loss": 0.2611,
+      "step": 5782
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022685204081632652,
+      "loss": 0.1687,
+      "step": 5783
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022683673469387752,
+      "loss": 0.1171,
+      "step": 5784
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022682142857142858,
+      "loss": 0.14,
+      "step": 5785
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022680612244897958,
+      "loss": 0.1488,
+      "step": 5786
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022679081632653059,
+      "loss": 0.1741,
+      "step": 5787
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.0002267755102040816,
+      "loss": 0.1873,
+      "step": 5788
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022676020408163265,
+      "loss": 0.1709,
+      "step": 5789
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022674489795918365,
+      "loss": 0.1135,
+      "step": 5790
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022672959183673468,
+      "loss": 0.1118,
+      "step": 5791
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022671428571428568,
+      "loss": 0.1515,
+      "step": 5792
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00022669897959183671,
+      "loss": 0.1506,
+      "step": 5793
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.00022668367346938775,
+      "loss": 0.282,
+      "step": 5794
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.00022666836734693875,
+      "loss": 0.1937,
+      "step": 5795
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.00022665306122448975,
+      "loss": 0.1848,
+      "step": 5796
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0002266377551020408,
+      "loss": 0.1072,
+      "step": 5797
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0002266224489795918,
+      "loss": 0.2205,
+      "step": 5798
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.00022660714285714284,
+      "loss": 0.1105,
+      "step": 5799
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.00022659183673469385,
+      "loss": 0.1994,
+      "step": 5800
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.00022657653061224488,
+      "loss": 0.269,
+      "step": 5801
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0002265612244897959,
+      "loss": 0.1743,
+      "step": 5802
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0002265459183673469,
+      "loss": 0.1089,
+      "step": 5803
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.00022653061224489791,
+      "loss": 0.1487,
+      "step": 5804
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022651530612244897,
+      "loss": 0.1756,
+      "step": 5805
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022649999999999998,
+      "loss": 0.1843,
+      "step": 5806
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.000226484693877551,
+      "loss": 0.2903,
+      "step": 5807
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.000226469387755102,
+      "loss": 0.2066,
+      "step": 5808
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022645408163265304,
+      "loss": 0.2443,
+      "step": 5809
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022643877551020407,
+      "loss": 0.1943,
+      "step": 5810
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022642346938775507,
+      "loss": 0.2092,
+      "step": 5811
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022640816326530608,
+      "loss": 0.12,
+      "step": 5812
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022639285714285714,
+      "loss": 0.2355,
+      "step": 5813
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022637755102040814,
+      "loss": 0.1336,
+      "step": 5814
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00022636224489795914,
+      "loss": 0.2622,
+      "step": 5815
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.0002263469387755102,
+      "loss": 0.1584,
+      "step": 5816
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.0002263316326530612,
+      "loss": 0.1381,
+      "step": 5817
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00022631632653061223,
+      "loss": 0.1145,
+      "step": 5818
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00022630102040816324,
+      "loss": 0.2413,
+      "step": 5819
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00022628571428571427,
+      "loss": 0.2443,
+      "step": 5820
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.0002262704081632653,
+      "loss": 0.1158,
+      "step": 5821
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.0002262551020408163,
+      "loss": 0.1772,
+      "step": 5822
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.0002262397959183673,
+      "loss": 0.0974,
+      "step": 5823
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00022622448979591836,
+      "loss": 0.1393,
+      "step": 5824
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00022620918367346937,
+      "loss": 0.1351,
+      "step": 5825
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.0002261938775510204,
+      "loss": 0.1764,
+      "step": 5826
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.0002261785714285714,
+      "loss": 0.1987,
+      "step": 5827
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00022616326530612243,
+      "loss": 0.097,
+      "step": 5828
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00022614795918367346,
+      "loss": 0.1117,
+      "step": 5829
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00022613265306122447,
+      "loss": 0.139,
+      "step": 5830
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00022611734693877547,
+      "loss": 0.1623,
+      "step": 5831
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00022610204081632653,
+      "loss": 0.2246,
+      "step": 5832
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00022608673469387753,
+      "loss": 0.2449,
+      "step": 5833
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00022607142857142856,
+      "loss": 0.2871,
+      "step": 5834
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.00022605612244897956,
+      "loss": 0.1361,
+      "step": 5835
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0002260408163265306,
+      "loss": 0.131,
+      "step": 5836
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.00022602551020408162,
+      "loss": 0.1052,
+      "step": 5837
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.00022601020408163263,
+      "loss": 0.2333,
+      "step": 5838
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.00022599489795918363,
+      "loss": 0.1417,
+      "step": 5839
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0002259795918367347,
+      "loss": 0.1166,
+      "step": 5840
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0002259642857142857,
+      "loss": 0.0996,
+      "step": 5841
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0002259489795918367,
+      "loss": 0.1914,
+      "step": 5842
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.00022593367346938773,
+      "loss": 0.2197,
+      "step": 5843
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.00022591836734693876,
+      "loss": 0.1822,
+      "step": 5844
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0002259030612244898,
+      "loss": 0.2117,
+      "step": 5845
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0002258877551020408,
+      "loss": 0.2142,
+      "step": 5846
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0002258724489795918,
+      "loss": 0.1822,
+      "step": 5847
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00022585714285714285,
+      "loss": 0.1227,
+      "step": 5848
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00022584183673469386,
+      "loss": 0.1104,
+      "step": 5849
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00022582653061224486,
+      "loss": 0.1865,
+      "step": 5850
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0002258112244897959,
+      "loss": 0.3703,
+      "step": 5851
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00022579591836734692,
+      "loss": 0.2198,
+      "step": 5852
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00022578061224489795,
+      "loss": 0.1177,
+      "step": 5853
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00022576530612244895,
+      "loss": 0.168,
+      "step": 5854
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00022574999999999996,
+      "loss": 0.1855,
+      "step": 5855
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022573469387755102,
+      "loss": 0.2816,
+      "step": 5856
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022571938775510202,
+      "loss": 0.326,
+      "step": 5857
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022570408163265302,
+      "loss": 0.1826,
+      "step": 5858
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022568877551020408,
+      "loss": 0.1046,
+      "step": 5859
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022567346938775508,
+      "loss": 0.2122,
+      "step": 5860
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022565816326530611,
+      "loss": 0.1804,
+      "step": 5861
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022564285714285712,
+      "loss": 0.1359,
+      "step": 5862
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022562755102040815,
+      "loss": 0.2217,
+      "step": 5863
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022561224489795918,
+      "loss": 0.2233,
+      "step": 5864
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022559693877551018,
+      "loss": 0.2227,
+      "step": 5865
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00022558163265306119,
+      "loss": 0.1977,
+      "step": 5866
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00022556632653061224,
+      "loss": 0.1784,
+      "step": 5867
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00022555102040816325,
+      "loss": 0.1779,
+      "step": 5868
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00022553571428571428,
+      "loss": 0.2693,
+      "step": 5869
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00022552040816326528,
+      "loss": 0.1417,
+      "step": 5870
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0002255051020408163,
+      "loss": 0.2097,
+      "step": 5871
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00022548979591836734,
+      "loss": 0.1744,
+      "step": 5872
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00022547448979591835,
+      "loss": 0.1485,
+      "step": 5873
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00022545918367346935,
+      "loss": 0.154,
+      "step": 5874
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0002254438775510204,
+      "loss": 0.1436,
+      "step": 5875
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0002254285714285714,
+      "loss": 0.3155,
+      "step": 5876
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0002254132653061224,
+      "loss": 0.1712,
+      "step": 5877
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.00022539795918367344,
+      "loss": 0.1072,
+      "step": 5878
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.00022538265306122447,
+      "loss": 0.2071,
+      "step": 5879
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0002253673469387755,
+      "loss": 0.1625,
+      "step": 5880
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0002253520408163265,
+      "loss": 0.1507,
+      "step": 5881
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0002253367346938775,
+      "loss": 0.3699,
+      "step": 5882
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.00022532142857142857,
+      "loss": 0.1623,
+      "step": 5883
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.00022530612244897957,
+      "loss": 0.1399,
+      "step": 5884
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.00022529081632653058,
+      "loss": 0.2264,
+      "step": 5885
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0002252755102040816,
+      "loss": 0.1137,
+      "step": 5886
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022526020408163264,
+      "loss": 0.1203,
+      "step": 5887
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022524489795918367,
+      "loss": 0.2773,
+      "step": 5888
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022522959183673467,
+      "loss": 0.1561,
+      "step": 5889
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022521428571428567,
+      "loss": 0.1704,
+      "step": 5890
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022519897959183673,
+      "loss": 0.1402,
+      "step": 5891
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022518367346938774,
+      "loss": 0.1709,
+      "step": 5892
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022516836734693874,
+      "loss": 0.1197,
+      "step": 5893
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022515306122448977,
+      "loss": 0.3783,
+      "step": 5894
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.0002251377551020408,
+      "loss": 0.3206,
+      "step": 5895
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00022512244897959183,
+      "loss": 0.1306,
+      "step": 5896
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00022510714285714283,
+      "loss": 0.1824,
+      "step": 5897
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00022509183673469384,
+      "loss": 0.0996,
+      "step": 5898
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0002250765306122449,
+      "loss": 0.1221,
+      "step": 5899
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0002250612244897959,
+      "loss": 0.1741,
+      "step": 5900
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0002250459183673469,
+      "loss": 0.1868,
+      "step": 5901
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00022503061224489793,
+      "loss": 0.2349,
+      "step": 5902
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00022501530612244896,
+      "loss": 0.1785,
+      "step": 5903
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.000225,
+      "loss": 0.1659,
+      "step": 5904
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.000224984693877551,
+      "loss": 0.128,
+      "step": 5905
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00022496938775510203,
+      "loss": 0.1506,
+      "step": 5906
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00022495408163265306,
+      "loss": 0.2888,
+      "step": 5907
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022493877551020406,
+      "loss": 0.1125,
+      "step": 5908
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022492346938775507,
+      "loss": 0.2015,
+      "step": 5909
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022490816326530612,
+      "loss": 0.1974,
+      "step": 5910
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022489285714285713,
+      "loss": 0.1433,
+      "step": 5911
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022487755102040813,
+      "loss": 0.213,
+      "step": 5912
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022486224489795916,
+      "loss": 0.2541,
+      "step": 5913
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.0002248469387755102,
+      "loss": 0.1361,
+      "step": 5914
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022483163265306122,
+      "loss": 0.1337,
+      "step": 5915
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022481632653061222,
+      "loss": 0.257,
+      "step": 5916
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00022480102040816323,
+      "loss": 0.1168,
+      "step": 5917
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.00022478571428571429,
+      "loss": 0.0994,
+      "step": 5918
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0002247704081632653,
+      "loss": 0.2813,
+      "step": 5919
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0002247551020408163,
+      "loss": 0.3053,
+      "step": 5920
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.00022473979591836732,
+      "loss": 0.1742,
+      "step": 5921
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.00022472448979591835,
+      "loss": 0.1186,
+      "step": 5922
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.00022470918367346938,
+      "loss": 0.1345,
+      "step": 5923
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0002246938775510204,
+      "loss": 0.1758,
+      "step": 5924
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0002246785714285714,
+      "loss": 0.1172,
+      "step": 5925
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.00022466326530612245,
+      "loss": 0.2195,
+      "step": 5926
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.00022464795918367345,
+      "loss": 0.1392,
+      "step": 5927
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022463265306122446,
+      "loss": 0.1293,
+      "step": 5928
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022461734693877549,
+      "loss": 0.1865,
+      "step": 5929
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022460204081632652,
+      "loss": 0.211,
+      "step": 5930
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022458673469387755,
+      "loss": 0.2056,
+      "step": 5931
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022457142857142855,
+      "loss": 0.2119,
+      "step": 5932
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022455612244897955,
+      "loss": 0.2069,
+      "step": 5933
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.0002245408163265306,
+      "loss": 0.1665,
+      "step": 5934
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022452551020408162,
+      "loss": 0.1181,
+      "step": 5935
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022451020408163262,
+      "loss": 0.1982,
+      "step": 5936
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00022449489795918365,
+      "loss": 0.144,
+      "step": 5937
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022447959183673468,
+      "loss": 0.2812,
+      "step": 5938
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022446428571428568,
+      "loss": 0.3772,
+      "step": 5939
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022444897959183671,
+      "loss": 0.2107,
+      "step": 5940
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022443367346938772,
+      "loss": 0.1591,
+      "step": 5941
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022441836734693878,
+      "loss": 0.1549,
+      "step": 5942
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022440306122448978,
+      "loss": 0.1851,
+      "step": 5943
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022438775510204078,
+      "loss": 0.3594,
+      "step": 5944
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.0002243724489795918,
+      "loss": 0.1864,
+      "step": 5945
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022435714285714284,
+      "loss": 0.1254,
+      "step": 5946
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022434183673469385,
+      "loss": 0.2293,
+      "step": 5947
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00022432653061224488,
+      "loss": 0.2282,
+      "step": 5948
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00022431122448979588,
+      "loss": 0.0824,
+      "step": 5949
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00022429591836734694,
+      "loss": 0.1931,
+      "step": 5950
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00022428061224489794,
+      "loss": 0.3275,
+      "step": 5951
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00022426530612244895,
+      "loss": 0.176,
+      "step": 5952
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00022425,
+      "loss": 0.1352,
+      "step": 5953
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.000224234693877551,
+      "loss": 0.2578,
+      "step": 5954
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.000224219387755102,
+      "loss": 0.2524,
+      "step": 5955
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00022420408163265304,
+      "loss": 0.1896,
+      "step": 5956
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00022418877551020407,
+      "loss": 0.1516,
+      "step": 5957
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.0002241734693877551,
+      "loss": 0.1982,
+      "step": 5958
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.0002241581632653061,
+      "loss": 0.1792,
+      "step": 5959
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.0002241428571428571,
+      "loss": 0.1316,
+      "step": 5960
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00022412755102040817,
+      "loss": 0.1802,
+      "step": 5961
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00022411224489795917,
+      "loss": 0.1113,
+      "step": 5962
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00022409693877551017,
+      "loss": 0.3105,
+      "step": 5963
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.0002240816326530612,
+      "loss": 0.2205,
+      "step": 5964
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00022406632653061223,
+      "loss": 0.1118,
+      "step": 5965
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00022405102040816326,
+      "loss": 0.1265,
+      "step": 5966
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00022403571428571427,
+      "loss": 0.1945,
+      "step": 5967
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00022402040816326527,
+      "loss": 0.1022,
+      "step": 5968
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00022400510204081633,
+      "loss": 0.3007,
+      "step": 5969
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00022398979591836733,
+      "loss": 0.1626,
+      "step": 5970
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00022397448979591834,
+      "loss": 0.2654,
+      "step": 5971
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00022395918367346937,
+      "loss": 0.1834,
+      "step": 5972
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0002239438775510204,
+      "loss": 0.1025,
+      "step": 5973
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0002239285714285714,
+      "loss": 0.2116,
+      "step": 5974
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00022391326530612243,
+      "loss": 0.2526,
+      "step": 5975
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00022389795918367343,
+      "loss": 0.2451,
+      "step": 5976
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0002238826530612245,
+      "loss": 0.1632,
+      "step": 5977
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0002238673469387755,
+      "loss": 0.1443,
+      "step": 5978
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0002238520408163265,
+      "loss": 0.183,
+      "step": 5979
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00022383673469387753,
+      "loss": 0.1132,
+      "step": 5980
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00022382142857142856,
+      "loss": 0.2109,
+      "step": 5981
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00022380612244897956,
+      "loss": 0.2351,
+      "step": 5982
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.0002237908163265306,
+      "loss": 0.1839,
+      "step": 5983
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.0002237755102040816,
+      "loss": 0.2428,
+      "step": 5984
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00022376020408163265,
+      "loss": 0.2753,
+      "step": 5985
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00022374489795918366,
+      "loss": 0.1593,
+      "step": 5986
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00022372959183673466,
+      "loss": 0.1274,
+      "step": 5987
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.0002237142857142857,
+      "loss": 0.2316,
+      "step": 5988
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00022369897959183672,
+      "loss": 0.1736,
+      "step": 5989
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022368367346938773,
+      "loss": 0.1481,
+      "step": 5990
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022366836734693876,
+      "loss": 0.1645,
+      "step": 5991
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022365306122448976,
+      "loss": 0.1391,
+      "step": 5992
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022363775510204082,
+      "loss": 0.1118,
+      "step": 5993
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022362244897959182,
+      "loss": 0.2284,
+      "step": 5994
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022360714285714282,
+      "loss": 0.2375,
+      "step": 5995
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022359183673469388,
+      "loss": 0.1614,
+      "step": 5996
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022357653061224489,
+      "loss": 0.1247,
+      "step": 5997
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.0002235612244897959,
+      "loss": 0.1297,
+      "step": 5998
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00022354591836734692,
+      "loss": 0.1672,
+      "step": 5999
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022353061224489795,
+      "loss": 0.2301,
+      "step": 6000
+    },
+    {
+      "epoch": 5.83,
+      "eval_cer": 0.2735236291572634,
+      "eval_loss": 1.0216587781906128,
+      "eval_runtime": 230.9176,
+      "eval_samples_per_second": 38.169,
+      "eval_steps_per_second": 4.772,
+      "eval_wer": 0.6719232332553362,
+      "step": 6000
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022351530612244895,
+      "loss": 0.2079,
+      "step": 6001
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022349999999999998,
+      "loss": 0.1926,
+      "step": 6002
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.000223484693877551,
+      "loss": 0.1428,
+      "step": 6003
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022346938775510205,
+      "loss": 0.1607,
+      "step": 6004
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022345408163265305,
+      "loss": 0.1605,
+      "step": 6005
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022343877551020405,
+      "loss": 0.2239,
+      "step": 6006
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022342346938775508,
+      "loss": 0.1735,
+      "step": 6007
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022340816326530611,
+      "loss": 0.2269,
+      "step": 6008
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022339285714285712,
+      "loss": 0.1559,
+      "step": 6009
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00022337755102040815,
+      "loss": 0.2765,
+      "step": 6010
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00022336224489795915,
+      "loss": 0.1575,
+      "step": 6011
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.0002233469387755102,
+      "loss": 0.1428,
+      "step": 6012
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.0002233316326530612,
+      "loss": 0.3163,
+      "step": 6013
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00022331632653061222,
+      "loss": 0.1713,
+      "step": 6014
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00022330102040816325,
+      "loss": 0.1577,
+      "step": 6015
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00022328571428571428,
+      "loss": 0.1151,
+      "step": 6016
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00022327040816326528,
+      "loss": 0.1318,
+      "step": 6017
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.0002232551020408163,
+      "loss": 0.1997,
+      "step": 6018
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00022323979591836731,
+      "loss": 0.3238,
+      "step": 6019
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00022322448979591837,
+      "loss": 0.2095,
+      "step": 6020
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022320918367346938,
+      "loss": 0.1905,
+      "step": 6021
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022319387755102038,
+      "loss": 0.1331,
+      "step": 6022
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022317857142857138,
+      "loss": 0.1949,
+      "step": 6023
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022316326530612244,
+      "loss": 0.1439,
+      "step": 6024
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022314795918367344,
+      "loss": 0.2257,
+      "step": 6025
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022313265306122447,
+      "loss": 0.1616,
+      "step": 6026
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022311734693877548,
+      "loss": 0.1878,
+      "step": 6027
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022310204081632653,
+      "loss": 0.1768,
+      "step": 6028
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022308673469387754,
+      "loss": 0.1748,
+      "step": 6029
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00022307142857142854,
+      "loss": 0.1806,
+      "step": 6030
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.00022305612244897955,
+      "loss": 0.1355,
+      "step": 6031
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0002230408163265306,
+      "loss": 0.3034,
+      "step": 6032
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0002230255102040816,
+      "loss": 0.1861,
+      "step": 6033
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.00022301020408163264,
+      "loss": 0.163,
+      "step": 6034
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.00022299489795918364,
+      "loss": 0.1078,
+      "step": 6035
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.00022297959183673467,
+      "loss": 0.1158,
+      "step": 6036
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0002229642857142857,
+      "loss": 0.1334,
+      "step": 6037
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0002229489795918367,
+      "loss": 0.3704,
+      "step": 6038
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0002229336734693877,
+      "loss": 0.2521,
+      "step": 6039
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.00022291836734693877,
+      "loss": 0.1329,
+      "step": 6040
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022290306122448977,
+      "loss": 0.2116,
+      "step": 6041
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0002228877551020408,
+      "loss": 0.2595,
+      "step": 6042
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022287244897959183,
+      "loss": 0.165,
+      "step": 6043
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022285714285714283,
+      "loss": 0.1762,
+      "step": 6044
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022284183673469386,
+      "loss": 0.1864,
+      "step": 6045
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022282653061224487,
+      "loss": 0.2222,
+      "step": 6046
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022281122448979593,
+      "loss": 0.1125,
+      "step": 6047
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022279591836734693,
+      "loss": 0.1518,
+      "step": 6048
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022278061224489793,
+      "loss": 0.168,
+      "step": 6049
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022276530612244896,
+      "loss": 0.36,
+      "step": 6050
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00022275,
+      "loss": 0.1497,
+      "step": 6051
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.000222734693877551,
+      "loss": 0.2331,
+      "step": 6052
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.00022271938775510203,
+      "loss": 0.3236,
+      "step": 6053
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.00022270408163265303,
+      "loss": 0.1989,
+      "step": 6054
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0002226887755102041,
+      "loss": 0.1207,
+      "step": 6055
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0002226734693877551,
+      "loss": 0.1644,
+      "step": 6056
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0002226581632653061,
+      "loss": 0.225,
+      "step": 6057
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0002226428571428571,
+      "loss": 0.1776,
+      "step": 6058
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.00022262755102040816,
+      "loss": 0.2094,
+      "step": 6059
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.00022261224489795916,
+      "loss": 0.172,
+      "step": 6060
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0002225969387755102,
+      "loss": 0.1136,
+      "step": 6061
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.0002225816326530612,
+      "loss": 0.0901,
+      "step": 6062
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00022256632653061222,
+      "loss": 0.2015,
+      "step": 6063
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00022255102040816325,
+      "loss": 0.242,
+      "step": 6064
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00022253571428571426,
+      "loss": 0.1606,
+      "step": 6065
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00022252040816326526,
+      "loss": 0.1601,
+      "step": 6066
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00022250510204081632,
+      "loss": 0.13,
+      "step": 6067
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00022248979591836732,
+      "loss": 0.1511,
+      "step": 6068
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00022247448979591835,
+      "loss": 0.3119,
+      "step": 6069
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00022245918367346936,
+      "loss": 0.1432,
+      "step": 6070
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.0002224438775510204,
+      "loss": 0.1892,
+      "step": 6071
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022242857142857142,
+      "loss": 0.187,
+      "step": 6072
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022241326530612242,
+      "loss": 0.136,
+      "step": 6073
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022239795918367342,
+      "loss": 0.1827,
+      "step": 6074
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022238265306122448,
+      "loss": 0.2292,
+      "step": 6075
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022236734693877549,
+      "loss": 0.2503,
+      "step": 6076
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022235204081632652,
+      "loss": 0.262,
+      "step": 6077
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022233673469387752,
+      "loss": 0.2458,
+      "step": 6078
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022232142857142855,
+      "loss": 0.1996,
+      "step": 6079
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022230612244897958,
+      "loss": 0.1532,
+      "step": 6080
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00022229081632653058,
+      "loss": 0.1761,
+      "step": 6081
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.0002222755102040816,
+      "loss": 0.209,
+      "step": 6082
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.00022226020408163265,
+      "loss": 0.1865,
+      "step": 6083
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.00022224489795918365,
+      "loss": 0.1497,
+      "step": 6084
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.00022222959183673468,
+      "loss": 0.2214,
+      "step": 6085
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0002222142857142857,
+      "loss": 0.16,
+      "step": 6086
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.00022219897959183671,
+      "loss": 0.1177,
+      "step": 6087
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.00022218367346938774,
+      "loss": 0.2819,
+      "step": 6088
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.00022216836734693875,
+      "loss": 0.2898,
+      "step": 6089
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0002221530612244898,
+      "loss": 0.2076,
+      "step": 6090
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0002221377551020408,
+      "loss": 0.2164,
+      "step": 6091
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0002221224489795918,
+      "loss": 0.1466,
+      "step": 6092
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00022210714285714282,
+      "loss": 0.1236,
+      "step": 6093
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00022209183673469387,
+      "loss": 0.1882,
+      "step": 6094
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00022207653061224488,
+      "loss": 0.2497,
+      "step": 6095
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.0002220612244897959,
+      "loss": 0.1646,
+      "step": 6096
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.0002220459183673469,
+      "loss": 0.1744,
+      "step": 6097
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00022203061224489794,
+      "loss": 0.1398,
+      "step": 6098
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00022201530612244897,
+      "loss": 0.1937,
+      "step": 6099
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00022199999999999998,
+      "loss": 0.1958,
+      "step": 6100
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00022198469387755098,
+      "loss": 0.1939,
+      "step": 6101
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00022196938775510204,
+      "loss": 0.1347,
+      "step": 6102
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00022195408163265304,
+      "loss": 0.1717,
+      "step": 6103
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00022193877551020407,
+      "loss": 0.2238,
+      "step": 6104
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00022192346938775507,
+      "loss": 0.2256,
+      "step": 6105
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.0002219081632653061,
+      "loss": 0.1233,
+      "step": 6106
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00022189285714285713,
+      "loss": 0.2511,
+      "step": 6107
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00022187755102040814,
+      "loss": 0.2563,
+      "step": 6108
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00022186224489795914,
+      "loss": 0.2315,
+      "step": 6109
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.0002218469387755102,
+      "loss": 0.1874,
+      "step": 6110
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.0002218316326530612,
+      "loss": 0.2249,
+      "step": 6111
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00022181632653061223,
+      "loss": 0.2227,
+      "step": 6112
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00022180102040816324,
+      "loss": 0.2477,
+      "step": 6113
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00022178571428571427,
+      "loss": 0.2539,
+      "step": 6114
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0002217704081632653,
+      "loss": 0.162,
+      "step": 6115
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0002217551020408163,
+      "loss": 0.1791,
+      "step": 6116
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0002217397959183673,
+      "loss": 0.2405,
+      "step": 6117
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00022172448979591836,
+      "loss": 0.1556,
+      "step": 6118
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00022170918367346937,
+      "loss": 0.3229,
+      "step": 6119
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00022169387755102037,
+      "loss": 0.2594,
+      "step": 6120
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0002216785714285714,
+      "loss": 0.1637,
+      "step": 6121
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00022166326530612243,
+      "loss": 0.2027,
+      "step": 6122
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00022164795918367346,
+      "loss": 0.2046,
+      "step": 6123
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022163265306122446,
+      "loss": 0.1975,
+      "step": 6124
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022161734693877547,
+      "loss": 0.3431,
+      "step": 6125
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022160204081632653,
+      "loss": 0.2779,
+      "step": 6126
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022158673469387753,
+      "loss": 0.1465,
+      "step": 6127
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022157142857142853,
+      "loss": 0.1943,
+      "step": 6128
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022155612244897956,
+      "loss": 0.1125,
+      "step": 6129
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.0002215408163265306,
+      "loss": 0.1739,
+      "step": 6130
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022152551020408162,
+      "loss": 0.2027,
+      "step": 6131
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022151020408163263,
+      "loss": 0.2148,
+      "step": 6132
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00022149489795918366,
+      "loss": 0.1992,
+      "step": 6133
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0002214795918367347,
+      "loss": 0.2126,
+      "step": 6134
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0002214642857142857,
+      "loss": 0.1693,
+      "step": 6135
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0002214489795918367,
+      "loss": 0.1722,
+      "step": 6136
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.00022143367346938775,
+      "loss": 0.2076,
+      "step": 6137
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.00022141836734693876,
+      "loss": 0.3692,
+      "step": 6138
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0002214030612244898,
+      "loss": 0.2199,
+      "step": 6139
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0002213877551020408,
+      "loss": 0.1325,
+      "step": 6140
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.00022137244897959182,
+      "loss": 0.2393,
+      "step": 6141
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.00022135714285714285,
+      "loss": 0.1511,
+      "step": 6142
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.00022134183673469385,
+      "loss": 0.1188,
+      "step": 6143
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022132653061224486,
+      "loss": 0.1808,
+      "step": 6144
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022131122448979592,
+      "loss": 0.2323,
+      "step": 6145
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022129591836734692,
+      "loss": 0.1778,
+      "step": 6146
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022128061224489795,
+      "loss": 0.1595,
+      "step": 6147
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022126530612244895,
+      "loss": 0.103,
+      "step": 6148
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022124999999999998,
+      "loss": 0.1284,
+      "step": 6149
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022123469387755101,
+      "loss": 0.2162,
+      "step": 6150
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022121938775510202,
+      "loss": 0.2704,
+      "step": 6151
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022120408163265302,
+      "loss": 0.228,
+      "step": 6152
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022118877551020408,
+      "loss": 0.1231,
+      "step": 6153
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00022117346938775508,
+      "loss": 0.1992,
+      "step": 6154
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022115816326530609,
+      "loss": 0.2232,
+      "step": 6155
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022114285714285712,
+      "loss": 0.1949,
+      "step": 6156
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022112755102040815,
+      "loss": 0.1534,
+      "step": 6157
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022111224489795918,
+      "loss": 0.1658,
+      "step": 6158
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022109693877551018,
+      "loss": 0.1245,
+      "step": 6159
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022108163265306118,
+      "loss": 0.2487,
+      "step": 6160
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022106632653061224,
+      "loss": 0.2537,
+      "step": 6161
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022105102040816325,
+      "loss": 0.2462,
+      "step": 6162
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022103571428571425,
+      "loss": 0.2519,
+      "step": 6163
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00022102040816326528,
+      "loss": 0.21,
+      "step": 6164
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0002210051020408163,
+      "loss": 0.1595,
+      "step": 6165
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.00022098979591836734,
+      "loss": 0.1244,
+      "step": 6166
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.00022097448979591834,
+      "loss": 0.148,
+      "step": 6167
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.00022095918367346935,
+      "loss": 0.254,
+      "step": 6168
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0002209438775510204,
+      "loss": 0.2146,
+      "step": 6169
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0002209285714285714,
+      "loss": 0.208,
+      "step": 6170
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0002209132653061224,
+      "loss": 0.1934,
+      "step": 6171
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.00022089795918367344,
+      "loss": 0.1671,
+      "step": 6172
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.00022088265306122447,
+      "loss": 0.1875,
+      "step": 6173
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0002208673469387755,
+      "loss": 0.1113,
+      "step": 6174
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.0002208520408163265,
+      "loss": 0.1716,
+      "step": 6175
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.0002208367346938775,
+      "loss": 0.1695,
+      "step": 6176
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00022082142857142857,
+      "loss": 0.189,
+      "step": 6177
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00022080612244897957,
+      "loss": 0.1803,
+      "step": 6178
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00022079081632653058,
+      "loss": 0.1172,
+      "step": 6179
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00022077551020408163,
+      "loss": 0.1672,
+      "step": 6180
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00022076020408163264,
+      "loss": 0.2828,
+      "step": 6181
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00022074489795918364,
+      "loss": 0.2552,
+      "step": 6182
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00022072959183673467,
+      "loss": 0.1066,
+      "step": 6183
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.0002207142857142857,
+      "loss": 0.0932,
+      "step": 6184
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00022069897959183673,
+      "loss": 0.1275,
+      "step": 6185
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.00022068367346938773,
+      "loss": 0.2177,
+      "step": 6186
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.00022066836734693874,
+      "loss": 0.3904,
+      "step": 6187
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0002206530612244898,
+      "loss": 0.1904,
+      "step": 6188
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0002206377551020408,
+      "loss": 0.1016,
+      "step": 6189
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0002206224489795918,
+      "loss": 0.1715,
+      "step": 6190
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.00022060714285714283,
+      "loss": 0.1994,
+      "step": 6191
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.00022059183673469386,
+      "loss": 0.0943,
+      "step": 6192
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0002205765306122449,
+      "loss": 0.2128,
+      "step": 6193
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0002205612244897959,
+      "loss": 0.1162,
+      "step": 6194
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0002205459183673469,
+      "loss": 0.1608,
+      "step": 6195
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022053061224489796,
+      "loss": 0.174,
+      "step": 6196
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022051530612244896,
+      "loss": 0.1119,
+      "step": 6197
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022049999999999997,
+      "loss": 0.0857,
+      "step": 6198
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.000220484693877551,
+      "loss": 0.2922,
+      "step": 6199
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022046938775510203,
+      "loss": 0.1644,
+      "step": 6200
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022045408163265306,
+      "loss": 0.101,
+      "step": 6201
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022043877551020406,
+      "loss": 0.1248,
+      "step": 6202
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022042346938775506,
+      "loss": 0.1155,
+      "step": 6203
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022040816326530612,
+      "loss": 0.1248,
+      "step": 6204
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00022039285714285713,
+      "loss": 0.163,
+      "step": 6205
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022037755102040813,
+      "loss": 0.126,
+      "step": 6206
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022036224489795916,
+      "loss": 0.1497,
+      "step": 6207
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.0002203469387755102,
+      "loss": 0.1022,
+      "step": 6208
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022033163265306122,
+      "loss": 0.1684,
+      "step": 6209
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022031632653061222,
+      "loss": 0.091,
+      "step": 6210
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022030102040816323,
+      "loss": 0.1871,
+      "step": 6211
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022028571428571429,
+      "loss": 0.1687,
+      "step": 6212
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.0002202704081632653,
+      "loss": 0.1271,
+      "step": 6213
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.0002202551020408163,
+      "loss": 0.1624,
+      "step": 6214
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022023979591836732,
+      "loss": 0.1514,
+      "step": 6215
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022022448979591835,
+      "loss": 0.0894,
+      "step": 6216
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.00022020918367346936,
+      "loss": 0.1727,
+      "step": 6217
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.0002201938775510204,
+      "loss": 0.2913,
+      "step": 6218
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.0002201785714285714,
+      "loss": 0.1527,
+      "step": 6219
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.00022016326530612245,
+      "loss": 0.1824,
+      "step": 6220
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.00022014795918367345,
+      "loss": 0.1961,
+      "step": 6221
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.00022013265306122445,
+      "loss": 0.16,
+      "step": 6222
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.0002201173469387755,
+      "loss": 0.0979,
+      "step": 6223
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.00022010204081632652,
+      "loss": 0.2373,
+      "step": 6224
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.00022008673469387752,
+      "loss": 0.1585,
+      "step": 6225
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.00022007142857142855,
+      "loss": 0.2056,
+      "step": 6226
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00022005612244897958,
+      "loss": 0.136,
+      "step": 6227
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.0002200408163265306,
+      "loss": 0.1151,
+      "step": 6228
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00022002551020408161,
+      "loss": 0.1696,
+      "step": 6229
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00022001020408163262,
+      "loss": 0.2581,
+      "step": 6230
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00021999489795918368,
+      "loss": 0.2352,
+      "step": 6231
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00021997959183673468,
+      "loss": 0.1909,
+      "step": 6232
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00021996428571428568,
+      "loss": 0.2943,
+      "step": 6233
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.0002199489795918367,
+      "loss": 0.1118,
+      "step": 6234
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00021993367346938774,
+      "loss": 0.1389,
+      "step": 6235
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00021991836734693877,
+      "loss": 0.1335,
+      "step": 6236
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021990306122448978,
+      "loss": 0.1767,
+      "step": 6237
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021988775510204078,
+      "loss": 0.1444,
+      "step": 6238
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021987244897959184,
+      "loss": 0.1321,
+      "step": 6239
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021985714285714284,
+      "loss": 0.124,
+      "step": 6240
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021984183673469385,
+      "loss": 0.0941,
+      "step": 6241
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021982653061224488,
+      "loss": 0.1739,
+      "step": 6242
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0002198112244897959,
+      "loss": 0.2025,
+      "step": 6243
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021979591836734694,
+      "loss": 0.1646,
+      "step": 6244
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021978061224489794,
+      "loss": 0.1085,
+      "step": 6245
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021976530612244894,
+      "loss": 0.1199,
+      "step": 6246
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.00021975,
+      "loss": 0.0771,
+      "step": 6247
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.000219734693877551,
+      "loss": 0.1478,
+      "step": 6248
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.000219719387755102,
+      "loss": 0.1315,
+      "step": 6249
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.00021970408163265304,
+      "loss": 0.1652,
+      "step": 6250
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.00021968877551020407,
+      "loss": 0.1159,
+      "step": 6251
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.00021967346938775507,
+      "loss": 0.1077,
+      "step": 6252
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.0002196581632653061,
+      "loss": 0.1489,
+      "step": 6253
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.0002196428571428571,
+      "loss": 0.241,
+      "step": 6254
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.00021962755102040816,
+      "loss": 0.1241,
+      "step": 6255
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.00021961224489795917,
+      "loss": 0.2368,
+      "step": 6256
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.00021959693877551017,
+      "loss": 0.1725,
+      "step": 6257
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.0002195816326530612,
+      "loss": 0.1481,
+      "step": 6258
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021956632653061223,
+      "loss": 0.1895,
+      "step": 6259
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021955102040816324,
+      "loss": 0.0863,
+      "step": 6260
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021953571428571427,
+      "loss": 0.1062,
+      "step": 6261
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021952040816326527,
+      "loss": 0.3566,
+      "step": 6262
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021950510204081633,
+      "loss": 0.1796,
+      "step": 6263
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021948979591836733,
+      "loss": 0.1076,
+      "step": 6264
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021947448979591833,
+      "loss": 0.0982,
+      "step": 6265
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021945918367346937,
+      "loss": 0.1363,
+      "step": 6266
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.0002194438775510204,
+      "loss": 0.1649,
+      "step": 6267
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0002194285714285714,
+      "loss": 0.2282,
+      "step": 6268
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.00021941326530612243,
+      "loss": 0.1336,
+      "step": 6269
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.00021939795918367346,
+      "loss": 0.1469,
+      "step": 6270
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0002193826530612245,
+      "loss": 0.1087,
+      "step": 6271
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0002193673469387755,
+      "loss": 0.126,
+      "step": 6272
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0002193520408163265,
+      "loss": 0.0913,
+      "step": 6273
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.00021933673469387756,
+      "loss": 0.2125,
+      "step": 6274
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.00021932142857142856,
+      "loss": 0.1558,
+      "step": 6275
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.00021930612244897956,
+      "loss": 0.1803,
+      "step": 6276
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0002192908163265306,
+      "loss": 0.1764,
+      "step": 6277
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021927551020408162,
+      "loss": 0.141,
+      "step": 6278
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021926020408163263,
+      "loss": 0.0942,
+      "step": 6279
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021924489795918366,
+      "loss": 0.1693,
+      "step": 6280
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021922959183673466,
+      "loss": 0.1583,
+      "step": 6281
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021921428571428572,
+      "loss": 0.1752,
+      "step": 6282
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021919897959183672,
+      "loss": 0.1555,
+      "step": 6283
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021918367346938773,
+      "loss": 0.1077,
+      "step": 6284
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021916836734693876,
+      "loss": 0.1049,
+      "step": 6285
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.0002191530612244898,
+      "loss": 0.1549,
+      "step": 6286
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.0002191377551020408,
+      "loss": 0.1297,
+      "step": 6287
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021912244897959182,
+      "loss": 0.1401,
+      "step": 6288
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021910714285714282,
+      "loss": 0.2293,
+      "step": 6289
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021909183673469388,
+      "loss": 0.0948,
+      "step": 6290
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021907653061224489,
+      "loss": 0.1021,
+      "step": 6291
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.0002190612244897959,
+      "loss": 0.1261,
+      "step": 6292
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021904591836734692,
+      "loss": 0.3576,
+      "step": 6293
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021903061224489795,
+      "loss": 0.1,
+      "step": 6294
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021901530612244895,
+      "loss": 0.2531,
+      "step": 6295
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021899999999999998,
+      "loss": 0.088,
+      "step": 6296
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.000218984693877551,
+      "loss": 0.0842,
+      "step": 6297
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021896938775510204,
+      "loss": 0.0999,
+      "step": 6298
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00021895408163265305,
+      "loss": 0.1556,
+      "step": 6299
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00021893877551020405,
+      "loss": 0.2357,
+      "step": 6300
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00021892346938775505,
+      "loss": 0.1374,
+      "step": 6301
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.0002189081632653061,
+      "loss": 0.121,
+      "step": 6302
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00021889285714285712,
+      "loss": 0.1047,
+      "step": 6303
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00021887755102040815,
+      "loss": 0.1963,
+      "step": 6304
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00021886224489795915,
+      "loss": 0.122,
+      "step": 6305
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.0002188469387755102,
+      "loss": 0.2544,
+      "step": 6306
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.0002188316326530612,
+      "loss": 0.1391,
+      "step": 6307
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00021881632653061221,
+      "loss": 0.1196,
+      "step": 6308
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021880102040816322,
+      "loss": 0.1513,
+      "step": 6309
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021878571428571428,
+      "loss": 0.1332,
+      "step": 6310
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021877040816326528,
+      "loss": 0.1552,
+      "step": 6311
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.0002187551020408163,
+      "loss": 0.1923,
+      "step": 6312
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021873979591836734,
+      "loss": 0.2036,
+      "step": 6313
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021872448979591834,
+      "loss": 0.1234,
+      "step": 6314
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021870918367346937,
+      "loss": 0.1608,
+      "step": 6315
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021869387755102038,
+      "loss": 0.1461,
+      "step": 6316
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021867857142857144,
+      "loss": 0.1021,
+      "step": 6317
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021866326530612244,
+      "loss": 0.1657,
+      "step": 6318
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00021864795918367344,
+      "loss": 0.1632,
+      "step": 6319
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.00021863265306122447,
+      "loss": 0.1872,
+      "step": 6320
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0002186173469387755,
+      "loss": 0.1457,
+      "step": 6321
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0002186020408163265,
+      "loss": 0.1355,
+      "step": 6322
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.00021858673469387754,
+      "loss": 0.1391,
+      "step": 6323
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.00021857142857142854,
+      "loss": 0.2182,
+      "step": 6324
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0002185561224489796,
+      "loss": 0.17,
+      "step": 6325
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0002185408163265306,
+      "loss": 0.1253,
+      "step": 6326
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0002185255102040816,
+      "loss": 0.1282,
+      "step": 6327
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.00021851020408163264,
+      "loss": 0.1252,
+      "step": 6328
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.00021849489795918367,
+      "loss": 0.1396,
+      "step": 6329
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00021847959183673467,
+      "loss": 0.2039,
+      "step": 6330
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0002184642857142857,
+      "loss": 0.1237,
+      "step": 6331
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0002184489795918367,
+      "loss": 0.1394,
+      "step": 6332
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00021843367346938776,
+      "loss": 0.1359,
+      "step": 6333
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00021841836734693876,
+      "loss": 0.0911,
+      "step": 6334
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00021840306122448977,
+      "loss": 0.1152,
+      "step": 6335
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00021838775510204077,
+      "loss": 0.1575,
+      "step": 6336
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00021837244897959183,
+      "loss": 0.2319,
+      "step": 6337
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00021835714285714283,
+      "loss": 0.3155,
+      "step": 6338
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00021834183673469386,
+      "loss": 0.1493,
+      "step": 6339
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021832653061224487,
+      "loss": 0.1332,
+      "step": 6340
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.0002183112244897959,
+      "loss": 0.1805,
+      "step": 6341
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021829591836734693,
+      "loss": 0.1646,
+      "step": 6342
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021828061224489793,
+      "loss": 0.3064,
+      "step": 6343
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021826530612244893,
+      "loss": 0.219,
+      "step": 6344
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021825,
+      "loss": 0.106,
+      "step": 6345
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.000218234693877551,
+      "loss": 0.1337,
+      "step": 6346
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021821938775510203,
+      "loss": 0.2127,
+      "step": 6347
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021820408163265303,
+      "loss": 0.1368,
+      "step": 6348
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021818877551020406,
+      "loss": 0.1866,
+      "step": 6349
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0002181734693877551,
+      "loss": 0.2369,
+      "step": 6350
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0002181581632653061,
+      "loss": 0.1728,
+      "step": 6351
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0002181428571428571,
+      "loss": 0.1802,
+      "step": 6352
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00021812755102040816,
+      "loss": 0.156,
+      "step": 6353
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00021811224489795916,
+      "loss": 0.0913,
+      "step": 6354
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0002180969387755102,
+      "loss": 0.2111,
+      "step": 6355
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0002180816326530612,
+      "loss": 0.2672,
+      "step": 6356
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00021806632653061222,
+      "loss": 0.1142,
+      "step": 6357
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00021805102040816325,
+      "loss": 0.1118,
+      "step": 6358
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00021803571428571426,
+      "loss": 0.1096,
+      "step": 6359
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00021802040816326532,
+      "loss": 0.121,
+      "step": 6360
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021800510204081632,
+      "loss": 0.1896,
+      "step": 6361
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021798979591836732,
+      "loss": 0.1217,
+      "step": 6362
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021797448979591835,
+      "loss": 0.0988,
+      "step": 6363
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021795918367346938,
+      "loss": 0.1513,
+      "step": 6364
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.0002179438775510204,
+      "loss": 0.1805,
+      "step": 6365
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021792857142857142,
+      "loss": 0.1734,
+      "step": 6366
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021791326530612242,
+      "loss": 0.194,
+      "step": 6367
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021789795918367348,
+      "loss": 0.2191,
+      "step": 6368
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021788265306122448,
+      "loss": 0.189,
+      "step": 6369
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00021786734693877549,
+      "loss": 0.1406,
+      "step": 6370
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.0002178520408163265,
+      "loss": 0.1312,
+      "step": 6371
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021783673469387755,
+      "loss": 0.1287,
+      "step": 6372
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021782142857142855,
+      "loss": 0.1505,
+      "step": 6373
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021780612244897958,
+      "loss": 0.2307,
+      "step": 6374
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021779081632653058,
+      "loss": 0.2972,
+      "step": 6375
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021777551020408161,
+      "loss": 0.1524,
+      "step": 6376
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021776020408163264,
+      "loss": 0.1489,
+      "step": 6377
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021774489795918365,
+      "loss": 0.0822,
+      "step": 6378
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021772959183673465,
+      "loss": 0.1384,
+      "step": 6379
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.0002177142857142857,
+      "loss": 0.295,
+      "step": 6380
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.0002176989795918367,
+      "loss": 0.2841,
+      "step": 6381
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00021768367346938774,
+      "loss": 0.1982,
+      "step": 6382
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00021766836734693875,
+      "loss": 0.1551,
+      "step": 6383
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00021765306122448978,
+      "loss": 0.102,
+      "step": 6384
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.0002176377551020408,
+      "loss": 0.0973,
+      "step": 6385
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.0002176224489795918,
+      "loss": 0.0913,
+      "step": 6386
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00021760714285714281,
+      "loss": 0.1313,
+      "step": 6387
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00021759183673469387,
+      "loss": 0.1634,
+      "step": 6388
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00021757653061224488,
+      "loss": 0.1616,
+      "step": 6389
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.0002175612244897959,
+      "loss": 0.1621,
+      "step": 6390
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.0002175459183673469,
+      "loss": 0.103,
+      "step": 6391
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021753061224489794,
+      "loss": 0.0935,
+      "step": 6392
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021751530612244897,
+      "loss": 0.1922,
+      "step": 6393
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021749999999999997,
+      "loss": 0.1248,
+      "step": 6394
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021748469387755098,
+      "loss": 0.14,
+      "step": 6395
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021746938775510204,
+      "loss": 0.1166,
+      "step": 6396
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021745408163265304,
+      "loss": 0.1111,
+      "step": 6397
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021743877551020404,
+      "loss": 0.1153,
+      "step": 6398
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021742346938775507,
+      "loss": 0.2111,
+      "step": 6399
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0002174081632653061,
+      "loss": 0.222,
+      "step": 6400
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021739285714285713,
+      "loss": 0.1276,
+      "step": 6401
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00021737755102040814,
+      "loss": 0.1915,
+      "step": 6402
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00021736224489795914,
+      "loss": 0.1391,
+      "step": 6403
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.0002173469387755102,
+      "loss": 0.0654,
+      "step": 6404
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.0002173316326530612,
+      "loss": 0.1563,
+      "step": 6405
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.0002173163265306122,
+      "loss": 0.1874,
+      "step": 6406
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00021730102040816326,
+      "loss": 0.1453,
+      "step": 6407
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00021728571428571427,
+      "loss": 0.1314,
+      "step": 6408
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.0002172704081632653,
+      "loss": 0.1551,
+      "step": 6409
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.0002172551020408163,
+      "loss": 0.1523,
+      "step": 6410
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00021723979591836733,
+      "loss": 0.15,
+      "step": 6411
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021722448979591836,
+      "loss": 0.2152,
+      "step": 6412
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021720918367346936,
+      "loss": 0.2004,
+      "step": 6413
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021719387755102037,
+      "loss": 0.2006,
+      "step": 6414
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021717857142857143,
+      "loss": 0.1752,
+      "step": 6415
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021716326530612243,
+      "loss": 0.0873,
+      "step": 6416
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021714795918367346,
+      "loss": 0.1341,
+      "step": 6417
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021713265306122446,
+      "loss": 0.3317,
+      "step": 6418
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.0002171173469387755,
+      "loss": 0.1346,
+      "step": 6419
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021710204081632652,
+      "loss": 0.114,
+      "step": 6420
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021708673469387753,
+      "loss": 0.1884,
+      "step": 6421
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00021707142857142853,
+      "loss": 0.1209,
+      "step": 6422
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0002170561224489796,
+      "loss": 0.0835,
+      "step": 6423
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0002170408163265306,
+      "loss": 0.2258,
+      "step": 6424
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.00021702551020408162,
+      "loss": 0.1266,
+      "step": 6425
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.00021701020408163263,
+      "loss": 0.1596,
+      "step": 6426
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.00021699489795918366,
+      "loss": 0.1756,
+      "step": 6427
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0002169795918367347,
+      "loss": 0.2396,
+      "step": 6428
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0002169642857142857,
+      "loss": 0.1016,
+      "step": 6429
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0002169489795918367,
+      "loss": 0.1485,
+      "step": 6430
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.00021693367346938775,
+      "loss": 0.1858,
+      "step": 6431
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.00021691836734693876,
+      "loss": 0.2576,
+      "step": 6432
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021690306122448976,
+      "loss": 0.1276,
+      "step": 6433
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.0002168877551020408,
+      "loss": 0.147,
+      "step": 6434
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021687244897959182,
+      "loss": 0.1756,
+      "step": 6435
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021685714285714285,
+      "loss": 0.1108,
+      "step": 6436
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021684183673469385,
+      "loss": 0.187,
+      "step": 6437
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021682653061224486,
+      "loss": 0.2211,
+      "step": 6438
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021681122448979592,
+      "loss": 0.2109,
+      "step": 6439
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021679591836734692,
+      "loss": 0.1354,
+      "step": 6440
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021678061224489792,
+      "loss": 0.1105,
+      "step": 6441
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021676530612244895,
+      "loss": 0.1113,
+      "step": 6442
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021674999999999998,
+      "loss": 0.2719,
+      "step": 6443
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021673469387755101,
+      "loss": 0.1257,
+      "step": 6444
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021671938775510202,
+      "loss": 0.1775,
+      "step": 6445
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021670408163265302,
+      "loss": 0.1511,
+      "step": 6446
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021668877551020408,
+      "loss": 0.1009,
+      "step": 6447
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021667346938775508,
+      "loss": 0.1118,
+      "step": 6448
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021665816326530609,
+      "loss": 0.3218,
+      "step": 6449
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021664285714285714,
+      "loss": 0.2116,
+      "step": 6450
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021662755102040815,
+      "loss": 0.2495,
+      "step": 6451
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021661224489795918,
+      "loss": 0.2876,
+      "step": 6452
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00021659693877551018,
+      "loss": 0.1239,
+      "step": 6453
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.0002165816326530612,
+      "loss": 0.2241,
+      "step": 6454
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00021656632653061224,
+      "loss": 0.1709,
+      "step": 6455
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00021655102040816324,
+      "loss": 0.1381,
+      "step": 6456
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00021653571428571425,
+      "loss": 0.1552,
+      "step": 6457
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.0002165204081632653,
+      "loss": 0.1707,
+      "step": 6458
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.0002165051020408163,
+      "loss": 0.1065,
+      "step": 6459
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.0002164897959183673,
+      "loss": 0.2324,
+      "step": 6460
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00021647448979591834,
+      "loss": 0.2021,
+      "step": 6461
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00021645918367346937,
+      "loss": 0.1837,
+      "step": 6462
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.0002164438775510204,
+      "loss": 0.209,
+      "step": 6463
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0002164285714285714,
+      "loss": 0.1595,
+      "step": 6464
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0002164132653061224,
+      "loss": 0.1538,
+      "step": 6465
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00021639795918367347,
+      "loss": 0.1981,
+      "step": 6466
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00021638265306122447,
+      "loss": 0.0891,
+      "step": 6467
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00021636734693877548,
+      "loss": 0.1462,
+      "step": 6468
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0002163520408163265,
+      "loss": 0.1475,
+      "step": 6469
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00021633673469387754,
+      "loss": 0.1489,
+      "step": 6470
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00021632142857142857,
+      "loss": 0.1582,
+      "step": 6471
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00021630612244897957,
+      "loss": 0.1282,
+      "step": 6472
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00021629081632653057,
+      "loss": 0.1163,
+      "step": 6473
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00021627551020408163,
+      "loss": 0.2508,
+      "step": 6474
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00021626020408163264,
+      "loss": 0.1664,
+      "step": 6475
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00021624489795918364,
+      "loss": 0.1388,
+      "step": 6476
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00021622959183673467,
+      "loss": 0.1582,
+      "step": 6477
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.0002162142857142857,
+      "loss": 0.1002,
+      "step": 6478
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00021619897959183673,
+      "loss": 0.0807,
+      "step": 6479
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00021618367346938773,
+      "loss": 0.1263,
+      "step": 6480
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00021616836734693874,
+      "loss": 0.229,
+      "step": 6481
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.0002161530612244898,
+      "loss": 0.2009,
+      "step": 6482
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.0002161377551020408,
+      "loss": 0.1763,
+      "step": 6483
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.0002161224489795918,
+      "loss": 0.1444,
+      "step": 6484
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00021610714285714283,
+      "loss": 0.1152,
+      "step": 6485
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00021609183673469386,
+      "loss": 0.0867,
+      "step": 6486
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.0002160765306122449,
+      "loss": 0.184,
+      "step": 6487
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.0002160612244897959,
+      "loss": 0.1174,
+      "step": 6488
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.0002160459183673469,
+      "loss": 0.1721,
+      "step": 6489
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00021603061224489796,
+      "loss": 0.1558,
+      "step": 6490
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00021601530612244896,
+      "loss": 0.0938,
+      "step": 6491
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00021599999999999996,
+      "loss": 0.1012,
+      "step": 6492
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.000215984693877551,
+      "loss": 0.254,
+      "step": 6493
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00021596938775510203,
+      "loss": 0.231,
+      "step": 6494
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00021595408163265303,
+      "loss": 0.094,
+      "step": 6495
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00021593877551020406,
+      "loss": 0.2155,
+      "step": 6496
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.0002159234693877551,
+      "loss": 0.131,
+      "step": 6497
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00021590816326530612,
+      "loss": 0.1582,
+      "step": 6498
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00021589285714285712,
+      "loss": 0.2713,
+      "step": 6499
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00021587755102040813,
+      "loss": 0.1325,
+      "step": 6500
+    },
+    {
+      "epoch": 6.31,
+      "eval_cer": 0.2654695500281553,
+      "eval_loss": 0.9578253030776978,
+      "eval_runtime": 229.2456,
+      "eval_samples_per_second": 38.448,
+      "eval_steps_per_second": 4.807,
+      "eval_wer": 0.6691438991969592,
+      "step": 6500
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00021586224489795919,
+      "loss": 0.2077,
+      "step": 6501
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.0002158469387755102,
+      "loss": 0.1443,
+      "step": 6502
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.0002158316326530612,
+      "loss": 0.1847,
+      "step": 6503
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00021581632653061222,
+      "loss": 0.1519,
+      "step": 6504
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00021580102040816325,
+      "loss": 0.1879,
+      "step": 6505
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00021578571428571428,
+      "loss": 0.1398,
+      "step": 6506
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0002157704081632653,
+      "loss": 0.1292,
+      "step": 6507
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0002157551020408163,
+      "loss": 0.2009,
+      "step": 6508
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00021573979591836735,
+      "loss": 0.1782,
+      "step": 6509
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00021572448979591835,
+      "loss": 0.1851,
+      "step": 6510
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00021570918367346936,
+      "loss": 0.1735,
+      "step": 6511
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00021569387755102039,
+      "loss": 0.1534,
+      "step": 6512
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00021567857142857142,
+      "loss": 0.1334,
+      "step": 6513
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00021566326530612245,
+      "loss": 0.1489,
+      "step": 6514
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021564795918367345,
+      "loss": 0.1704,
+      "step": 6515
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021563265306122445,
+      "loss": 0.1381,
+      "step": 6516
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.0002156173469387755,
+      "loss": 0.1542,
+      "step": 6517
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021560204081632652,
+      "loss": 0.1697,
+      "step": 6518
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021558673469387752,
+      "loss": 0.2146,
+      "step": 6519
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021557142857142855,
+      "loss": 0.148,
+      "step": 6520
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021555612244897958,
+      "loss": 0.1379,
+      "step": 6521
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021554081632653058,
+      "loss": 0.1459,
+      "step": 6522
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021552551020408161,
+      "loss": 0.1359,
+      "step": 6523
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021551020408163262,
+      "loss": 0.2608,
+      "step": 6524
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021549489795918367,
+      "loss": 0.1817,
+      "step": 6525
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021547959183673468,
+      "loss": 0.1297,
+      "step": 6526
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021546428571428568,
+      "loss": 0.1952,
+      "step": 6527
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.0002154489795918367,
+      "loss": 0.1864,
+      "step": 6528
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021543367346938774,
+      "loss": 0.2058,
+      "step": 6529
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021541836734693875,
+      "loss": 0.2136,
+      "step": 6530
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021540306122448978,
+      "loss": 0.194,
+      "step": 6531
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021538775510204078,
+      "loss": 0.1104,
+      "step": 6532
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021537244897959184,
+      "loss": 0.0925,
+      "step": 6533
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021535714285714284,
+      "loss": 0.199,
+      "step": 6534
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021534183673469384,
+      "loss": 0.125,
+      "step": 6535
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00021532653061224488,
+      "loss": 0.207,
+      "step": 6536
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.0002153112244897959,
+      "loss": 0.2662,
+      "step": 6537
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.0002152959183673469,
+      "loss": 0.1402,
+      "step": 6538
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00021528061224489794,
+      "loss": 0.0774,
+      "step": 6539
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00021526530612244897,
+      "loss": 0.1343,
+      "step": 6540
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00021525,
+      "loss": 0.1703,
+      "step": 6541
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.000215234693877551,
+      "loss": 0.1211,
+      "step": 6542
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.000215219387755102,
+      "loss": 0.2893,
+      "step": 6543
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00021520408163265307,
+      "loss": 0.2459,
+      "step": 6544
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00021518877551020407,
+      "loss": 0.0956,
+      "step": 6545
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021517346938775507,
+      "loss": 0.0985,
+      "step": 6546
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.0002151581632653061,
+      "loss": 0.1851,
+      "step": 6547
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021514285714285713,
+      "loss": 0.1043,
+      "step": 6548
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021512755102040816,
+      "loss": 0.3,
+      "step": 6549
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021511224489795917,
+      "loss": 0.1569,
+      "step": 6550
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021509693877551017,
+      "loss": 0.1514,
+      "step": 6551
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021508163265306123,
+      "loss": 0.2275,
+      "step": 6552
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021506632653061223,
+      "loss": 0.1171,
+      "step": 6553
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021505102040816324,
+      "loss": 0.1455,
+      "step": 6554
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00021503571428571427,
+      "loss": 0.2744,
+      "step": 6555
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0002150204081632653,
+      "loss": 0.2196,
+      "step": 6556
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0002150051020408163,
+      "loss": 0.1184,
+      "step": 6557
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.00021498979591836733,
+      "loss": 0.1557,
+      "step": 6558
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.00021497448979591833,
+      "loss": 0.1348,
+      "step": 6559
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0002149591836734694,
+      "loss": 0.1157,
+      "step": 6560
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0002149438775510204,
+      "loss": 0.1138,
+      "step": 6561
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0002149285714285714,
+      "loss": 0.1882,
+      "step": 6562
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.00021491326530612243,
+      "loss": 0.1968,
+      "step": 6563
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.00021489795918367346,
+      "loss": 0.148,
+      "step": 6564
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.00021488265306122446,
+      "loss": 0.1181,
+      "step": 6565
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0002148673469387755,
+      "loss": 0.1767,
+      "step": 6566
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.0002148520408163265,
+      "loss": 0.257,
+      "step": 6567
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00021483673469387755,
+      "loss": 0.3738,
+      "step": 6568
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00021482142857142856,
+      "loss": 0.2054,
+      "step": 6569
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00021480612244897956,
+      "loss": 0.1909,
+      "step": 6570
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.0002147908163265306,
+      "loss": 0.2006,
+      "step": 6571
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00021477551020408162,
+      "loss": 0.2032,
+      "step": 6572
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00021476020408163263,
+      "loss": 0.0834,
+      "step": 6573
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00021474489795918366,
+      "loss": 0.1975,
+      "step": 6574
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00021472959183673466,
+      "loss": 0.2899,
+      "step": 6575
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00021471428571428572,
+      "loss": 0.1734,
+      "step": 6576
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00021469897959183672,
+      "loss": 0.1357,
+      "step": 6577
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00021468367346938772,
+      "loss": 0.11,
+      "step": 6578
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00021466836734693873,
+      "loss": 0.0947,
+      "step": 6579
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00021465306122448979,
+      "loss": 0.1783,
+      "step": 6580
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.0002146377551020408,
+      "loss": 0.1499,
+      "step": 6581
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00021462244897959182,
+      "loss": 0.1991,
+      "step": 6582
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00021460714285714282,
+      "loss": 0.1554,
+      "step": 6583
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00021459183673469388,
+      "loss": 0.1405,
+      "step": 6584
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00021457653061224488,
+      "loss": 0.1079,
+      "step": 6585
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.0002145612244897959,
+      "loss": 0.1452,
+      "step": 6586
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00021454591836734695,
+      "loss": 0.3181,
+      "step": 6587
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00021453061224489795,
+      "loss": 0.1307,
+      "step": 6588
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00021451530612244895,
+      "loss": 0.3581,
+      "step": 6589
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00021449999999999998,
+      "loss": 0.1472,
+      "step": 6590
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.000214484693877551,
+      "loss": 0.0997,
+      "step": 6591
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00021446938775510202,
+      "loss": 0.1179,
+      "step": 6592
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00021445408163265305,
+      "loss": 0.1781,
+      "step": 6593
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00021443877551020405,
+      "loss": 0.0994,
+      "step": 6594
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.0002144234693877551,
+      "loss": 0.1793,
+      "step": 6595
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.0002144081632653061,
+      "loss": 0.1618,
+      "step": 6596
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00021439285714285712,
+      "loss": 0.1335,
+      "step": 6597
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00021437755102040815,
+      "loss": 0.1205,
+      "step": 6598
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00021436224489795918,
+      "loss": 0.3165,
+      "step": 6599
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00021434693877551018,
+      "loss": 0.1437,
+      "step": 6600
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.0002143316326530612,
+      "loss": 0.1142,
+      "step": 6601
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00021431632653061221,
+      "loss": 0.2056,
+      "step": 6602
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00021430102040816327,
+      "loss": 0.1988,
+      "step": 6603
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00021428571428571427,
+      "loss": 0.1283,
+      "step": 6604
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00021427040816326528,
+      "loss": 0.1153,
+      "step": 6605
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.0002142551020408163,
+      "loss": 0.1663,
+      "step": 6606
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00021423979591836734,
+      "loss": 0.1033,
+      "step": 6607
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00021422448979591834,
+      "loss": 0.0989,
+      "step": 6608
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00021420918367346937,
+      "loss": 0.1537,
+      "step": 6609
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00021419387755102038,
+      "loss": 0.1008,
+      "step": 6610
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00021417857142857143,
+      "loss": 0.2016,
+      "step": 6611
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00021416326530612244,
+      "loss": 0.2678,
+      "step": 6612
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00021414795918367344,
+      "loss": 0.2725,
+      "step": 6613
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00021413265306122444,
+      "loss": 0.1501,
+      "step": 6614
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.0002141173469387755,
+      "loss": 0.1325,
+      "step": 6615
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.0002141020408163265,
+      "loss": 0.1905,
+      "step": 6616
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00021408673469387754,
+      "loss": 0.1759,
+      "step": 6617
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.00021407142857142854,
+      "loss": 0.1468,
+      "step": 6618
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.00021405612244897957,
+      "loss": 0.1253,
+      "step": 6619
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0002140408163265306,
+      "loss": 0.2433,
+      "step": 6620
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0002140255102040816,
+      "loss": 0.1777,
+      "step": 6621
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0002140102040816326,
+      "loss": 0.241,
+      "step": 6622
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.00021399489795918367,
+      "loss": 0.1885,
+      "step": 6623
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.00021397959183673467,
+      "loss": 0.1514,
+      "step": 6624
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0002139642857142857,
+      "loss": 0.2185,
+      "step": 6625
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0002139489795918367,
+      "loss": 0.2224,
+      "step": 6626
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.00021393367346938773,
+      "loss": 0.1337,
+      "step": 6627
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.00021391836734693876,
+      "loss": 0.0846,
+      "step": 6628
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00021390306122448977,
+      "loss": 0.0923,
+      "step": 6629
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00021388775510204082,
+      "loss": 0.1464,
+      "step": 6630
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00021387244897959183,
+      "loss": 0.173,
+      "step": 6631
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00021385714285714283,
+      "loss": 0.14,
+      "step": 6632
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00021384183673469386,
+      "loss": 0.169,
+      "step": 6633
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.0002138265306122449,
+      "loss": 0.3105,
+      "step": 6634
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.0002138112244897959,
+      "loss": 0.1417,
+      "step": 6635
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00021379591836734693,
+      "loss": 0.2315,
+      "step": 6636
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00021378061224489793,
+      "loss": 0.1905,
+      "step": 6637
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.000213765306122449,
+      "loss": 0.1544,
+      "step": 6638
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00021375,
+      "loss": 0.1939,
+      "step": 6639
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.000213734693877551,
+      "loss": 0.1175,
+      "step": 6640
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.000213719387755102,
+      "loss": 0.1507,
+      "step": 6641
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00021370408163265306,
+      "loss": 0.1034,
+      "step": 6642
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00021368877551020406,
+      "loss": 0.1271,
+      "step": 6643
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.0002136734693877551,
+      "loss": 0.1261,
+      "step": 6644
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.0002136581632653061,
+      "loss": 0.1892,
+      "step": 6645
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00021364285714285715,
+      "loss": 0.1431,
+      "step": 6646
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00021362755102040815,
+      "loss": 0.2011,
+      "step": 6647
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00021361224489795916,
+      "loss": 0.1304,
+      "step": 6648
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021359693877551016,
+      "loss": 0.2775,
+      "step": 6649
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021358163265306122,
+      "loss": 0.2064,
+      "step": 6650
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021356632653061222,
+      "loss": 0.1471,
+      "step": 6651
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021355102040816325,
+      "loss": 0.1421,
+      "step": 6652
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021353571428571426,
+      "loss": 0.1437,
+      "step": 6653
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.0002135204081632653,
+      "loss": 0.0716,
+      "step": 6654
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021350510204081632,
+      "loss": 0.1473,
+      "step": 6655
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021348979591836732,
+      "loss": 0.1383,
+      "step": 6656
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021347448979591832,
+      "loss": 0.2147,
+      "step": 6657
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00021345918367346938,
+      "loss": 0.1113,
+      "step": 6658
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021344387755102039,
+      "loss": 0.1523,
+      "step": 6659
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021342857142857142,
+      "loss": 0.1058,
+      "step": 6660
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021341326530612242,
+      "loss": 0.1257,
+      "step": 6661
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021339795918367345,
+      "loss": 0.1603,
+      "step": 6662
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021338265306122448,
+      "loss": 0.1711,
+      "step": 6663
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021336734693877548,
+      "loss": 0.1789,
+      "step": 6664
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.0002133520408163265,
+      "loss": 0.1503,
+      "step": 6665
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021333673469387755,
+      "loss": 0.1725,
+      "step": 6666
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021332142857142855,
+      "loss": 0.1648,
+      "step": 6667
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021330612244897958,
+      "loss": 0.1749,
+      "step": 6668
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00021329081632653058,
+      "loss": 0.1627,
+      "step": 6669
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0002132755102040816,
+      "loss": 0.0996,
+      "step": 6670
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00021326020408163264,
+      "loss": 0.2344,
+      "step": 6671
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00021324489795918365,
+      "loss": 0.0808,
+      "step": 6672
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00021322959183673465,
+      "loss": 0.1344,
+      "step": 6673
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0002132142857142857,
+      "loss": 0.2055,
+      "step": 6674
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0002131989795918367,
+      "loss": 0.2478,
+      "step": 6675
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00021318367346938772,
+      "loss": 0.1118,
+      "step": 6676
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00021316836734693877,
+      "loss": 0.1697,
+      "step": 6677
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00021315306122448978,
+      "loss": 0.2153,
+      "step": 6678
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0002131377551020408,
+      "loss": 0.1116,
+      "step": 6679
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.0002131224489795918,
+      "loss": 0.148,
+      "step": 6680
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00021310714285714284,
+      "loss": 0.2156,
+      "step": 6681
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00021309183673469387,
+      "loss": 0.1511,
+      "step": 6682
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00021307653061224487,
+      "loss": 0.1013,
+      "step": 6683
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00021306122448979588,
+      "loss": 0.2153,
+      "step": 6684
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00021304591836734694,
+      "loss": 0.1551,
+      "step": 6685
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00021303061224489794,
+      "loss": 0.1572,
+      "step": 6686
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00021301530612244897,
+      "loss": 0.2334,
+      "step": 6687
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00021299999999999997,
+      "loss": 0.2328,
+      "step": 6688
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.000212984693877551,
+      "loss": 0.1666,
+      "step": 6689
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00021296938775510203,
+      "loss": 0.2124,
+      "step": 6690
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00021295408163265304,
+      "loss": 0.1878,
+      "step": 6691
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00021293877551020404,
+      "loss": 0.1178,
+      "step": 6692
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0002129234693877551,
+      "loss": 0.1424,
+      "step": 6693
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0002129081632653061,
+      "loss": 0.1102,
+      "step": 6694
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00021289285714285713,
+      "loss": 0.1203,
+      "step": 6695
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00021287755102040814,
+      "loss": 0.1137,
+      "step": 6696
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00021286224489795917,
+      "loss": 0.1675,
+      "step": 6697
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0002128469387755102,
+      "loss": 0.1281,
+      "step": 6698
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0002128316326530612,
+      "loss": 0.23,
+      "step": 6699
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0002128163265306122,
+      "loss": 0.2263,
+      "step": 6700
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00021280102040816326,
+      "loss": 0.1127,
+      "step": 6701
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00021278571428571427,
+      "loss": 0.1774,
+      "step": 6702
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.0002127704081632653,
+      "loss": 0.1288,
+      "step": 6703
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.0002127551020408163,
+      "loss": 0.1845,
+      "step": 6704
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00021273979591836733,
+      "loss": 0.2315,
+      "step": 6705
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00021272448979591836,
+      "loss": 0.1625,
+      "step": 6706
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00021270918367346936,
+      "loss": 0.1583,
+      "step": 6707
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00021269387755102037,
+      "loss": 0.1308,
+      "step": 6708
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00021267857142857142,
+      "loss": 0.0999,
+      "step": 6709
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00021266326530612243,
+      "loss": 0.2079,
+      "step": 6710
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00021264795918367343,
+      "loss": 0.1369,
+      "step": 6711
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00021263265306122446,
+      "loss": 0.2199,
+      "step": 6712
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.0002126173469387755,
+      "loss": 0.2149,
+      "step": 6713
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00021260204081632652,
+      "loss": 0.0924,
+      "step": 6714
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00021258673469387753,
+      "loss": 0.1297,
+      "step": 6715
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00021257142857142853,
+      "loss": 0.1145,
+      "step": 6716
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.0002125561224489796,
+      "loss": 0.1173,
+      "step": 6717
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.0002125408163265306,
+      "loss": 0.21,
+      "step": 6718
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.0002125255102040816,
+      "loss": 0.1907,
+      "step": 6719
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00021251020408163263,
+      "loss": 0.1344,
+      "step": 6720
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021249489795918366,
+      "loss": 0.2305,
+      "step": 6721
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.0002124795918367347,
+      "loss": 0.1753,
+      "step": 6722
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.0002124642857142857,
+      "loss": 0.1117,
+      "step": 6723
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021244897959183672,
+      "loss": 0.1575,
+      "step": 6724
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021243367346938775,
+      "loss": 0.193,
+      "step": 6725
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021241836734693875,
+      "loss": 0.2006,
+      "step": 6726
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021240306122448976,
+      "loss": 0.0884,
+      "step": 6727
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021238775510204082,
+      "loss": 0.129,
+      "step": 6728
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021237244897959182,
+      "loss": 0.0925,
+      "step": 6729
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021235714285714285,
+      "loss": 0.166,
+      "step": 6730
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00021234183673469385,
+      "loss": 0.1647,
+      "step": 6731
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021232653061224488,
+      "loss": 0.1114,
+      "step": 6732
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021231122448979591,
+      "loss": 0.0943,
+      "step": 6733
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021229591836734692,
+      "loss": 0.1368,
+      "step": 6734
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021228061224489792,
+      "loss": 0.0737,
+      "step": 6735
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021226530612244898,
+      "loss": 0.1522,
+      "step": 6736
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021224999999999998,
+      "loss": 0.2486,
+      "step": 6737
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021223469387755099,
+      "loss": 0.1264,
+      "step": 6738
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021221938775510202,
+      "loss": 0.1195,
+      "step": 6739
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021220408163265305,
+      "loss": 0.0877,
+      "step": 6740
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00021218877551020408,
+      "loss": 0.165,
+      "step": 6741
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021217346938775508,
+      "loss": 0.1966,
+      "step": 6742
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021215816326530608,
+      "loss": 0.3681,
+      "step": 6743
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021214285714285714,
+      "loss": 0.1659,
+      "step": 6744
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021212755102040815,
+      "loss": 0.1656,
+      "step": 6745
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021211224489795915,
+      "loss": 0.1213,
+      "step": 6746
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021209693877551018,
+      "loss": 0.1445,
+      "step": 6747
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.0002120816326530612,
+      "loss": 0.0938,
+      "step": 6748
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021206632653061224,
+      "loss": 0.4941,
+      "step": 6749
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021205102040816324,
+      "loss": 0.1504,
+      "step": 6750
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00021203571428571425,
+      "loss": 0.2067,
+      "step": 6751
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0002120204081632653,
+      "loss": 0.2139,
+      "step": 6752
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0002120051020408163,
+      "loss": 0.0984,
+      "step": 6753
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0002119897959183673,
+      "loss": 0.1575,
+      "step": 6754
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.00021197448979591834,
+      "loss": 0.1341,
+      "step": 6755
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.00021195918367346937,
+      "loss": 0.1565,
+      "step": 6756
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0002119438775510204,
+      "loss": 0.146,
+      "step": 6757
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0002119285714285714,
+      "loss": 0.1879,
+      "step": 6758
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0002119132653061224,
+      "loss": 0.1426,
+      "step": 6759
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.00021189795918367347,
+      "loss": 0.1259,
+      "step": 6760
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.00021188265306122447,
+      "loss": 0.1108,
+      "step": 6761
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00021186734693877547,
+      "loss": 0.4364,
+      "step": 6762
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.0002118520408163265,
+      "loss": 0.1878,
+      "step": 6763
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00021183673469387754,
+      "loss": 0.2267,
+      "step": 6764
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00021182142857142857,
+      "loss": 0.15,
+      "step": 6765
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00021180612244897957,
+      "loss": 0.1525,
+      "step": 6766
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.0002117908163265306,
+      "loss": 0.1632,
+      "step": 6767
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00021177551020408163,
+      "loss": 0.2253,
+      "step": 6768
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00021176020408163263,
+      "loss": 0.1956,
+      "step": 6769
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00021174489795918364,
+      "loss": 0.2017,
+      "step": 6770
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.0002117295918367347,
+      "loss": 0.2062,
+      "step": 6771
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.0002117142857142857,
+      "loss": 0.1025,
+      "step": 6772
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0002116989795918367,
+      "loss": 0.1421,
+      "step": 6773
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00021168367346938773,
+      "loss": 0.1423,
+      "step": 6774
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00021166836734693876,
+      "loss": 0.2186,
+      "step": 6775
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0002116530612244898,
+      "loss": 0.245,
+      "step": 6776
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0002116377551020408,
+      "loss": 0.2151,
+      "step": 6777
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0002116224489795918,
+      "loss": 0.1224,
+      "step": 6778
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00021160714285714286,
+      "loss": 0.1788,
+      "step": 6779
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00021159183673469386,
+      "loss": 0.2047,
+      "step": 6780
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00021157653061224487,
+      "loss": 0.1394,
+      "step": 6781
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0002115612244897959,
+      "loss": 0.1597,
+      "step": 6782
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021154591836734693,
+      "loss": 0.1648,
+      "step": 6783
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021153061224489796,
+      "loss": 0.1,
+      "step": 6784
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021151530612244896,
+      "loss": 0.171,
+      "step": 6785
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021149999999999996,
+      "loss": 0.1226,
+      "step": 6786
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021148469387755102,
+      "loss": 0.1772,
+      "step": 6787
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021146938775510202,
+      "loss": 0.1625,
+      "step": 6788
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021145408163265303,
+      "loss": 0.1977,
+      "step": 6789
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021143877551020406,
+      "loss": 0.0919,
+      "step": 6790
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.0002114234693877551,
+      "loss": 0.1295,
+      "step": 6791
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00021140816326530612,
+      "loss": 0.1121,
+      "step": 6792
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00021139285714285712,
+      "loss": 0.1591,
+      "step": 6793
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00021137755102040813,
+      "loss": 0.1381,
+      "step": 6794
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00021136224489795918,
+      "loss": 0.1915,
+      "step": 6795
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.0002113469387755102,
+      "loss": 0.1345,
+      "step": 6796
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.0002113316326530612,
+      "loss": 0.1422,
+      "step": 6797
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00021131632653061222,
+      "loss": 0.0847,
+      "step": 6798
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00021130102040816325,
+      "loss": 0.3378,
+      "step": 6799
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00021128571428571426,
+      "loss": 0.1714,
+      "step": 6800
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.0002112704081632653,
+      "loss": 0.1265,
+      "step": 6801
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.0002112551020408163,
+      "loss": 0.0943,
+      "step": 6802
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00021123979591836735,
+      "loss": 0.1649,
+      "step": 6803
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021122448979591835,
+      "loss": 0.1039,
+      "step": 6804
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021120918367346935,
+      "loss": 0.2368,
+      "step": 6805
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021119387755102039,
+      "loss": 0.1699,
+      "step": 6806
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021117857142857142,
+      "loss": 0.1325,
+      "step": 6807
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021116326530612242,
+      "loss": 0.1127,
+      "step": 6808
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021114795918367345,
+      "loss": 0.1397,
+      "step": 6809
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021113265306122445,
+      "loss": 0.1098,
+      "step": 6810
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0002111173469387755,
+      "loss": 0.2397,
+      "step": 6811
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021110204081632651,
+      "loss": 0.197,
+      "step": 6812
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00021108673469387752,
+      "loss": 0.1851,
+      "step": 6813
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021107142857142858,
+      "loss": 0.1491,
+      "step": 6814
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021105612244897958,
+      "loss": 0.1406,
+      "step": 6815
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021104081632653058,
+      "loss": 0.1438,
+      "step": 6816
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.0002110255102040816,
+      "loss": 0.1691,
+      "step": 6817
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021101020408163264,
+      "loss": 0.2632,
+      "step": 6818
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021099489795918367,
+      "loss": 0.1846,
+      "step": 6819
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021097959183673468,
+      "loss": 0.1192,
+      "step": 6820
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021096428571428568,
+      "loss": 0.129,
+      "step": 6821
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021094897959183674,
+      "loss": 0.0904,
+      "step": 6822
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00021093367346938774,
+      "loss": 0.1436,
+      "step": 6823
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00021091836734693875,
+      "loss": 0.1807,
+      "step": 6824
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00021090306122448978,
+      "loss": 0.2207,
+      "step": 6825
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.0002108877551020408,
+      "loss": 0.1127,
+      "step": 6826
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00021087244897959184,
+      "loss": 0.0983,
+      "step": 6827
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00021085714285714284,
+      "loss": 0.1736,
+      "step": 6828
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00021084183673469384,
+      "loss": 0.0859,
+      "step": 6829
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.0002108265306122449,
+      "loss": 0.1623,
+      "step": 6830
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.0002108112244897959,
+      "loss": 0.3549,
+      "step": 6831
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.0002107959183673469,
+      "loss": 0.1528,
+      "step": 6832
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00021078061224489794,
+      "loss": 0.1052,
+      "step": 6833
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00021076530612244897,
+      "loss": 0.1598,
+      "step": 6834
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00021074999999999997,
+      "loss": 0.0913,
+      "step": 6835
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.000210734693877551,
+      "loss": 0.1804,
+      "step": 6836
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.000210719387755102,
+      "loss": 0.2478,
+      "step": 6837
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00021070408163265306,
+      "loss": 0.215,
+      "step": 6838
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00021068877551020407,
+      "loss": 0.1096,
+      "step": 6839
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00021067346938775507,
+      "loss": 0.1859,
+      "step": 6840
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.0002106581632653061,
+      "loss": 0.0816,
+      "step": 6841
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00021064285714285713,
+      "loss": 0.1942,
+      "step": 6842
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00021062755102040814,
+      "loss": 0.1826,
+      "step": 6843
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00021061224489795917,
+      "loss": 0.2292,
+      "step": 6844
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00021059693877551017,
+      "loss": 0.167,
+      "step": 6845
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00021058163265306123,
+      "loss": 0.1659,
+      "step": 6846
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00021056632653061223,
+      "loss": 0.1487,
+      "step": 6847
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00021055102040816323,
+      "loss": 0.137,
+      "step": 6848
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00021053571428571426,
+      "loss": 0.2007,
+      "step": 6849
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.0002105204081632653,
+      "loss": 0.23,
+      "step": 6850
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.0002105051020408163,
+      "loss": 0.2029,
+      "step": 6851
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00021048979591836733,
+      "loss": 0.1471,
+      "step": 6852
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00021047448979591833,
+      "loss": 0.0888,
+      "step": 6853
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.0002104591836734694,
+      "loss": 0.1243,
+      "step": 6854
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.0002104438775510204,
+      "loss": 0.1169,
+      "step": 6855
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.0002104285714285714,
+      "loss": 0.2364,
+      "step": 6856
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00021041326530612246,
+      "loss": 0.171,
+      "step": 6857
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00021039795918367346,
+      "loss": 0.1736,
+      "step": 6858
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00021038265306122446,
+      "loss": 0.1171,
+      "step": 6859
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.0002103673469387755,
+      "loss": 0.101,
+      "step": 6860
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00021035204081632652,
+      "loss": 0.1322,
+      "step": 6861
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00021033673469387753,
+      "loss": 0.1976,
+      "step": 6862
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00021032142857142856,
+      "loss": 0.1309,
+      "step": 6863
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00021030612244897956,
+      "loss": 0.1421,
+      "step": 6864
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021029081632653062,
+      "loss": 0.155,
+      "step": 6865
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021027551020408162,
+      "loss": 0.125,
+      "step": 6866
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021026020408163262,
+      "loss": 0.1366,
+      "step": 6867
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021024489795918366,
+      "loss": 0.1768,
+      "step": 6868
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021022959183673469,
+      "loss": 0.1992,
+      "step": 6869
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.0002102142857142857,
+      "loss": 0.1001,
+      "step": 6870
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021019897959183672,
+      "loss": 0.0887,
+      "step": 6871
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021018367346938772,
+      "loss": 0.1161,
+      "step": 6872
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021016836734693878,
+      "loss": 0.2752,
+      "step": 6873
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00021015306122448978,
+      "loss": 0.2936,
+      "step": 6874
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.0002101377551020408,
+      "loss": 0.1231,
+      "step": 6875
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00021012244897959182,
+      "loss": 0.1101,
+      "step": 6876
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00021010714285714285,
+      "loss": 0.1158,
+      "step": 6877
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00021009183673469385,
+      "loss": 0.1395,
+      "step": 6878
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00021007653061224488,
+      "loss": 0.1005,
+      "step": 6879
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.0002100612244897959,
+      "loss": 0.254,
+      "step": 6880
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00021004591836734694,
+      "loss": 0.2594,
+      "step": 6881
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00021003061224489795,
+      "loss": 0.1377,
+      "step": 6882
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00021001530612244895,
+      "loss": 0.1083,
+      "step": 6883
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00020999999999999998,
+      "loss": 0.1511,
+      "step": 6884
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.000209984693877551,
+      "loss": 0.1299,
+      "step": 6885
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00020996938775510202,
+      "loss": 0.1228,
+      "step": 6886
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00020995408163265305,
+      "loss": 0.2655,
+      "step": 6887
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00020993877551020405,
+      "loss": 0.2079,
+      "step": 6888
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0002099234693877551,
+      "loss": 0.1598,
+      "step": 6889
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0002099081632653061,
+      "loss": 0.1038,
+      "step": 6890
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00020989285714285711,
+      "loss": 0.1231,
+      "step": 6891
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00020987755102040812,
+      "loss": 0.1362,
+      "step": 6892
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00020986224489795918,
+      "loss": 0.2274,
+      "step": 6893
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00020984693877551018,
+      "loss": 0.3021,
+      "step": 6894
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0002098316326530612,
+      "loss": 0.1392,
+      "step": 6895
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.0002098163265306122,
+      "loss": 0.1607,
+      "step": 6896
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020980102040816324,
+      "loss": 0.0912,
+      "step": 6897
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020978571428571427,
+      "loss": 0.2051,
+      "step": 6898
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020977040816326528,
+      "loss": 0.2225,
+      "step": 6899
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020975510204081628,
+      "loss": 0.2232,
+      "step": 6900
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020973979591836734,
+      "loss": 0.1679,
+      "step": 6901
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020972448979591834,
+      "loss": 0.1181,
+      "step": 6902
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020970918367346937,
+      "loss": 0.1394,
+      "step": 6903
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.0002096938775510204,
+      "loss": 0.1238,
+      "step": 6904
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.0002096785714285714,
+      "loss": 0.1716,
+      "step": 6905
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020966326530612244,
+      "loss": 0.2438,
+      "step": 6906
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.00020964795918367344,
+      "loss": 0.2803,
+      "step": 6907
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.0002096326530612245,
+      "loss": 0.1478,
+      "step": 6908
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.0002096173469387755,
+      "loss": 0.106,
+      "step": 6909
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.0002096020408163265,
+      "loss": 0.1568,
+      "step": 6910
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.00020958673469387754,
+      "loss": 0.1229,
+      "step": 6911
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.00020957142857142857,
+      "loss": 0.2121,
+      "step": 6912
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.00020955612244897957,
+      "loss": 0.186,
+      "step": 6913
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.0002095408163265306,
+      "loss": 0.0951,
+      "step": 6914
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.0002095255102040816,
+      "loss": 0.154,
+      "step": 6915
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.00020951020408163266,
+      "loss": 0.2314,
+      "step": 6916
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020949489795918366,
+      "loss": 0.1333,
+      "step": 6917
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020947959183673467,
+      "loss": 0.1545,
+      "step": 6918
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020946428571428567,
+      "loss": 0.1177,
+      "step": 6919
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020944897959183673,
+      "loss": 0.1751,
+      "step": 6920
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020943367346938773,
+      "loss": 0.1293,
+      "step": 6921
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020941836734693876,
+      "loss": 0.142,
+      "step": 6922
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020940306122448977,
+      "loss": 0.1751,
+      "step": 6923
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020938775510204082,
+      "loss": 0.1507,
+      "step": 6924
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020937244897959183,
+      "loss": 0.1314,
+      "step": 6925
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020935714285714283,
+      "loss": 0.1287,
+      "step": 6926
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00020934183673469383,
+      "loss": 0.195,
+      "step": 6927
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.0002093265306122449,
+      "loss": 0.1612,
+      "step": 6928
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.0002093112244897959,
+      "loss": 0.151,
+      "step": 6929
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00020929591836734693,
+      "loss": 0.2356,
+      "step": 6930
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00020928061224489793,
+      "loss": 0.1399,
+      "step": 6931
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00020926530612244896,
+      "loss": 0.1352,
+      "step": 6932
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00020925,
+      "loss": 0.1505,
+      "step": 6933
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.000209234693877551,
+      "loss": 0.0865,
+      "step": 6934
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.000209219387755102,
+      "loss": 0.1411,
+      "step": 6935
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00020920408163265306,
+      "loss": 0.1225,
+      "step": 6936
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00020918877551020406,
+      "loss": 0.2547,
+      "step": 6937
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.0002091734693877551,
+      "loss": 0.1691,
+      "step": 6938
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.0002091581632653061,
+      "loss": 0.1199,
+      "step": 6939
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00020914285714285712,
+      "loss": 0.1621,
+      "step": 6940
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00020912755102040815,
+      "loss": 0.1169,
+      "step": 6941
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00020911224489795916,
+      "loss": 0.1067,
+      "step": 6942
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00020909693877551016,
+      "loss": 0.2305,
+      "step": 6943
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00020908163265306122,
+      "loss": 0.1616,
+      "step": 6944
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00020906632653061222,
+      "loss": 0.1193,
+      "step": 6945
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00020905102040816325,
+      "loss": 0.1987,
+      "step": 6946
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00020903571428571426,
+      "loss": 0.1513,
+      "step": 6947
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020902040816326529,
+      "loss": 0.1563,
+      "step": 6948
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020900510204081632,
+      "loss": 0.2781,
+      "step": 6949
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020898979591836732,
+      "loss": 0.198,
+      "step": 6950
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020897448979591838,
+      "loss": 0.1236,
+      "step": 6951
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020895918367346938,
+      "loss": 0.1798,
+      "step": 6952
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020894387755102038,
+      "loss": 0.2181,
+      "step": 6953
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.0002089285714285714,
+      "loss": 0.1693,
+      "step": 6954
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020891326530612245,
+      "loss": 0.1349,
+      "step": 6955
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020889795918367345,
+      "loss": 0.3774,
+      "step": 6956
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020888265306122448,
+      "loss": 0.1098,
+      "step": 6957
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020886734693877548,
+      "loss": 0.1464,
+      "step": 6958
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020885204081632651,
+      "loss": 0.1169,
+      "step": 6959
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020883673469387754,
+      "loss": 0.1807,
+      "step": 6960
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020882142857142855,
+      "loss": 0.1761,
+      "step": 6961
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020880612244897955,
+      "loss": 0.2125,
+      "step": 6962
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.0002087908163265306,
+      "loss": 0.2525,
+      "step": 6963
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.0002087755102040816,
+      "loss": 0.2697,
+      "step": 6964
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020876020408163264,
+      "loss": 0.2154,
+      "step": 6965
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020874489795918365,
+      "loss": 0.116,
+      "step": 6966
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020872959183673468,
+      "loss": 0.1536,
+      "step": 6967
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.0002087142857142857,
+      "loss": 0.243,
+      "step": 6968
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.0002086989795918367,
+      "loss": 0.2775,
+      "step": 6969
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00020868367346938771,
+      "loss": 0.1667,
+      "step": 6970
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00020866836734693877,
+      "loss": 0.2083,
+      "step": 6971
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00020865306122448978,
+      "loss": 0.1143,
+      "step": 6972
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.0002086377551020408,
+      "loss": 0.2213,
+      "step": 6973
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.0002086224489795918,
+      "loss": 0.3567,
+      "step": 6974
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00020860714285714284,
+      "loss": 0.1346,
+      "step": 6975
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00020859183673469387,
+      "loss": 0.128,
+      "step": 6976
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00020857653061224487,
+      "loss": 0.1791,
+      "step": 6977
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00020856122448979588,
+      "loss": 0.3155,
+      "step": 6978
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00020854591836734693,
+      "loss": 0.1431,
+      "step": 6979
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00020853061224489794,
+      "loss": 0.224,
+      "step": 6980
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00020851530612244894,
+      "loss": 0.1563,
+      "step": 6981
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00020849999999999997,
+      "loss": 0.1675,
+      "step": 6982
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.000208484693877551,
+      "loss": 0.1292,
+      "step": 6983
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00020846938775510203,
+      "loss": 0.1734,
+      "step": 6984
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00020845408163265304,
+      "loss": 0.1011,
+      "step": 6985
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00020843877551020404,
+      "loss": 0.1206,
+      "step": 6986
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.0002084234693877551,
+      "loss": 0.1934,
+      "step": 6987
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.0002084081632653061,
+      "loss": 0.1277,
+      "step": 6988
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.0002083928571428571,
+      "loss": 0.108,
+      "step": 6989
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00020837755102040814,
+      "loss": 0.1074,
+      "step": 6990
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00020836224489795917,
+      "loss": 0.1087,
+      "step": 6991
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.0002083469387755102,
+      "loss": 0.1047,
+      "step": 6992
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.0002083316326530612,
+      "loss": 0.1301,
+      "step": 6993
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00020831632653061223,
+      "loss": 0.1865,
+      "step": 6994
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00020830102040816326,
+      "loss": 0.0986,
+      "step": 6995
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00020828571428571426,
+      "loss": 0.1551,
+      "step": 6996
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00020827040816326527,
+      "loss": 0.1025,
+      "step": 6997
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00020825510204081633,
+      "loss": 0.1209,
+      "step": 6998
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00020823979591836733,
+      "loss": 0.2518,
+      "step": 6999
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00020822448979591836,
+      "loss": 0.1145,
+      "step": 7000
+    },
+    {
+      "epoch": 6.8,
+      "eval_cer": 0.2593032272888285,
+      "eval_loss": 0.9128949046134949,
+      "eval_runtime": 244.101,
+      "eval_samples_per_second": 36.108,
+      "eval_steps_per_second": 4.515,
+      "eval_wer": 0.6679904206351822,
+      "step": 7000
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00020820918367346936,
+      "loss": 0.1134,
+      "step": 7001
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0002081938775510204,
+      "loss": 0.1185,
+      "step": 7002
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00020817857142857142,
+      "loss": 0.1925,
+      "step": 7003
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00020816326530612243,
+      "loss": 0.1438,
+      "step": 7004
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00020814795918367343,
+      "loss": 0.1598,
+      "step": 7005
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0002081326530612245,
+      "loss": 0.239,
+      "step": 7006
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0002081173469387755,
+      "loss": 0.1656,
+      "step": 7007
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00020810204081632652,
+      "loss": 0.1332,
+      "step": 7008
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00020808673469387753,
+      "loss": 0.242,
+      "step": 7009
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00020807142857142856,
+      "loss": 0.2191,
+      "step": 7010
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0002080561224489796,
+      "loss": 0.1196,
+      "step": 7011
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0002080408163265306,
+      "loss": 0.2426,
+      "step": 7012
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0002080255102040816,
+      "loss": 0.1873,
+      "step": 7013
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00020801020408163265,
+      "loss": 0.1031,
+      "step": 7014
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00020799489795918366,
+      "loss": 0.1672,
+      "step": 7015
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00020797959183673466,
+      "loss": 0.1301,
+      "step": 7016
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0002079642857142857,
+      "loss": 0.1644,
+      "step": 7017
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00020794897959183672,
+      "loss": 0.2202,
+      "step": 7018
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00020793367346938775,
+      "loss": 0.2083,
+      "step": 7019
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020791836734693875,
+      "loss": 0.1642,
+      "step": 7020
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020790306122448976,
+      "loss": 0.0884,
+      "step": 7021
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020788775510204081,
+      "loss": 0.121,
+      "step": 7022
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020787244897959182,
+      "loss": 0.0913,
+      "step": 7023
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020785714285714282,
+      "loss": 0.348,
+      "step": 7024
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020784183673469385,
+      "loss": 0.1987,
+      "step": 7025
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020782653061224488,
+      "loss": 0.2084,
+      "step": 7026
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.0002078112244897959,
+      "loss": 0.1506,
+      "step": 7027
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020779591836734692,
+      "loss": 0.1272,
+      "step": 7028
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020778061224489792,
+      "loss": 0.0814,
+      "step": 7029
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020776530612244898,
+      "loss": 0.2913,
+      "step": 7030
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020774999999999998,
+      "loss": 0.2452,
+      "step": 7031
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020773469387755098,
+      "loss": 0.2213,
+      "step": 7032
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020771938775510202,
+      "loss": 0.0989,
+      "step": 7033
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020770408163265305,
+      "loss": 0.107,
+      "step": 7034
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020768877551020408,
+      "loss": 0.1109,
+      "step": 7035
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020767346938775508,
+      "loss": 0.1751,
+      "step": 7036
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020765816326530608,
+      "loss": 0.1992,
+      "step": 7037
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020764285714285714,
+      "loss": 0.1178,
+      "step": 7038
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020762755102040814,
+      "loss": 0.116,
+      "step": 7039
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00020761224489795915,
+      "loss": 0.1402,
+      "step": 7040
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0002075969387755102,
+      "loss": 0.0882,
+      "step": 7041
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0002075816326530612,
+      "loss": 0.0819,
+      "step": 7042
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.00020756632653061224,
+      "loss": 0.2111,
+      "step": 7043
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.00020755102040816324,
+      "loss": 0.1366,
+      "step": 7044
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.00020753571428571427,
+      "loss": 0.1387,
+      "step": 7045
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0002075204081632653,
+      "loss": 0.213,
+      "step": 7046
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0002075051020408163,
+      "loss": 0.103,
+      "step": 7047
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0002074897959183673,
+      "loss": 0.1275,
+      "step": 7048
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.00020747448979591837,
+      "loss": 0.2016,
+      "step": 7049
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.00020745918367346937,
+      "loss": 0.1841,
+      "step": 7050
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020744387755102038,
+      "loss": 0.1871,
+      "step": 7051
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.0002074285714285714,
+      "loss": 0.1358,
+      "step": 7052
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020741326530612244,
+      "loss": 0.1557,
+      "step": 7053
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020739795918367347,
+      "loss": 0.1652,
+      "step": 7054
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020738265306122447,
+      "loss": 0.1134,
+      "step": 7055
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020736734693877547,
+      "loss": 0.1846,
+      "step": 7056
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020735204081632653,
+      "loss": 0.1638,
+      "step": 7057
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020733673469387753,
+      "loss": 0.1817,
+      "step": 7058
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020732142857142854,
+      "loss": 0.1465,
+      "step": 7059
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00020730612244897957,
+      "loss": 0.116,
+      "step": 7060
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.0002072908163265306,
+      "loss": 0.1428,
+      "step": 7061
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00020727551020408163,
+      "loss": 0.3541,
+      "step": 7062
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00020726020408163263,
+      "loss": 0.181,
+      "step": 7063
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00020724489795918364,
+      "loss": 0.1274,
+      "step": 7064
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.0002072295918367347,
+      "loss": 0.1377,
+      "step": 7065
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.0002072142857142857,
+      "loss": 0.1502,
+      "step": 7066
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.0002071989795918367,
+      "loss": 0.1989,
+      "step": 7067
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00020718367346938773,
+      "loss": 0.1875,
+      "step": 7068
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00020716836734693876,
+      "loss": 0.1588,
+      "step": 7069
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.0002071530612244898,
+      "loss": 0.1905,
+      "step": 7070
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.0002071377551020408,
+      "loss": 0.1267,
+      "step": 7071
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.0002071224489795918,
+      "loss": 0.1075,
+      "step": 7072
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020710714285714286,
+      "loss": 0.1446,
+      "step": 7073
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020709183673469386,
+      "loss": 0.1834,
+      "step": 7074
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020707653061224486,
+      "loss": 0.2717,
+      "step": 7075
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.0002070612244897959,
+      "loss": 0.1955,
+      "step": 7076
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020704591836734693,
+      "loss": 0.2304,
+      "step": 7077
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020703061224489793,
+      "loss": 0.1546,
+      "step": 7078
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020701530612244896,
+      "loss": 0.0992,
+      "step": 7079
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020699999999999996,
+      "loss": 0.2749,
+      "step": 7080
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020698469387755102,
+      "loss": 0.2435,
+      "step": 7081
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00020696938775510202,
+      "loss": 0.1312,
+      "step": 7082
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00020695408163265303,
+      "loss": 0.1153,
+      "step": 7083
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00020693877551020409,
+      "loss": 0.1497,
+      "step": 7084
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.0002069234693877551,
+      "loss": 0.1916,
+      "step": 7085
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.0002069081632653061,
+      "loss": 0.1062,
+      "step": 7086
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00020689285714285712,
+      "loss": 0.2345,
+      "step": 7087
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00020687755102040815,
+      "loss": 0.2019,
+      "step": 7088
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00020686224489795918,
+      "loss": 0.1081,
+      "step": 7089
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.0002068469387755102,
+      "loss": 0.2093,
+      "step": 7090
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.0002068316326530612,
+      "loss": 0.1542,
+      "step": 7091
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020681632653061225,
+      "loss": 0.0906,
+      "step": 7092
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020680102040816325,
+      "loss": 0.2692,
+      "step": 7093
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020678571428571426,
+      "loss": 0.2065,
+      "step": 7094
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020677040816326529,
+      "loss": 0.249,
+      "step": 7095
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020675510204081632,
+      "loss": 0.2286,
+      "step": 7096
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020673979591836735,
+      "loss": 0.1595,
+      "step": 7097
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020672448979591835,
+      "loss": 0.0881,
+      "step": 7098
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020670918367346935,
+      "loss": 0.2038,
+      "step": 7099
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.0002066938775510204,
+      "loss": 0.1748,
+      "step": 7100
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00020667857142857141,
+      "loss": 0.1433,
+      "step": 7101
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020666326530612242,
+      "loss": 0.1668,
+      "step": 7102
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020664795918367345,
+      "loss": 0.116,
+      "step": 7103
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020663265306122448,
+      "loss": 0.127,
+      "step": 7104
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0002066173469387755,
+      "loss": 0.2582,
+      "step": 7105
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0002066020408163265,
+      "loss": 0.1503,
+      "step": 7106
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020658673469387752,
+      "loss": 0.1322,
+      "step": 7107
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020657142857142857,
+      "loss": 0.1238,
+      "step": 7108
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020655612244897958,
+      "loss": 0.117,
+      "step": 7109
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020654081632653058,
+      "loss": 0.1444,
+      "step": 7110
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0002065255102040816,
+      "loss": 0.1994,
+      "step": 7111
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020651020408163264,
+      "loss": 0.2796,
+      "step": 7112
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020649489795918365,
+      "loss": 0.1454,
+      "step": 7113
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020647959183673468,
+      "loss": 0.143,
+      "step": 7114
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020646428571428568,
+      "loss": 0.13,
+      "step": 7115
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020644897959183674,
+      "loss": 0.1366,
+      "step": 7116
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020643367346938774,
+      "loss": 0.1417,
+      "step": 7117
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020641836734693874,
+      "loss": 0.2115,
+      "step": 7118
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020640306122448977,
+      "loss": 0.1826,
+      "step": 7119
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.0002063877551020408,
+      "loss": 0.1417,
+      "step": 7120
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.0002063724489795918,
+      "loss": 0.1373,
+      "step": 7121
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020635714285714284,
+      "loss": 0.1766,
+      "step": 7122
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.00020634183673469384,
+      "loss": 0.1571,
+      "step": 7123
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0002063265306122449,
+      "loss": 0.3198,
+      "step": 7124
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0002063112244897959,
+      "loss": 0.2306,
+      "step": 7125
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0002062959183673469,
+      "loss": 0.2506,
+      "step": 7126
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.00020628061224489794,
+      "loss": 0.1654,
+      "step": 7127
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.00020626530612244897,
+      "loss": 0.1698,
+      "step": 7128
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.00020624999999999997,
+      "loss": 0.1301,
+      "step": 7129
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.000206234693877551,
+      "loss": 0.2533,
+      "step": 7130
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.00020621938775510203,
+      "loss": 0.221,
+      "step": 7131
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.00020620408163265306,
+      "loss": 0.212,
+      "step": 7132
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00020618877551020407,
+      "loss": 0.1522,
+      "step": 7133
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00020617346938775507,
+      "loss": 0.1963,
+      "step": 7134
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00020615816326530613,
+      "loss": 0.1081,
+      "step": 7135
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00020614285714285713,
+      "loss": 0.1099,
+      "step": 7136
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00020612755102040813,
+      "loss": 0.2047,
+      "step": 7137
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00020611224489795917,
+      "loss": 0.1052,
+      "step": 7138
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.0002060969387755102,
+      "loss": 0.1106,
+      "step": 7139
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.0002060816326530612,
+      "loss": 0.1516,
+      "step": 7140
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00020606632653061223,
+      "loss": 0.1175,
+      "step": 7141
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00020605102040816323,
+      "loss": 0.1041,
+      "step": 7142
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.0002060357142857143,
+      "loss": 0.1889,
+      "step": 7143
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0002060204081632653,
+      "loss": 0.1252,
+      "step": 7144
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0002060051020408163,
+      "loss": 0.1088,
+      "step": 7145
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00020598979591836733,
+      "loss": 0.0948,
+      "step": 7146
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00020597448979591836,
+      "loss": 0.1993,
+      "step": 7147
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00020595918367346936,
+      "loss": 0.0845,
+      "step": 7148
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0002059438775510204,
+      "loss": 0.358,
+      "step": 7149
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0002059285714285714,
+      "loss": 0.2117,
+      "step": 7150
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00020591326530612245,
+      "loss": 0.1685,
+      "step": 7151
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00020589795918367346,
+      "loss": 0.1294,
+      "step": 7152
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00020588265306122446,
+      "loss": 0.1435,
+      "step": 7153
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.0002058673469387755,
+      "loss": 0.1496,
+      "step": 7154
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020585204081632652,
+      "loss": 0.1449,
+      "step": 7155
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020583673469387753,
+      "loss": 0.2899,
+      "step": 7156
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020582142857142856,
+      "loss": 0.3075,
+      "step": 7157
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020580612244897956,
+      "loss": 0.1261,
+      "step": 7158
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020579081632653062,
+      "loss": 0.1686,
+      "step": 7159
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020577551020408162,
+      "loss": 0.0977,
+      "step": 7160
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020576020408163262,
+      "loss": 0.243,
+      "step": 7161
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020574489795918365,
+      "loss": 0.2178,
+      "step": 7162
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020572959183673469,
+      "loss": 0.1585,
+      "step": 7163
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.0002057142857142857,
+      "loss": 0.154,
+      "step": 7164
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00020569897959183672,
+      "loss": 0.1593,
+      "step": 7165
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00020568367346938772,
+      "loss": 0.1499,
+      "step": 7166
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00020566836734693878,
+      "loss": 0.1099,
+      "step": 7167
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00020565306122448978,
+      "loss": 0.221,
+      "step": 7168
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.0002056377551020408,
+      "loss": 0.2627,
+      "step": 7169
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.0002056224489795918,
+      "loss": 0.1591,
+      "step": 7170
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00020560714285714285,
+      "loss": 0.1085,
+      "step": 7171
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00020559183673469385,
+      "loss": 0.1048,
+      "step": 7172
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00020557653061224488,
+      "loss": 0.1521,
+      "step": 7173
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00020556122448979589,
+      "loss": 0.1344,
+      "step": 7174
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00020554591836734692,
+      "loss": 0.1503,
+      "step": 7175
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00020553061224489795,
+      "loss": 0.1529,
+      "step": 7176
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00020551530612244895,
+      "loss": 0.122,
+      "step": 7177
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.0002055,
+      "loss": 0.2364,
+      "step": 7178
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.000205484693877551,
+      "loss": 0.0773,
+      "step": 7179
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00020546938775510201,
+      "loss": 0.1962,
+      "step": 7180
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00020545408163265305,
+      "loss": 0.2327,
+      "step": 7181
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00020543877551020408,
+      "loss": 0.1932,
+      "step": 7182
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00020542346938775508,
+      "loss": 0.099,
+      "step": 7183
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.0002054081632653061,
+      "loss": 0.136,
+      "step": 7184
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.0002053928571428571,
+      "loss": 0.1531,
+      "step": 7185
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020537755102040817,
+      "loss": 0.1453,
+      "step": 7186
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020536224489795917,
+      "loss": 0.1816,
+      "step": 7187
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020534693877551018,
+      "loss": 0.1029,
+      "step": 7188
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.0002053316326530612,
+      "loss": 0.1108,
+      "step": 7189
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020531632653061224,
+      "loss": 0.1688,
+      "step": 7190
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020530102040816324,
+      "loss": 0.1858,
+      "step": 7191
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020528571428571427,
+      "loss": 0.2079,
+      "step": 7192
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020527040816326528,
+      "loss": 0.1503,
+      "step": 7193
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020525510204081633,
+      "loss": 0.129,
+      "step": 7194
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.00020523979591836734,
+      "loss": 0.2286,
+      "step": 7195
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.00020522448979591834,
+      "loss": 0.1539,
+      "step": 7196
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.00020520918367346934,
+      "loss": 0.0986,
+      "step": 7197
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0002051938775510204,
+      "loss": 0.1012,
+      "step": 7198
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0002051785714285714,
+      "loss": 0.2898,
+      "step": 7199
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.00020516326530612244,
+      "loss": 0.1902,
+      "step": 7200
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.00020514795918367344,
+      "loss": 0.2351,
+      "step": 7201
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0002051326530612245,
+      "loss": 0.2113,
+      "step": 7202
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0002051173469387755,
+      "loss": 0.1596,
+      "step": 7203
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0002051020408163265,
+      "loss": 0.1503,
+      "step": 7204
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.0002050867346938775,
+      "loss": 0.1692,
+      "step": 7205
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00020507142857142856,
+      "loss": 0.2612,
+      "step": 7206
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00020505612244897957,
+      "loss": 0.1385,
+      "step": 7207
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.0002050408163265306,
+      "loss": 0.1867,
+      "step": 7208
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.0002050255102040816,
+      "loss": 0.1977,
+      "step": 7209
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00020501020408163263,
+      "loss": 0.25,
+      "step": 7210
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00020499489795918366,
+      "loss": 0.2491,
+      "step": 7211
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00020497959183673467,
+      "loss": 0.1287,
+      "step": 7212
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00020496428571428567,
+      "loss": 0.1089,
+      "step": 7213
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00020494897959183673,
+      "loss": 0.1051,
+      "step": 7214
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00020493367346938773,
+      "loss": 0.1512,
+      "step": 7215
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00020491836734693876,
+      "loss": 0.1188,
+      "step": 7216
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00020490306122448977,
+      "loss": 0.1527,
+      "step": 7217
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0002048877551020408,
+      "loss": 0.1485,
+      "step": 7218
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00020487244897959183,
+      "loss": 0.1493,
+      "step": 7219
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00020485714285714283,
+      "loss": 0.1901,
+      "step": 7220
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0002048418367346939,
+      "loss": 0.0928,
+      "step": 7221
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0002048265306122449,
+      "loss": 0.073,
+      "step": 7222
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0002048112244897959,
+      "loss": 0.2398,
+      "step": 7223
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00020479591836734692,
+      "loss": 0.152,
+      "step": 7224
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00020478061224489796,
+      "loss": 0.1205,
+      "step": 7225
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020476530612244896,
+      "loss": 0.1376,
+      "step": 7226
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020475,
+      "loss": 0.1288,
+      "step": 7227
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.000204734693877551,
+      "loss": 0.1031,
+      "step": 7228
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020471938775510205,
+      "loss": 0.2255,
+      "step": 7229
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020470408163265305,
+      "loss": 0.1702,
+      "step": 7230
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020468877551020406,
+      "loss": 0.1322,
+      "step": 7231
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020467346938775506,
+      "loss": 0.1042,
+      "step": 7232
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020465816326530612,
+      "loss": 0.1228,
+      "step": 7233
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020464285714285712,
+      "loss": 0.0824,
+      "step": 7234
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00020462755102040815,
+      "loss": 0.2344,
+      "step": 7235
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020461224489795916,
+      "loss": 0.2448,
+      "step": 7236
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.0002045969387755102,
+      "loss": 0.1064,
+      "step": 7237
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020458163265306122,
+      "loss": 0.1377,
+      "step": 7238
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020456632653061222,
+      "loss": 0.1008,
+      "step": 7239
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020455102040816322,
+      "loss": 0.0587,
+      "step": 7240
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020453571428571428,
+      "loss": 0.1559,
+      "step": 7241
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020452040816326529,
+      "loss": 0.3059,
+      "step": 7242
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020450510204081632,
+      "loss": 0.2018,
+      "step": 7243
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020448979591836732,
+      "loss": 0.1379,
+      "step": 7244
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020447448979591835,
+      "loss": 0.1484,
+      "step": 7245
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00020445918367346938,
+      "loss": 0.1183,
+      "step": 7246
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00020444387755102038,
+      "loss": 0.1479,
+      "step": 7247
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.0002044285714285714,
+      "loss": 0.2181,
+      "step": 7248
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00020441326530612244,
+      "loss": 0.1106,
+      "step": 7249
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00020439795918367345,
+      "loss": 0.1053,
+      "step": 7250
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00020438265306122448,
+      "loss": 0.1657,
+      "step": 7251
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00020436734693877548,
+      "loss": 0.0962,
+      "step": 7252
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.0002043520408163265,
+      "loss": 0.1075,
+      "step": 7253
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00020433673469387754,
+      "loss": 0.1849,
+      "step": 7254
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00020432142857142855,
+      "loss": 0.1909,
+      "step": 7255
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00020430612244897955,
+      "loss": 0.1493,
+      "step": 7256
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0002042908163265306,
+      "loss": 0.1501,
+      "step": 7257
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0002042755102040816,
+      "loss": 0.1236,
+      "step": 7258
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00020426020408163261,
+      "loss": 0.1679,
+      "step": 7259
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00020424489795918365,
+      "loss": 0.2073,
+      "step": 7260
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00020422959183673468,
+      "loss": 0.203,
+      "step": 7261
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0002042142857142857,
+      "loss": 0.1103,
+      "step": 7262
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0002041989795918367,
+      "loss": 0.1147,
+      "step": 7263
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0002041836734693877,
+      "loss": 0.1968,
+      "step": 7264
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00020416836734693877,
+      "loss": 0.0676,
+      "step": 7265
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00020415306122448977,
+      "loss": 0.1378,
+      "step": 7266
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00020413775510204078,
+      "loss": 0.1936,
+      "step": 7267
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00020412244897959184,
+      "loss": 0.1572,
+      "step": 7268
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00020410714285714284,
+      "loss": 0.0962,
+      "step": 7269
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00020409183673469387,
+      "loss": 0.1679,
+      "step": 7270
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00020407653061224487,
+      "loss": 0.2005,
+      "step": 7271
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.0002040612244897959,
+      "loss": 0.056,
+      "step": 7272
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00020404591836734693,
+      "loss": 0.3335,
+      "step": 7273
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00020403061224489794,
+      "loss": 0.0968,
+      "step": 7274
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00020401530612244894,
+      "loss": 0.1464,
+      "step": 7275
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.000204,
+      "loss": 0.1756,
+      "step": 7276
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.000203984693877551,
+      "loss": 0.1551,
+      "step": 7277
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00020396938775510203,
+      "loss": 0.1214,
+      "step": 7278
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00020395408163265304,
+      "loss": 0.164,
+      "step": 7279
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00020393877551020407,
+      "loss": 0.1728,
+      "step": 7280
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0002039234693877551,
+      "loss": 0.1057,
+      "step": 7281
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0002039081632653061,
+      "loss": 0.1403,
+      "step": 7282
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0002038928571428571,
+      "loss": 0.0678,
+      "step": 7283
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00020387755102040816,
+      "loss": 0.1318,
+      "step": 7284
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00020386224489795916,
+      "loss": 0.1589,
+      "step": 7285
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0002038469387755102,
+      "loss": 0.1802,
+      "step": 7286
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0002038316326530612,
+      "loss": 0.1767,
+      "step": 7287
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020381632653061223,
+      "loss": 0.0917,
+      "step": 7288
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020380102040816326,
+      "loss": 0.1365,
+      "step": 7289
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020378571428571426,
+      "loss": 0.1842,
+      "step": 7290
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020377040816326527,
+      "loss": 0.1846,
+      "step": 7291
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020375510204081632,
+      "loss": 0.1843,
+      "step": 7292
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020373979591836733,
+      "loss": 0.1281,
+      "step": 7293
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020372448979591833,
+      "loss": 0.1048,
+      "step": 7294
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020370918367346936,
+      "loss": 0.1173,
+      "step": 7295
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.0002036938775510204,
+      "loss": 0.1498,
+      "step": 7296
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00020367857142857142,
+      "loss": 0.1192,
+      "step": 7297
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00020366326530612243,
+      "loss": 0.2359,
+      "step": 7298
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00020364795918367343,
+      "loss": 0.1176,
+      "step": 7299
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.0002036326530612245,
+      "loss": 0.1083,
+      "step": 7300
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.0002036173469387755,
+      "loss": 0.1019,
+      "step": 7301
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.0002036020408163265,
+      "loss": 0.1842,
+      "step": 7302
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00020358673469387752,
+      "loss": 0.1425,
+      "step": 7303
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00020357142857142856,
+      "loss": 0.1666,
+      "step": 7304
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00020355612244897959,
+      "loss": 0.1656,
+      "step": 7305
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.0002035408163265306,
+      "loss": 0.1274,
+      "step": 7306
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.0002035255102040816,
+      "loss": 0.1352,
+      "step": 7307
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020351020408163265,
+      "loss": 0.1181,
+      "step": 7308
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020349489795918365,
+      "loss": 0.1107,
+      "step": 7309
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020347959183673466,
+      "loss": 0.1978,
+      "step": 7310
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020346428571428572,
+      "loss": 0.14,
+      "step": 7311
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020344897959183672,
+      "loss": 0.2152,
+      "step": 7312
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020343367346938775,
+      "loss": 0.1337,
+      "step": 7313
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020341836734693875,
+      "loss": 0.1772,
+      "step": 7314
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020340306122448978,
+      "loss": 0.065,
+      "step": 7315
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020338775510204081,
+      "loss": 0.1999,
+      "step": 7316
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020337244897959182,
+      "loss": 0.1953,
+      "step": 7317
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00020335714285714282,
+      "loss": 0.1833,
+      "step": 7318
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020334183673469388,
+      "loss": 0.1567,
+      "step": 7319
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020332653061224488,
+      "loss": 0.1357,
+      "step": 7320
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020331122448979589,
+      "loss": 0.0756,
+      "step": 7321
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020329591836734692,
+      "loss": 0.1797,
+      "step": 7322
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020328061224489795,
+      "loss": 0.1894,
+      "step": 7323
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020326530612244898,
+      "loss": 0.1448,
+      "step": 7324
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020324999999999998,
+      "loss": 0.1827,
+      "step": 7325
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020323469387755098,
+      "loss": 0.1767,
+      "step": 7326
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020321938775510204,
+      "loss": 0.0889,
+      "step": 7327
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00020320408163265304,
+      "loss": 0.0667,
+      "step": 7328
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00020318877551020405,
+      "loss": 0.2231,
+      "step": 7329
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00020317346938775508,
+      "loss": 0.0831,
+      "step": 7330
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.0002031581632653061,
+      "loss": 0.1337,
+      "step": 7331
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00020314285714285714,
+      "loss": 0.143,
+      "step": 7332
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00020312755102040814,
+      "loss": 0.0757,
+      "step": 7333
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00020311224489795915,
+      "loss": 0.119,
+      "step": 7334
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.0002030969387755102,
+      "loss": 0.1722,
+      "step": 7335
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.0002030816326530612,
+      "loss": 0.1673,
+      "step": 7336
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.0002030663265306122,
+      "loss": 0.17,
+      "step": 7337
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00020305102040816324,
+      "loss": 0.1101,
+      "step": 7338
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00020303571428571427,
+      "loss": 0.0935,
+      "step": 7339
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0002030204081632653,
+      "loss": 0.1109,
+      "step": 7340
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0002030051020408163,
+      "loss": 0.1081,
+      "step": 7341
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0002029897959183673,
+      "loss": 0.1897,
+      "step": 7342
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00020297448979591837,
+      "loss": 0.24,
+      "step": 7343
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00020295918367346937,
+      "loss": 0.1561,
+      "step": 7344
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00020294387755102037,
+      "loss": 0.0866,
+      "step": 7345
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0002029285714285714,
+      "loss": 0.1561,
+      "step": 7346
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00020291326530612244,
+      "loss": 0.126,
+      "step": 7347
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00020289795918367347,
+      "loss": 0.1762,
+      "step": 7348
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00020288265306122447,
+      "loss": 0.2615,
+      "step": 7349
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00020286734693877547,
+      "loss": 0.1043,
+      "step": 7350
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00020285204081632653,
+      "loss": 0.1534,
+      "step": 7351
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00020283673469387753,
+      "loss": 0.0892,
+      "step": 7352
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00020282142857142854,
+      "loss": 0.087,
+      "step": 7353
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00020280612244897957,
+      "loss": 0.2892,
+      "step": 7354
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.0002027908163265306,
+      "loss": 0.1413,
+      "step": 7355
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.0002027755102040816,
+      "loss": 0.1121,
+      "step": 7356
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00020276020408163263,
+      "loss": 0.1628,
+      "step": 7357
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00020274489795918366,
+      "loss": 0.0982,
+      "step": 7358
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.0002027295918367347,
+      "loss": 0.0803,
+      "step": 7359
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.0002027142857142857,
+      "loss": 0.2153,
+      "step": 7360
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.0002026989795918367,
+      "loss": 0.2675,
+      "step": 7361
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00020268367346938776,
+      "loss": 0.1411,
+      "step": 7362
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00020266836734693876,
+      "loss": 0.0893,
+      "step": 7363
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00020265306122448976,
+      "loss": 0.1529,
+      "step": 7364
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.0002026377551020408,
+      "loss": 0.1583,
+      "step": 7365
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00020262244897959183,
+      "loss": 0.1974,
+      "step": 7366
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00020260714285714286,
+      "loss": 0.1361,
+      "step": 7367
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00020259183673469386,
+      "loss": 0.1058,
+      "step": 7368
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00020257653061224486,
+      "loss": 0.1132,
+      "step": 7369
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00020256122448979592,
+      "loss": 0.1311,
+      "step": 7370
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00020254591836734692,
+      "loss": 0.1169,
+      "step": 7371
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00020253061224489793,
+      "loss": 0.1814,
+      "step": 7372
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00020251530612244896,
+      "loss": 0.1507,
+      "step": 7373
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.0002025,
+      "loss": 0.1679,
+      "step": 7374
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00020248469387755102,
+      "loss": 0.1222,
+      "step": 7375
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00020246938775510202,
+      "loss": 0.2062,
+      "step": 7376
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00020245408163265303,
+      "loss": 0.0784,
+      "step": 7377
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00020243877551020408,
+      "loss": 0.121,
+      "step": 7378
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.0002024234693877551,
+      "loss": 0.1673,
+      "step": 7379
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0002024081632653061,
+      "loss": 0.1624,
+      "step": 7380
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020239285714285712,
+      "loss": 0.1053,
+      "step": 7381
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020237755102040815,
+      "loss": 0.1175,
+      "step": 7382
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020236224489795918,
+      "loss": 0.1943,
+      "step": 7383
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020234693877551019,
+      "loss": 0.1166,
+      "step": 7384
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0002023316326530612,
+      "loss": 0.2475,
+      "step": 7385
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020231632653061225,
+      "loss": 0.1427,
+      "step": 7386
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020230102040816325,
+      "loss": 0.2255,
+      "step": 7387
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020228571428571425,
+      "loss": 0.1898,
+      "step": 7388
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020227040816326528,
+      "loss": 0.1292,
+      "step": 7389
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00020225510204081632,
+      "loss": 0.0974,
+      "step": 7390
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00020223979591836732,
+      "loss": 0.1411,
+      "step": 7391
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00020222448979591835,
+      "loss": 0.3067,
+      "step": 7392
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00020220918367346935,
+      "loss": 0.1687,
+      "step": 7393
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.0002021938775510204,
+      "loss": 0.1368,
+      "step": 7394
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00020217857142857141,
+      "loss": 0.1759,
+      "step": 7395
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00020216326530612242,
+      "loss": 0.1709,
+      "step": 7396
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00020214795918367345,
+      "loss": 0.0977,
+      "step": 7397
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00020213265306122448,
+      "loss": 0.1492,
+      "step": 7398
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00020211734693877548,
+      "loss": 0.1199,
+      "step": 7399
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.0002021020408163265,
+      "loss": 0.1635,
+      "step": 7400
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020208673469387752,
+      "loss": 0.1783,
+      "step": 7401
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020207142857142857,
+      "loss": 0.0787,
+      "step": 7402
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020205612244897958,
+      "loss": 0.1085,
+      "step": 7403
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020204081632653058,
+      "loss": 0.2034,
+      "step": 7404
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020202551020408164,
+      "loss": 0.1437,
+      "step": 7405
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020201020408163264,
+      "loss": 0.2631,
+      "step": 7406
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020199489795918364,
+      "loss": 0.1422,
+      "step": 7407
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020197959183673468,
+      "loss": 0.1389,
+      "step": 7408
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.0002019642857142857,
+      "loss": 0.165,
+      "step": 7409
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00020194897959183674,
+      "loss": 0.275,
+      "step": 7410
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00020193367346938774,
+      "loss": 0.1876,
+      "step": 7411
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00020191836734693874,
+      "loss": 0.1575,
+      "step": 7412
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0002019030612244898,
+      "loss": 0.1642,
+      "step": 7413
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0002018877551020408,
+      "loss": 0.0911,
+      "step": 7414
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0002018724489795918,
+      "loss": 0.1335,
+      "step": 7415
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00020185714285714284,
+      "loss": 0.1551,
+      "step": 7416
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00020184183673469387,
+      "loss": 0.2294,
+      "step": 7417
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00020182653061224487,
+      "loss": 0.1272,
+      "step": 7418
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0002018112244897959,
+      "loss": 0.1228,
+      "step": 7419
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0002017959183673469,
+      "loss": 0.1198,
+      "step": 7420
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00020178061224489796,
+      "loss": 0.1244,
+      "step": 7421
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020176530612244897,
+      "loss": 0.1105,
+      "step": 7422
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020174999999999997,
+      "loss": 0.1518,
+      "step": 7423
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.000201734693877551,
+      "loss": 0.1089,
+      "step": 7424
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020171938775510203,
+      "loss": 0.0793,
+      "step": 7425
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020170408163265304,
+      "loss": 0.0745,
+      "step": 7426
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020168877551020407,
+      "loss": 0.0856,
+      "step": 7427
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020167346938775507,
+      "loss": 0.081,
+      "step": 7428
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020165816326530613,
+      "loss": 0.2092,
+      "step": 7429
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020164285714285713,
+      "loss": 0.1823,
+      "step": 7430
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00020162755102040813,
+      "loss": 0.1749,
+      "step": 7431
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.00020161224489795916,
+      "loss": 0.1193,
+      "step": 7432
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0002015969387755102,
+      "loss": 0.1035,
+      "step": 7433
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0002015816326530612,
+      "loss": 0.1084,
+      "step": 7434
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.00020156632653061223,
+      "loss": 0.1923,
+      "step": 7435
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.00020155102040816323,
+      "loss": 0.1731,
+      "step": 7436
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0002015357142857143,
+      "loss": 0.1058,
+      "step": 7437
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0002015204081632653,
+      "loss": 0.1223,
+      "step": 7438
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0002015051020408163,
+      "loss": 0.0684,
+      "step": 7439
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0002014897959183673,
+      "loss": 0.1499,
+      "step": 7440
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.00020147448979591836,
+      "loss": 0.0893,
+      "step": 7441
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020145918367346936,
+      "loss": 0.1961,
+      "step": 7442
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0002014438775510204,
+      "loss": 0.1071,
+      "step": 7443
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0002014285714285714,
+      "loss": 0.1447,
+      "step": 7444
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020141326530612245,
+      "loss": 0.0954,
+      "step": 7445
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020139795918367346,
+      "loss": 0.1425,
+      "step": 7446
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020138265306122446,
+      "loss": 0.1248,
+      "step": 7447
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020136734693877552,
+      "loss": 0.124,
+      "step": 7448
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020135204081632652,
+      "loss": 0.0801,
+      "step": 7449
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020133673469387752,
+      "loss": 0.0965,
+      "step": 7450
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020132142857142856,
+      "loss": 0.2198,
+      "step": 7451
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00020130612244897959,
+      "loss": 0.1411,
+      "step": 7452
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.0002012908163265306,
+      "loss": 0.1127,
+      "step": 7453
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00020127551020408162,
+      "loss": 0.2037,
+      "step": 7454
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00020126020408163262,
+      "loss": 0.263,
+      "step": 7455
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00020124489795918368,
+      "loss": 0.1782,
+      "step": 7456
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00020122959183673468,
+      "loss": 0.0936,
+      "step": 7457
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.0002012142857142857,
+      "loss": 0.0922,
+      "step": 7458
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00020119897959183672,
+      "loss": 0.1148,
+      "step": 7459
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00020118367346938775,
+      "loss": 0.1705,
+      "step": 7460
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00020116836734693875,
+      "loss": 0.1279,
+      "step": 7461
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00020115306122448978,
+      "loss": 0.1488,
+      "step": 7462
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00020113775510204079,
+      "loss": 0.0665,
+      "step": 7463
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00020112244897959184,
+      "loss": 0.1902,
+      "step": 7464
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00020110714285714285,
+      "loss": 0.0759,
+      "step": 7465
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00020109183673469385,
+      "loss": 0.1444,
+      "step": 7466
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00020107653061224488,
+      "loss": 0.3435,
+      "step": 7467
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.0002010612244897959,
+      "loss": 0.1477,
+      "step": 7468
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00020104591836734692,
+      "loss": 0.0768,
+      "step": 7469
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00020103061224489795,
+      "loss": 0.104,
+      "step": 7470
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00020101530612244895,
+      "loss": 0.0873,
+      "step": 7471
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.000201,
+      "loss": 0.0892,
+      "step": 7472
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.000200984693877551,
+      "loss": 0.3009,
+      "step": 7473
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00020096938775510201,
+      "loss": 0.1108,
+      "step": 7474
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00020095408163265302,
+      "loss": 0.1202,
+      "step": 7475
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00020093877551020407,
+      "loss": 0.0825,
+      "step": 7476
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00020092346938775508,
+      "loss": 0.0803,
+      "step": 7477
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.0002009081632653061,
+      "loss": 0.1368,
+      "step": 7478
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.0002008928571428571,
+      "loss": 0.1514,
+      "step": 7479
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00020087755102040814,
+      "loss": 0.1347,
+      "step": 7480
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00020086224489795917,
+      "loss": 0.138,
+      "step": 7481
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00020084693877551018,
+      "loss": 0.1257,
+      "step": 7482
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00020083163265306118,
+      "loss": 0.1704,
+      "step": 7483
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00020081632653061224,
+      "loss": 0.1723,
+      "step": 7484
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00020080102040816324,
+      "loss": 0.1603,
+      "step": 7485
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00020078571428571427,
+      "loss": 0.1934,
+      "step": 7486
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00020077040816326528,
+      "loss": 0.1086,
+      "step": 7487
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0002007551020408163,
+      "loss": 0.0967,
+      "step": 7488
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00020073979591836734,
+      "loss": 0.1457,
+      "step": 7489
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00020072448979591834,
+      "loss": 0.19,
+      "step": 7490
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00020070918367346934,
+      "loss": 0.2406,
+      "step": 7491
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0002006938775510204,
+      "loss": 0.2752,
+      "step": 7492
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0002006785714285714,
+      "loss": 0.2165,
+      "step": 7493
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00020066326530612243,
+      "loss": 0.1073,
+      "step": 7494
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00020064795918367347,
+      "loss": 0.1054,
+      "step": 7495
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00020063265306122447,
+      "loss": 0.0501,
+      "step": 7496
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0002006173469387755,
+      "loss": 0.1658,
+      "step": 7497
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0002006020408163265,
+      "loss": 0.2377,
+      "step": 7498
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00020058673469387756,
+      "loss": 0.1528,
+      "step": 7499
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00020057142857142856,
+      "loss": 0.1202,
+      "step": 7500
+    },
+    {
+      "epoch": 7.28,
+      "eval_cer": 0.26193599034972664,
+      "eval_loss": 0.964619517326355,
+      "eval_runtime": 250.6269,
+      "eval_samples_per_second": 35.168,
+      "eval_steps_per_second": 4.397,
+      "eval_wer": 0.6748508717002274,
+      "step": 7500
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00020055612244897957,
+      "loss": 0.147,
+      "step": 7501
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0002005408163265306,
+      "loss": 0.1767,
+      "step": 7502
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00020052551020408163,
+      "loss": 0.1342,
+      "step": 7503
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00020051020408163263,
+      "loss": 0.139,
+      "step": 7504
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00020049489795918366,
+      "loss": 0.2645,
+      "step": 7505
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00020047959183673467,
+      "loss": 0.0988,
+      "step": 7506
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00020046428571428572,
+      "loss": 0.1979,
+      "step": 7507
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00020044897959183673,
+      "loss": 0.0833,
+      "step": 7508
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00020043367346938773,
+      "loss": 0.1151,
+      "step": 7509
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00020041836734693873,
+      "loss": 0.1138,
+      "step": 7510
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.0002004030612244898,
+      "loss": 0.2403,
+      "step": 7511
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.0002003877551020408,
+      "loss": 0.2175,
+      "step": 7512
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00020037244897959183,
+      "loss": 0.1271,
+      "step": 7513
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00020035714285714283,
+      "loss": 0.0804,
+      "step": 7514
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00020034183673469386,
+      "loss": 0.2043,
+      "step": 7515
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.0002003265306122449,
+      "loss": 0.1002,
+      "step": 7516
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.0002003112244897959,
+      "loss": 0.1562,
+      "step": 7517
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.0002002959183673469,
+      "loss": 0.0868,
+      "step": 7518
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00020028061224489795,
+      "loss": 0.0944,
+      "step": 7519
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00020026530612244896,
+      "loss": 0.1547,
+      "step": 7520
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00020025,
+      "loss": 0.1042,
+      "step": 7521
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.000200234693877551,
+      "loss": 0.1365,
+      "step": 7522
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00020021938775510202,
+      "loss": 0.1805,
+      "step": 7523
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00020020408163265305,
+      "loss": 0.131,
+      "step": 7524
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020018877551020406,
+      "loss": 0.1761,
+      "step": 7525
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020017346938775506,
+      "loss": 0.2127,
+      "step": 7526
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020015816326530612,
+      "loss": 0.0915,
+      "step": 7527
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020014285714285712,
+      "loss": 0.1175,
+      "step": 7528
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020012755102040815,
+      "loss": 0.0813,
+      "step": 7529
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020011224489795916,
+      "loss": 0.1434,
+      "step": 7530
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020009693877551019,
+      "loss": 0.1422,
+      "step": 7531
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020008163265306122,
+      "loss": 0.213,
+      "step": 7532
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020006632653061222,
+      "loss": 0.0788,
+      "step": 7533
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00020005102040816322,
+      "loss": 0.1543,
+      "step": 7534
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00020003571428571428,
+      "loss": 0.235,
+      "step": 7535
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00020002040816326528,
+      "loss": 0.2194,
+      "step": 7536
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.0002000051020408163,
+      "loss": 0.1478,
+      "step": 7537
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00019998979591836735,
+      "loss": 0.1307,
+      "step": 7538
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00019997448979591835,
+      "loss": 0.0925,
+      "step": 7539
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00019995918367346938,
+      "loss": 0.1511,
+      "step": 7540
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00019994387755102038,
+      "loss": 0.1255,
+      "step": 7541
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00019992857142857144,
+      "loss": 0.1948,
+      "step": 7542
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00019991326530612244,
+      "loss": 0.2208,
+      "step": 7543
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00019989795918367345,
+      "loss": 0.2194,
+      "step": 7544
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00019988265306122445,
+      "loss": 0.179,
+      "step": 7545
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0001998673469387755,
+      "loss": 0.1271,
+      "step": 7546
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0001998520408163265,
+      "loss": 0.1085,
+      "step": 7547
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00019983673469387754,
+      "loss": 0.2198,
+      "step": 7548
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00019982142857142855,
+      "loss": 0.0957,
+      "step": 7549
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00019980612244897958,
+      "loss": 0.1537,
+      "step": 7550
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0001997908163265306,
+      "loss": 0.0754,
+      "step": 7551
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0001997755102040816,
+      "loss": 0.1201,
+      "step": 7552
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00019976020408163261,
+      "loss": 0.0851,
+      "step": 7553
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00019974489795918367,
+      "loss": 0.1537,
+      "step": 7554
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00019972959183673467,
+      "loss": 0.1915,
+      "step": 7555
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.0001997142857142857,
+      "loss": 0.1107,
+      "step": 7556
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.0001996989795918367,
+      "loss": 0.1566,
+      "step": 7557
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00019968367346938774,
+      "loss": 0.0842,
+      "step": 7558
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00019966836734693877,
+      "loss": 0.0928,
+      "step": 7559
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00019965306122448977,
+      "loss": 0.2029,
+      "step": 7560
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00019963775510204078,
+      "loss": 0.1455,
+      "step": 7561
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00019962244897959183,
+      "loss": 0.1778,
+      "step": 7562
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00019960714285714284,
+      "loss": 0.084,
+      "step": 7563
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00019959183673469387,
+      "loss": 0.1754,
+      "step": 7564
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00019957653061224487,
+      "loss": 0.1325,
+      "step": 7565
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.0001995612244897959,
+      "loss": 0.0845,
+      "step": 7566
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00019954591836734693,
+      "loss": 0.1454,
+      "step": 7567
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00019953061224489794,
+      "loss": 0.1451,
+      "step": 7568
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00019951530612244894,
+      "loss": 0.2263,
+      "step": 7569
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.0001995,
+      "loss": 0.0987,
+      "step": 7570
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.000199484693877551,
+      "loss": 0.0882,
+      "step": 7571
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.000199469387755102,
+      "loss": 0.0655,
+      "step": 7572
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00019945408163265303,
+      "loss": 0.1422,
+      "step": 7573
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00019943877551020407,
+      "loss": 0.15,
+      "step": 7574
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.0001994234693877551,
+      "loss": 0.1111,
+      "step": 7575
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.0001994081632653061,
+      "loss": 0.1858,
+      "step": 7576
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.0001993928571428571,
+      "loss": 0.0752,
+      "step": 7577
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00019937755102040816,
+      "loss": 0.1419,
+      "step": 7578
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00019936224489795916,
+      "loss": 0.1953,
+      "step": 7579
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00019934693877551017,
+      "loss": 0.1124,
+      "step": 7580
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.0001993316326530612,
+      "loss": 0.2397,
+      "step": 7581
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00019931632653061223,
+      "loss": 0.0979,
+      "step": 7582
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00019930102040816326,
+      "loss": 0.183,
+      "step": 7583
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00019928571428571426,
+      "loss": 0.0758,
+      "step": 7584
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.0001992704081632653,
+      "loss": 0.2071,
+      "step": 7585
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00019925510204081632,
+      "loss": 0.1528,
+      "step": 7586
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00019923979591836733,
+      "loss": 0.11,
+      "step": 7587
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00019922448979591833,
+      "loss": 0.0995,
+      "step": 7588
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0001992091836734694,
+      "loss": 0.1201,
+      "step": 7589
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0001991938775510204,
+      "loss": 0.2846,
+      "step": 7590
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00019917857142857142,
+      "loss": 0.2177,
+      "step": 7591
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00019916326530612243,
+      "loss": 0.2279,
+      "step": 7592
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00019914795918367346,
+      "loss": 0.2054,
+      "step": 7593
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0001991326530612245,
+      "loss": 0.0888,
+      "step": 7594
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0001991173469387755,
+      "loss": 0.1216,
+      "step": 7595
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0001991020408163265,
+      "loss": 0.1064,
+      "step": 7596
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019908673469387755,
+      "loss": 0.0881,
+      "step": 7597
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019907142857142855,
+      "loss": 0.2769,
+      "step": 7598
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019905612244897956,
+      "loss": 0.0952,
+      "step": 7599
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.0001990408163265306,
+      "loss": 0.1633,
+      "step": 7600
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019902551020408162,
+      "loss": 0.1346,
+      "step": 7601
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019901020408163265,
+      "loss": 0.1092,
+      "step": 7602
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019899489795918365,
+      "loss": 0.094,
+      "step": 7603
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019897959183673466,
+      "loss": 0.1615,
+      "step": 7604
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019896428571428571,
+      "loss": 0.1746,
+      "step": 7605
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019894897959183672,
+      "loss": 0.1114,
+      "step": 7606
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019893367346938772,
+      "loss": 0.1144,
+      "step": 7607
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019891836734693875,
+      "loss": 0.1536,
+      "step": 7608
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019890306122448978,
+      "loss": 0.1249,
+      "step": 7609
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.0001988877551020408,
+      "loss": 0.1238,
+      "step": 7610
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019887244897959182,
+      "loss": 0.1705,
+      "step": 7611
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019885714285714282,
+      "loss": 0.2113,
+      "step": 7612
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019884183673469388,
+      "loss": 0.1434,
+      "step": 7613
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019882653061224488,
+      "loss": 0.1031,
+      "step": 7614
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019881122448979588,
+      "loss": 0.1149,
+      "step": 7615
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00019879591836734691,
+      "loss": 0.1333,
+      "step": 7616
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019878061224489795,
+      "loss": 0.2854,
+      "step": 7617
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019876530612244898,
+      "loss": 0.1595,
+      "step": 7618
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019874999999999998,
+      "loss": 0.093,
+      "step": 7619
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019873469387755098,
+      "loss": 0.0918,
+      "step": 7620
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019871938775510204,
+      "loss": 0.1737,
+      "step": 7621
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019870408163265304,
+      "loss": 0.121,
+      "step": 7622
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019868877551020405,
+      "loss": 0.2587,
+      "step": 7623
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019867346938775508,
+      "loss": 0.2361,
+      "step": 7624
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.0001986581632653061,
+      "loss": 0.1328,
+      "step": 7625
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019864285714285714,
+      "loss": 0.127,
+      "step": 7626
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019862755102040814,
+      "loss": 0.082,
+      "step": 7627
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00019861224489795915,
+      "loss": 0.076,
+      "step": 7628
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.0001985969387755102,
+      "loss": 0.235,
+      "step": 7629
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.0001985816326530612,
+      "loss": 0.1902,
+      "step": 7630
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.0001985663265306122,
+      "loss": 0.1169,
+      "step": 7631
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00019855102040816327,
+      "loss": 0.1343,
+      "step": 7632
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00019853571428571427,
+      "loss": 0.101,
+      "step": 7633
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00019852040816326527,
+      "loss": 0.1291,
+      "step": 7634
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.0001985051020408163,
+      "loss": 0.1776,
+      "step": 7635
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00019848979591836734,
+      "loss": 0.1769,
+      "step": 7636
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00019847448979591837,
+      "loss": 0.1096,
+      "step": 7637
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019845918367346937,
+      "loss": 0.1827,
+      "step": 7638
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019844387755102037,
+      "loss": 0.0994,
+      "step": 7639
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019842857142857143,
+      "loss": 0.1341,
+      "step": 7640
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019841326530612243,
+      "loss": 0.1192,
+      "step": 7641
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019839795918367344,
+      "loss": 0.2136,
+      "step": 7642
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019838265306122447,
+      "loss": 0.1562,
+      "step": 7643
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.0001983673469387755,
+      "loss": 0.1732,
+      "step": 7644
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019835204081632653,
+      "loss": 0.174,
+      "step": 7645
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019833673469387753,
+      "loss": 0.1106,
+      "step": 7646
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019832142857142854,
+      "loss": 0.0947,
+      "step": 7647
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0001983061224489796,
+      "loss": 0.1126,
+      "step": 7648
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0001982908163265306,
+      "loss": 0.1482,
+      "step": 7649
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0001982755102040816,
+      "loss": 0.1529,
+      "step": 7650
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.00019826020408163263,
+      "loss": 0.1068,
+      "step": 7651
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.00019824489795918366,
+      "loss": 0.1839,
+      "step": 7652
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0001982295918367347,
+      "loss": 0.1586,
+      "step": 7653
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0001982142857142857,
+      "loss": 0.1493,
+      "step": 7654
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0001981989795918367,
+      "loss": 0.1915,
+      "step": 7655
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.00019818367346938776,
+      "loss": 0.1706,
+      "step": 7656
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.00019816836734693876,
+      "loss": 0.1006,
+      "step": 7657
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.00019815306122448976,
+      "loss": 0.0893,
+      "step": 7658
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.0001981377551020408,
+      "loss": 0.1979,
+      "step": 7659
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019812244897959183,
+      "loss": 0.1383,
+      "step": 7660
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019810714285714286,
+      "loss": 0.2341,
+      "step": 7661
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019809183673469386,
+      "loss": 0.2005,
+      "step": 7662
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019807653061224486,
+      "loss": 0.1779,
+      "step": 7663
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019806122448979592,
+      "loss": 0.1765,
+      "step": 7664
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019804591836734692,
+      "loss": 0.1899,
+      "step": 7665
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019803061224489793,
+      "loss": 0.1893,
+      "step": 7666
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019801530612244896,
+      "loss": 0.2247,
+      "step": 7667
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.000198,
+      "loss": 0.1963,
+      "step": 7668
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.000197984693877551,
+      "loss": 0.1702,
+      "step": 7669
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00019796938775510202,
+      "loss": 0.1296,
+      "step": 7670
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00019795408163265303,
+      "loss": 0.1165,
+      "step": 7671
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00019793877551020408,
+      "loss": 0.1251,
+      "step": 7672
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.0001979234693877551,
+      "loss": 0.1525,
+      "step": 7673
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.0001979081632653061,
+      "loss": 0.1617,
+      "step": 7674
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00019789285714285715,
+      "loss": 0.0902,
+      "step": 7675
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00019787755102040815,
+      "loss": 0.2325,
+      "step": 7676
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00019786224489795915,
+      "loss": 0.1591,
+      "step": 7677
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00019784693877551019,
+      "loss": 0.1709,
+      "step": 7678
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00019783163265306122,
+      "loss": 0.1539,
+      "step": 7679
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00019781632653061225,
+      "loss": 0.1312,
+      "step": 7680
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00019780102040816325,
+      "loss": 0.1564,
+      "step": 7681
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00019778571428571425,
+      "loss": 0.1721,
+      "step": 7682
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.0001977704081632653,
+      "loss": 0.1069,
+      "step": 7683
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00019775510204081631,
+      "loss": 0.1589,
+      "step": 7684
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00019773979591836732,
+      "loss": 0.2397,
+      "step": 7685
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00019772448979591835,
+      "loss": 0.203,
+      "step": 7686
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00019770918367346938,
+      "loss": 0.1955,
+      "step": 7687
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.0001976938775510204,
+      "loss": 0.117,
+      "step": 7688
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.0001976785714285714,
+      "loss": 0.0818,
+      "step": 7689
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019766326530612242,
+      "loss": 0.1409,
+      "step": 7690
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019764795918367347,
+      "loss": 0.0639,
+      "step": 7691
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019763265306122448,
+      "loss": 0.1828,
+      "step": 7692
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019761734693877548,
+      "loss": 0.1568,
+      "step": 7693
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.0001976020408163265,
+      "loss": 0.1855,
+      "step": 7694
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019758673469387754,
+      "loss": 0.115,
+      "step": 7695
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019757142857142855,
+      "loss": 0.1107,
+      "step": 7696
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019755612244897958,
+      "loss": 0.1212,
+      "step": 7697
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019754081632653058,
+      "loss": 0.2463,
+      "step": 7698
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019752551020408164,
+      "loss": 0.155,
+      "step": 7699
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00019751020408163264,
+      "loss": 0.1208,
+      "step": 7700
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00019749489795918364,
+      "loss": 0.0879,
+      "step": 7701
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00019747959183673467,
+      "loss": 0.168,
+      "step": 7702
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.0001974642857142857,
+      "loss": 0.1242,
+      "step": 7703
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.0001974489795918367,
+      "loss": 0.2742,
+      "step": 7704
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00019743367346938774,
+      "loss": 0.3187,
+      "step": 7705
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00019741836734693874,
+      "loss": 0.1066,
+      "step": 7706
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.0001974030612244898,
+      "loss": 0.1614,
+      "step": 7707
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.0001973877551020408,
+      "loss": 0.1764,
+      "step": 7708
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.0001973724489795918,
+      "loss": 0.0852,
+      "step": 7709
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00019735714285714284,
+      "loss": 0.1081,
+      "step": 7710
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00019734183673469387,
+      "loss": 0.2742,
+      "step": 7711
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00019732653061224487,
+      "loss": 0.1152,
+      "step": 7712
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.0001973112244897959,
+      "loss": 0.1591,
+      "step": 7713
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.0001972959183673469,
+      "loss": 0.2207,
+      "step": 7714
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00019728061224489796,
+      "loss": 0.1363,
+      "step": 7715
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00019726530612244897,
+      "loss": 0.1749,
+      "step": 7716
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00019724999999999997,
+      "loss": 0.1581,
+      "step": 7717
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00019723469387755097,
+      "loss": 0.1805,
+      "step": 7718
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00019721938775510203,
+      "loss": 0.09,
+      "step": 7719
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00019720408163265303,
+      "loss": 0.1775,
+      "step": 7720
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00019718877551020406,
+      "loss": 0.0955,
+      "step": 7721
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0001971734693877551,
+      "loss": 0.1851,
+      "step": 7722
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00019715816326530613,
+      "loss": 0.2186,
+      "step": 7723
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00019714285714285713,
+      "loss": 0.0942,
+      "step": 7724
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00019712755102040813,
+      "loss": 0.1501,
+      "step": 7725
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0001971122448979592,
+      "loss": 0.1107,
+      "step": 7726
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0001970969387755102,
+      "loss": 0.0753,
+      "step": 7727
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0001970816326530612,
+      "loss": 0.0756,
+      "step": 7728
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00019706632653061223,
+      "loss": 0.3081,
+      "step": 7729
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00019705102040816326,
+      "loss": 0.1172,
+      "step": 7730
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00019703571428571426,
+      "loss": 0.1198,
+      "step": 7731
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.0001970204081632653,
+      "loss": 0.1947,
+      "step": 7732
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.0001970051020408163,
+      "loss": 0.1007,
+      "step": 7733
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00019698979591836735,
+      "loss": 0.0832,
+      "step": 7734
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00019697448979591836,
+      "loss": 0.1627,
+      "step": 7735
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00019695918367346936,
+      "loss": 0.286,
+      "step": 7736
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.0001969438775510204,
+      "loss": 0.1709,
+      "step": 7737
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00019692857142857142,
+      "loss": 0.1343,
+      "step": 7738
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00019691326530612243,
+      "loss": 0.0845,
+      "step": 7739
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00019689795918367346,
+      "loss": 0.1096,
+      "step": 7740
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019688265306122446,
+      "loss": 0.1202,
+      "step": 7741
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019686734693877552,
+      "loss": 0.2646,
+      "step": 7742
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019685204081632652,
+      "loss": 0.1046,
+      "step": 7743
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019683673469387752,
+      "loss": 0.1336,
+      "step": 7744
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019682142857142855,
+      "loss": 0.156,
+      "step": 7745
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019680612244897958,
+      "loss": 0.0896,
+      "step": 7746
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.0001967908163265306,
+      "loss": 0.1188,
+      "step": 7747
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019677551020408162,
+      "loss": 0.2366,
+      "step": 7748
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019676020408163262,
+      "loss": 0.1198,
+      "step": 7749
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00019674489795918368,
+      "loss": 0.3039,
+      "step": 7750
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019672959183673468,
+      "loss": 0.1494,
+      "step": 7751
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.0001967142857142857,
+      "loss": 0.1254,
+      "step": 7752
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.0001966989795918367,
+      "loss": 0.0849,
+      "step": 7753
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019668367346938775,
+      "loss": 0.2684,
+      "step": 7754
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019666836734693875,
+      "loss": 0.1958,
+      "step": 7755
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019665306122448978,
+      "loss": 0.15,
+      "step": 7756
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019663775510204079,
+      "loss": 0.128,
+      "step": 7757
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019662244897959182,
+      "loss": 0.1243,
+      "step": 7758
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019660714285714285,
+      "loss": 0.1684,
+      "step": 7759
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019659183673469385,
+      "loss": 0.1863,
+      "step": 7760
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019657653061224485,
+      "loss": 0.166,
+      "step": 7761
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.0001965612244897959,
+      "loss": 0.1431,
+      "step": 7762
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00019654591836734691,
+      "loss": 0.1346,
+      "step": 7763
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00019653061224489794,
+      "loss": 0.1015,
+      "step": 7764
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00019651530612244898,
+      "loss": 0.0766,
+      "step": 7765
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00019649999999999998,
+      "loss": 0.1551,
+      "step": 7766
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.000196484693877551,
+      "loss": 0.3549,
+      "step": 7767
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.000196469387755102,
+      "loss": 0.1881,
+      "step": 7768
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00019645408163265307,
+      "loss": 0.107,
+      "step": 7769
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00019643877551020407,
+      "loss": 0.1785,
+      "step": 7770
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00019642346938775508,
+      "loss": 0.1644,
+      "step": 7771
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.0001964081632653061,
+      "loss": 0.0734,
+      "step": 7772
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019639285714285714,
+      "loss": 0.2043,
+      "step": 7773
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019637755102040814,
+      "loss": 0.1902,
+      "step": 7774
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019636224489795917,
+      "loss": 0.208,
+      "step": 7775
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019634693877551018,
+      "loss": 0.2094,
+      "step": 7776
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019633163265306123,
+      "loss": 0.1747,
+      "step": 7777
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019631632653061224,
+      "loss": 0.0662,
+      "step": 7778
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019630102040816324,
+      "loss": 0.1828,
+      "step": 7779
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019628571428571427,
+      "loss": 0.1211,
+      "step": 7780
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.0001962704081632653,
+      "loss": 0.1688,
+      "step": 7781
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0001962551020408163,
+      "loss": 0.0959,
+      "step": 7782
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.00019623979591836734,
+      "loss": 0.1679,
+      "step": 7783
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.00019622448979591834,
+      "loss": 0.0608,
+      "step": 7784
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0001962091836734694,
+      "loss": 0.1352,
+      "step": 7785
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0001961938775510204,
+      "loss": 0.2644,
+      "step": 7786
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0001961785714285714,
+      "loss": 0.1607,
+      "step": 7787
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0001961632653061224,
+      "loss": 0.1773,
+      "step": 7788
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.00019614795918367346,
+      "loss": 0.1042,
+      "step": 7789
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.00019613265306122447,
+      "loss": 0.1747,
+      "step": 7790
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0001961173469387755,
+      "loss": 0.0782,
+      "step": 7791
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.0001961020408163265,
+      "loss": 0.1886,
+      "step": 7792
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019608673469387753,
+      "loss": 0.1678,
+      "step": 7793
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019607142857142856,
+      "loss": 0.1399,
+      "step": 7794
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019605612244897957,
+      "loss": 0.1488,
+      "step": 7795
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019604081632653057,
+      "loss": 0.1344,
+      "step": 7796
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019602551020408163,
+      "loss": 0.1691,
+      "step": 7797
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019601020408163263,
+      "loss": 0.2899,
+      "step": 7798
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019599489795918366,
+      "loss": 0.1444,
+      "step": 7799
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019597959183673466,
+      "loss": 0.1011,
+      "step": 7800
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.0001959642857142857,
+      "loss": 0.0948,
+      "step": 7801
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00019594897959183673,
+      "loss": 0.1318,
+      "step": 7802
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00019593367346938773,
+      "loss": 0.1597,
+      "step": 7803
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00019591836734693873,
+      "loss": 0.2692,
+      "step": 7804
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.0001959030612244898,
+      "loss": 0.1736,
+      "step": 7805
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.0001958877551020408,
+      "loss": 0.1844,
+      "step": 7806
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00019587244897959182,
+      "loss": 0.173,
+      "step": 7807
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00019585714285714283,
+      "loss": 0.1636,
+      "step": 7808
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00019584183673469386,
+      "loss": 0.1284,
+      "step": 7809
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.0001958265306122449,
+      "loss": 0.1972,
+      "step": 7810
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.0001958112244897959,
+      "loss": 0.2133,
+      "step": 7811
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00019579591836734695,
+      "loss": 0.1307,
+      "step": 7812
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019578061224489795,
+      "loss": 0.1529,
+      "step": 7813
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019576530612244896,
+      "loss": 0.0952,
+      "step": 7814
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019574999999999996,
+      "loss": 0.1162,
+      "step": 7815
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019573469387755102,
+      "loss": 0.1534,
+      "step": 7816
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019571938775510202,
+      "loss": 0.1705,
+      "step": 7817
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019570408163265305,
+      "loss": 0.1698,
+      "step": 7818
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019568877551020406,
+      "loss": 0.1608,
+      "step": 7819
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019567346938775509,
+      "loss": 0.132,
+      "step": 7820
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019565816326530612,
+      "loss": 0.1292,
+      "step": 7821
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019564285714285712,
+      "loss": 0.103,
+      "step": 7822
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019562755102040812,
+      "loss": 0.0909,
+      "step": 7823
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019561224489795918,
+      "loss": 0.1174,
+      "step": 7824
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019559693877551018,
+      "loss": 0.1222,
+      "step": 7825
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019558163265306122,
+      "loss": 0.1526,
+      "step": 7826
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019556632653061222,
+      "loss": 0.1859,
+      "step": 7827
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019555102040816325,
+      "loss": 0.1214,
+      "step": 7828
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019553571428571428,
+      "loss": 0.2026,
+      "step": 7829
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019552040816326528,
+      "loss": 0.165,
+      "step": 7830
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.0001955051020408163,
+      "loss": 0.1706,
+      "step": 7831
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019548979591836734,
+      "loss": 0.1582,
+      "step": 7832
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019547448979591835,
+      "loss": 0.0858,
+      "step": 7833
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00019545918367346938,
+      "loss": 0.0811,
+      "step": 7834
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00019544387755102038,
+      "loss": 0.2355,
+      "step": 7835
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.0001954285714285714,
+      "loss": 0.1891,
+      "step": 7836
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00019541326530612244,
+      "loss": 0.1047,
+      "step": 7837
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00019539795918367345,
+      "loss": 0.142,
+      "step": 7838
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00019538265306122445,
+      "loss": 0.1171,
+      "step": 7839
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.0001953673469387755,
+      "loss": 0.1007,
+      "step": 7840
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.0001953520408163265,
+      "loss": 0.0984,
+      "step": 7841
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00019533673469387754,
+      "loss": 0.2133,
+      "step": 7842
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00019532142857142854,
+      "loss": 0.1212,
+      "step": 7843
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00019530612244897958,
+      "loss": 0.1059,
+      "step": 7844
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.0001952908163265306,
+      "loss": 0.1072,
+      "step": 7845
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.0001952755102040816,
+      "loss": 0.0796,
+      "step": 7846
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.0001952602040816326,
+      "loss": 0.1565,
+      "step": 7847
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00019524489795918367,
+      "loss": 0.157,
+      "step": 7848
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00019522959183673467,
+      "loss": 0.1507,
+      "step": 7849
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00019521428571428568,
+      "loss": 0.1891,
+      "step": 7850
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.0001951989795918367,
+      "loss": 0.1038,
+      "step": 7851
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00019518367346938774,
+      "loss": 0.0849,
+      "step": 7852
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00019516836734693877,
+      "loss": 0.1025,
+      "step": 7853
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00019515306122448977,
+      "loss": 0.1662,
+      "step": 7854
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0001951377551020408,
+      "loss": 0.1856,
+      "step": 7855
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00019512244897959183,
+      "loss": 0.147,
+      "step": 7856
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00019510714285714284,
+      "loss": 0.1197,
+      "step": 7857
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00019509183673469384,
+      "loss": 0.0783,
+      "step": 7858
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0001950765306122449,
+      "loss": 0.124,
+      "step": 7859
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0001950612244897959,
+      "loss": 0.168,
+      "step": 7860
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00019504591836734693,
+      "loss": 0.2825,
+      "step": 7861
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00019503061224489794,
+      "loss": 0.2765,
+      "step": 7862
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00019501530612244897,
+      "loss": 0.1141,
+      "step": 7863
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.000195,
+      "loss": 0.1664,
+      "step": 7864
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.000194984693877551,
+      "loss": 0.1317,
+      "step": 7865
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.000194969387755102,
+      "loss": 0.266,
+      "step": 7866
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00019495408163265306,
+      "loss": 0.1771,
+      "step": 7867
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00019493877551020406,
+      "loss": 0.251,
+      "step": 7868
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.0001949234693877551,
+      "loss": 0.1124,
+      "step": 7869
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.0001949081632653061,
+      "loss": 0.149,
+      "step": 7870
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00019489285714285713,
+      "loss": 0.1868,
+      "step": 7871
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00019487755102040816,
+      "loss": 0.1062,
+      "step": 7872
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00019486224489795916,
+      "loss": 0.2189,
+      "step": 7873
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00019484693877551017,
+      "loss": 0.2232,
+      "step": 7874
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00019483163265306122,
+      "loss": 0.1048,
+      "step": 7875
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00019481632653061223,
+      "loss": 0.1226,
+      "step": 7876
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00019480102040816323,
+      "loss": 0.132,
+      "step": 7877
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00019478571428571426,
+      "loss": 0.0928,
+      "step": 7878
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.0001947704081632653,
+      "loss": 0.1516,
+      "step": 7879
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00019475510204081632,
+      "loss": 0.1792,
+      "step": 7880
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00019473979591836733,
+      "loss": 0.087,
+      "step": 7881
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00019472448979591833,
+      "loss": 0.2144,
+      "step": 7882
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.0001947091836734694,
+      "loss": 0.0687,
+      "step": 7883
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.0001946938775510204,
+      "loss": 0.0785,
+      "step": 7884
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.0001946785714285714,
+      "loss": 0.2549,
+      "step": 7885
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00019466326530612242,
+      "loss": 0.1385,
+      "step": 7886
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00019464795918367346,
+      "loss": 0.1299,
+      "step": 7887
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00019463265306122449,
+      "loss": 0.1498,
+      "step": 7888
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.0001946173469387755,
+      "loss": 0.1307,
+      "step": 7889
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.0001946020408163265,
+      "loss": 0.074,
+      "step": 7890
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00019458673469387755,
+      "loss": 0.0998,
+      "step": 7891
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00019457142857142855,
+      "loss": 0.1562,
+      "step": 7892
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00019455612244897956,
+      "loss": 0.0905,
+      "step": 7893
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.0001945408163265306,
+      "loss": 0.0983,
+      "step": 7894
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019452551020408162,
+      "loss": 0.1728,
+      "step": 7895
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019451020408163265,
+      "loss": 0.1091,
+      "step": 7896
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019449489795918365,
+      "loss": 0.195,
+      "step": 7897
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019447959183673466,
+      "loss": 0.2143,
+      "step": 7898
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.0001944642857142857,
+      "loss": 0.1158,
+      "step": 7899
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019444897959183672,
+      "loss": 0.1245,
+      "step": 7900
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019443367346938772,
+      "loss": 0.1319,
+      "step": 7901
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019441836734693878,
+      "loss": 0.1356,
+      "step": 7902
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019440306122448978,
+      "loss": 0.0826,
+      "step": 7903
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.0001943877551020408,
+      "loss": 0.2446,
+      "step": 7904
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00019437244897959182,
+      "loss": 0.179,
+      "step": 7905
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019435714285714285,
+      "loss": 0.1783,
+      "step": 7906
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019434183673469388,
+      "loss": 0.0892,
+      "step": 7907
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019432653061224488,
+      "loss": 0.1856,
+      "step": 7908
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019431122448979588,
+      "loss": 0.0789,
+      "step": 7909
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019429591836734694,
+      "loss": 0.1527,
+      "step": 7910
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019428061224489794,
+      "loss": 0.1205,
+      "step": 7911
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019426530612244895,
+      "loss": 0.0927,
+      "step": 7912
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019424999999999998,
+      "loss": 0.2032,
+      "step": 7913
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.000194234693877551,
+      "loss": 0.0896,
+      "step": 7914
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00019421938775510204,
+      "loss": 0.0803,
+      "step": 7915
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.00019420408163265304,
+      "loss": 0.1305,
+      "step": 7916
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.00019418877551020405,
+      "loss": 0.1434,
+      "step": 7917
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0001941734693877551,
+      "loss": 0.234,
+      "step": 7918
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0001941581632653061,
+      "loss": 0.1376,
+      "step": 7919
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0001941428571428571,
+      "loss": 0.1875,
+      "step": 7920
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.00019412755102040814,
+      "loss": 0.1332,
+      "step": 7921
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.00019411224489795917,
+      "loss": 0.0965,
+      "step": 7922
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0001940969387755102,
+      "loss": 0.107,
+      "step": 7923
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0001940816326530612,
+      "loss": 0.1228,
+      "step": 7924
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0001940663265306122,
+      "loss": 0.1054,
+      "step": 7925
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019405102040816327,
+      "loss": 0.1526,
+      "step": 7926
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019403571428571427,
+      "loss": 0.0839,
+      "step": 7927
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019402040816326527,
+      "loss": 0.2509,
+      "step": 7928
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.0001940051020408163,
+      "loss": 0.2464,
+      "step": 7929
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019398979591836733,
+      "loss": 0.1204,
+      "step": 7930
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019397448979591837,
+      "loss": 0.1758,
+      "step": 7931
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019395918367346937,
+      "loss": 0.1453,
+      "step": 7932
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019394387755102037,
+      "loss": 0.0924,
+      "step": 7933
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019392857142857143,
+      "loss": 0.0777,
+      "step": 7934
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019391326530612243,
+      "loss": 0.0657,
+      "step": 7935
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00019389795918367344,
+      "loss": 0.1582,
+      "step": 7936
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.00019388265306122447,
+      "loss": 0.1729,
+      "step": 7937
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0001938673469387755,
+      "loss": 0.1301,
+      "step": 7938
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0001938520408163265,
+      "loss": 0.1751,
+      "step": 7939
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.00019383673469387753,
+      "loss": 0.0852,
+      "step": 7940
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.00019382142857142854,
+      "loss": 0.1046,
+      "step": 7941
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0001938061224489796,
+      "loss": 0.147,
+      "step": 7942
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0001937908163265306,
+      "loss": 0.1783,
+      "step": 7943
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0001937755102040816,
+      "loss": 0.1169,
+      "step": 7944
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.00019376020408163263,
+      "loss": 0.1304,
+      "step": 7945
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.00019374489795918366,
+      "loss": 0.1933,
+      "step": 7946
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019372959183673466,
+      "loss": 0.1288,
+      "step": 7947
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.0001937142857142857,
+      "loss": 0.1668,
+      "step": 7948
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019369897959183673,
+      "loss": 0.1905,
+      "step": 7949
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019368367346938776,
+      "loss": 0.1338,
+      "step": 7950
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019366836734693876,
+      "loss": 0.1341,
+      "step": 7951
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019365306122448976,
+      "loss": 0.2134,
+      "step": 7952
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019363775510204082,
+      "loss": 0.0676,
+      "step": 7953
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019362244897959182,
+      "loss": 0.2708,
+      "step": 7954
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019360714285714283,
+      "loss": 0.2023,
+      "step": 7955
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00019359183673469386,
+      "loss": 0.0989,
+      "step": 7956
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.0001935765306122449,
+      "loss": 0.1306,
+      "step": 7957
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00019356122448979592,
+      "loss": 0.0929,
+      "step": 7958
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00019354591836734692,
+      "loss": 0.1129,
+      "step": 7959
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00019353061224489793,
+      "loss": 0.1278,
+      "step": 7960
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00019351530612244898,
+      "loss": 0.1812,
+      "step": 7961
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.0001935,
+      "loss": 0.2696,
+      "step": 7962
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.000193484693877551,
+      "loss": 0.1131,
+      "step": 7963
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00019346938775510202,
+      "loss": 0.2054,
+      "step": 7964
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00019345408163265305,
+      "loss": 0.1156,
+      "step": 7965
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00019343877551020408,
+      "loss": 0.1171,
+      "step": 7966
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00019342346938775509,
+      "loss": 0.1412,
+      "step": 7967
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.0001934081632653061,
+      "loss": 0.2122,
+      "step": 7968
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00019339285714285715,
+      "loss": 0.159,
+      "step": 7969
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00019337755102040815,
+      "loss": 0.1118,
+      "step": 7970
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00019336224489795915,
+      "loss": 0.1073,
+      "step": 7971
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00019334693877551018,
+      "loss": 0.1138,
+      "step": 7972
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00019333163265306121,
+      "loss": 0.3126,
+      "step": 7973
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00019331632653061222,
+      "loss": 0.135,
+      "step": 7974
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00019330102040816325,
+      "loss": 0.1314,
+      "step": 7975
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00019328571428571425,
+      "loss": 0.1608,
+      "step": 7976
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.0001932704081632653,
+      "loss": 0.0998,
+      "step": 7977
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.0001932551020408163,
+      "loss": 0.1105,
+      "step": 7978
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00019323979591836732,
+      "loss": 0.1497,
+      "step": 7979
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00019322448979591835,
+      "loss": 0.1176,
+      "step": 7980
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00019320918367346938,
+      "loss": 0.1923,
+      "step": 7981
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00019319387755102038,
+      "loss": 0.1098,
+      "step": 7982
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.0001931785714285714,
+      "loss": 0.1375,
+      "step": 7983
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00019316326530612242,
+      "loss": 0.082,
+      "step": 7984
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00019314795918367347,
+      "loss": 0.2081,
+      "step": 7985
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00019313265306122448,
+      "loss": 0.2778,
+      "step": 7986
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00019311734693877548,
+      "loss": 0.1329,
+      "step": 7987
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0001931020408163265,
+      "loss": 0.132,
+      "step": 7988
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00019308673469387754,
+      "loss": 0.1934,
+      "step": 7989
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00019307142857142854,
+      "loss": 0.16,
+      "step": 7990
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00019305612244897957,
+      "loss": 0.1293,
+      "step": 7991
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0001930408163265306,
+      "loss": 0.1843,
+      "step": 7992
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00019302551020408164,
+      "loss": 0.1135,
+      "step": 7993
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00019301020408163264,
+      "loss": 0.1781,
+      "step": 7994
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00019299489795918364,
+      "loss": 0.1639,
+      "step": 7995
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0001929795918367347,
+      "loss": 0.0957,
+      "step": 7996
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0001929642857142857,
+      "loss": 0.099,
+      "step": 7997
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.0001929489795918367,
+      "loss": 0.2168,
+      "step": 7998
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00019293367346938774,
+      "loss": 0.1416,
+      "step": 7999
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00019291836734693877,
+      "loss": 0.143,
+      "step": 8000
+    },
+    {
+      "epoch": 7.77,
+      "eval_cer": 0.25539890494463363,
+      "eval_loss": 0.9199783802032471,
+      "eval_runtime": 234.4328,
+      "eval_samples_per_second": 37.597,
+      "eval_steps_per_second": 4.701,
+      "eval_wer": 0.6554120115567567,
+      "step": 8000
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.0001929030612244898,
+      "loss": 0.1224,
+      "step": 8001
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.0001928877551020408,
+      "loss": 0.1462,
+      "step": 8002
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.0001928724489795918,
+      "loss": 0.122,
+      "step": 8003
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00019285714285714286,
+      "loss": 0.1848,
+      "step": 8004
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00019284183673469387,
+      "loss": 0.2153,
+      "step": 8005
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00019282653061224487,
+      "loss": 0.16,
+      "step": 8006
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.0001928112244897959,
+      "loss": 0.1373,
+      "step": 8007
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00019279591836734693,
+      "loss": 0.0973,
+      "step": 8008
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00019278061224489793,
+      "loss": 0.1177,
+      "step": 8009
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00019276530612244897,
+      "loss": 0.196,
+      "step": 8010
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00019274999999999997,
+      "loss": 0.1728,
+      "step": 8011
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00019273469387755103,
+      "loss": 0.1265,
+      "step": 8012
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00019271938775510203,
+      "loss": 0.1498,
+      "step": 8013
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00019270408163265303,
+      "loss": 0.0873,
+      "step": 8014
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00019268877551020406,
+      "loss": 0.1018,
+      "step": 8015
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.0001926734693877551,
+      "loss": 0.0813,
+      "step": 8016
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.0001926581632653061,
+      "loss": 0.1519,
+      "step": 8017
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00019264285714285713,
+      "loss": 0.1857,
+      "step": 8018
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00019262755102040813,
+      "loss": 0.1991,
+      "step": 8019
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.0001926122448979592,
+      "loss": 0.0905,
+      "step": 8020
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.0001925969387755102,
+      "loss": 0.0839,
+      "step": 8021
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.0001925816326530612,
+      "loss": 0.1796,
+      "step": 8022
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00019256632653061223,
+      "loss": 0.2591,
+      "step": 8023
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00019255102040816326,
+      "loss": 0.1972,
+      "step": 8024
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00019253571428571426,
+      "loss": 0.2252,
+      "step": 8025
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.0001925204081632653,
+      "loss": 0.0992,
+      "step": 8026
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.0001925051020408163,
+      "loss": 0.1444,
+      "step": 8027
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00019248979591836735,
+      "loss": 0.1169,
+      "step": 8028
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019247448979591836,
+      "loss": 0.1906,
+      "step": 8029
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019245918367346936,
+      "loss": 0.1653,
+      "step": 8030
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019244387755102036,
+      "loss": 0.1278,
+      "step": 8031
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019242857142857142,
+      "loss": 0.0816,
+      "step": 8032
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019241326530612242,
+      "loss": 0.137,
+      "step": 8033
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019239795918367345,
+      "loss": 0.0923,
+      "step": 8034
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019238265306122446,
+      "loss": 0.1806,
+      "step": 8035
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.0001923673469387755,
+      "loss": 0.2015,
+      "step": 8036
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019235204081632652,
+      "loss": 0.0881,
+      "step": 8037
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019233673469387752,
+      "loss": 0.1434,
+      "step": 8038
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00019232142857142858,
+      "loss": 0.0876,
+      "step": 8039
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019230612244897958,
+      "loss": 0.0986,
+      "step": 8040
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.0001922908163265306,
+      "loss": 0.1314,
+      "step": 8041
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019227551020408162,
+      "loss": 0.2021,
+      "step": 8042
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019226020408163265,
+      "loss": 0.1227,
+      "step": 8043
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019224489795918365,
+      "loss": 0.1204,
+      "step": 8044
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019222959183673468,
+      "loss": 0.0957,
+      "step": 8045
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019221428571428569,
+      "loss": 0.1585,
+      "step": 8046
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019219897959183674,
+      "loss": 0.1173,
+      "step": 8047
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019218367346938775,
+      "loss": 0.2278,
+      "step": 8048
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00019216836734693875,
+      "loss": 0.126,
+      "step": 8049
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00019215306122448978,
+      "loss": 0.1407,
+      "step": 8050
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0001921377551020408,
+      "loss": 0.088,
+      "step": 8051
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00019212244897959181,
+      "loss": 0.1839,
+      "step": 8052
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00019210714285714285,
+      "loss": 0.1841,
+      "step": 8053
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00019209183673469385,
+      "loss": 0.0906,
+      "step": 8054
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0001920765306122449,
+      "loss": 0.2047,
+      "step": 8055
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0001920612244897959,
+      "loss": 0.1559,
+      "step": 8056
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0001920459183673469,
+      "loss": 0.1118,
+      "step": 8057
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00019203061224489792,
+      "loss": 0.0789,
+      "step": 8058
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00019201530612244897,
+      "loss": 0.0871,
+      "step": 8059
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019199999999999998,
+      "loss": 0.2679,
+      "step": 8060
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.000191984693877551,
+      "loss": 0.2077,
+      "step": 8061
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.000191969387755102,
+      "loss": 0.1528,
+      "step": 8062
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019195408163265307,
+      "loss": 0.1433,
+      "step": 8063
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019193877551020407,
+      "loss": 0.1029,
+      "step": 8064
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019192346938775508,
+      "loss": 0.1377,
+      "step": 8065
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019190816326530608,
+      "loss": 0.1228,
+      "step": 8066
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019189285714285714,
+      "loss": 0.3086,
+      "step": 8067
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019187755102040814,
+      "loss": 0.101,
+      "step": 8068
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019186224489795917,
+      "loss": 0.1407,
+      "step": 8069
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00019184693877551017,
+      "loss": 0.0795,
+      "step": 8070
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.0001918316326530612,
+      "loss": 0.0809,
+      "step": 8071
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00019181632653061224,
+      "loss": 0.1222,
+      "step": 8072
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00019180102040816324,
+      "loss": 0.1714,
+      "step": 8073
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00019178571428571424,
+      "loss": 0.1481,
+      "step": 8074
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.0001917704081632653,
+      "loss": 0.1091,
+      "step": 8075
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.0001917551020408163,
+      "loss": 0.119,
+      "step": 8076
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00019173979591836733,
+      "loss": 0.1542,
+      "step": 8077
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00019172448979591834,
+      "loss": 0.0929,
+      "step": 8078
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00019170918367346937,
+      "loss": 0.2397,
+      "step": 8079
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.0001916938775510204,
+      "loss": 0.1755,
+      "step": 8080
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.0001916785714285714,
+      "loss": 0.1145,
+      "step": 8081
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00019166326530612246,
+      "loss": 0.1619,
+      "step": 8082
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00019164795918367346,
+      "loss": 0.1579,
+      "step": 8083
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00019163265306122447,
+      "loss": 0.1047,
+      "step": 8084
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.0001916173469387755,
+      "loss": 0.1715,
+      "step": 8085
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00019160204081632653,
+      "loss": 0.1665,
+      "step": 8086
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00019158673469387753,
+      "loss": 0.1269,
+      "step": 8087
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00019157142857142856,
+      "loss": 0.1349,
+      "step": 8088
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00019155612244897957,
+      "loss": 0.2762,
+      "step": 8089
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00019154081632653062,
+      "loss": 0.0721,
+      "step": 8090
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00019152551020408163,
+      "loss": 0.2802,
+      "step": 8091
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00019151020408163263,
+      "loss": 0.2959,
+      "step": 8092
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00019149489795918363,
+      "loss": 0.0893,
+      "step": 8093
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0001914795918367347,
+      "loss": 0.1536,
+      "step": 8094
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0001914642857142857,
+      "loss": 0.1131,
+      "step": 8095
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00019144897959183673,
+      "loss": 0.1609,
+      "step": 8096
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00019143367346938773,
+      "loss": 0.0894,
+      "step": 8097
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00019141836734693876,
+      "loss": 0.1735,
+      "step": 8098
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0001914030612244898,
+      "loss": 0.105,
+      "step": 8099
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0001913877551020408,
+      "loss": 0.1451,
+      "step": 8100
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.0001913724489795918,
+      "loss": 0.0945,
+      "step": 8101
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00019135714285714285,
+      "loss": 0.1568,
+      "step": 8102
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00019134183673469386,
+      "loss": 0.128,
+      "step": 8103
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.0001913265306122449,
+      "loss": 0.2304,
+      "step": 8104
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.0001913112244897959,
+      "loss": 0.247,
+      "step": 8105
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00019129591836734692,
+      "loss": 0.106,
+      "step": 8106
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00019128061224489795,
+      "loss": 0.1061,
+      "step": 8107
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00019126530612244896,
+      "loss": 0.1674,
+      "step": 8108
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00019124999999999996,
+      "loss": 0.1241,
+      "step": 8109
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00019123469387755102,
+      "loss": 0.2186,
+      "step": 8110
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00019121938775510202,
+      "loss": 0.2334,
+      "step": 8111
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019120408163265305,
+      "loss": 0.1141,
+      "step": 8112
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019118877551020405,
+      "loss": 0.1677,
+      "step": 8113
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019117346938775509,
+      "loss": 0.1303,
+      "step": 8114
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019115816326530612,
+      "loss": 0.0898,
+      "step": 8115
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019114285714285712,
+      "loss": 0.2143,
+      "step": 8116
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019112755102040812,
+      "loss": 0.1666,
+      "step": 8117
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019111224489795918,
+      "loss": 0.1499,
+      "step": 8118
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019109693877551018,
+      "loss": 0.1313,
+      "step": 8119
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019108163265306121,
+      "loss": 0.0806,
+      "step": 8120
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00019106632653061222,
+      "loss": 0.1117,
+      "step": 8121
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00019105102040816325,
+      "loss": 0.0803,
+      "step": 8122
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00019103571428571428,
+      "loss": 0.2033,
+      "step": 8123
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00019102040816326528,
+      "loss": 0.2343,
+      "step": 8124
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00019100510204081629,
+      "loss": 0.2486,
+      "step": 8125
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00019098979591836734,
+      "loss": 0.0882,
+      "step": 8126
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00019097448979591835,
+      "loss": 0.0986,
+      "step": 8127
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00019095918367346935,
+      "loss": 0.1392,
+      "step": 8128
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.0001909438775510204,
+      "loss": 0.2289,
+      "step": 8129
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.0001909285714285714,
+      "loss": 0.1163,
+      "step": 8130
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00019091326530612244,
+      "loss": 0.1476,
+      "step": 8131
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.00019089795918367345,
+      "loss": 0.113,
+      "step": 8132
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.00019088265306122448,
+      "loss": 0.0905,
+      "step": 8133
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.0001908673469387755,
+      "loss": 0.1188,
+      "step": 8134
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.0001908520408163265,
+      "loss": 0.1382,
+      "step": 8135
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.0001908367346938775,
+      "loss": 0.2231,
+      "step": 8136
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.00019082142857142857,
+      "loss": 0.1828,
+      "step": 8137
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.00019080612244897957,
+      "loss": 0.1934,
+      "step": 8138
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.0001907908163265306,
+      "loss": 0.1201,
+      "step": 8139
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.0001907755102040816,
+      "loss": 0.16,
+      "step": 8140
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.00019076020408163264,
+      "loss": 0.1724,
+      "step": 8141
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.00019074489795918367,
+      "loss": 0.1565,
+      "step": 8142
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019072959183673467,
+      "loss": 0.1641,
+      "step": 8143
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019071428571428568,
+      "loss": 0.1699,
+      "step": 8144
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019069897959183673,
+      "loss": 0.1576,
+      "step": 8145
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019068367346938774,
+      "loss": 0.1618,
+      "step": 8146
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019066836734693877,
+      "loss": 0.0839,
+      "step": 8147
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019065306122448977,
+      "loss": 0.3022,
+      "step": 8148
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.0001906377551020408,
+      "loss": 0.2084,
+      "step": 8149
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019062244897959183,
+      "loss": 0.1548,
+      "step": 8150
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019060714285714284,
+      "loss": 0.1193,
+      "step": 8151
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00019059183673469384,
+      "loss": 0.0957,
+      "step": 8152
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.0001905765306122449,
+      "loss": 0.1122,
+      "step": 8153
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.0001905612244897959,
+      "loss": 0.253,
+      "step": 8154
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.0001905459183673469,
+      "loss": 0.1356,
+      "step": 8155
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00019053061224489793,
+      "loss": 0.1307,
+      "step": 8156
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00019051530612244896,
+      "loss": 0.1083,
+      "step": 8157
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.0001905,
+      "loss": 0.0881,
+      "step": 8158
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.000190484693877551,
+      "loss": 0.1128,
+      "step": 8159
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.000190469387755102,
+      "loss": 0.229,
+      "step": 8160
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00019045408163265306,
+      "loss": 0.1815,
+      "step": 8161
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00019043877551020406,
+      "loss": 0.1242,
+      "step": 8162
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019042346938775507,
+      "loss": 0.1563,
+      "step": 8163
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.0001904081632653061,
+      "loss": 0.1329,
+      "step": 8164
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019039285714285713,
+      "loss": 0.1143,
+      "step": 8165
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019037755102040816,
+      "loss": 0.1644,
+      "step": 8166
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019036224489795916,
+      "loss": 0.2277,
+      "step": 8167
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019034693877551017,
+      "loss": 0.0843,
+      "step": 8168
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019033163265306122,
+      "loss": 0.1059,
+      "step": 8169
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019031632653061223,
+      "loss": 0.1121,
+      "step": 8170
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019030102040816323,
+      "loss": 0.1703,
+      "step": 8171
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00019028571428571426,
+      "loss": 0.1156,
+      "step": 8172
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.0001902704081632653,
+      "loss": 0.2181,
+      "step": 8173
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00019025510204081632,
+      "loss": 0.1516,
+      "step": 8174
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00019023979591836733,
+      "loss": 0.1353,
+      "step": 8175
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00019022448979591836,
+      "loss": 0.1237,
+      "step": 8176
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00019020918367346939,
+      "loss": 0.0988,
+      "step": 8177
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.0001901938775510204,
+      "loss": 0.0762,
+      "step": 8178
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.0001901785714285714,
+      "loss": 0.1959,
+      "step": 8179
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00019016326530612245,
+      "loss": 0.2543,
+      "step": 8180
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00019014795918367345,
+      "loss": 0.2195,
+      "step": 8181
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00019013265306122448,
+      "loss": 0.1233,
+      "step": 8182
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.0001901173469387755,
+      "loss": 0.185,
+      "step": 8183
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00019010204081632652,
+      "loss": 0.0798,
+      "step": 8184
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00019008673469387755,
+      "loss": 0.1066,
+      "step": 8185
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00019007142857142855,
+      "loss": 0.1801,
+      "step": 8186
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00019005612244897956,
+      "loss": 0.1372,
+      "step": 8187
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00019004081632653061,
+      "loss": 0.198,
+      "step": 8188
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00019002551020408162,
+      "loss": 0.1575,
+      "step": 8189
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00019001020408163262,
+      "loss": 0.1712,
+      "step": 8190
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00018999489795918365,
+      "loss": 0.165,
+      "step": 8191
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00018997959183673468,
+      "loss": 0.3095,
+      "step": 8192
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.0001899642857142857,
+      "loss": 0.2009,
+      "step": 8193
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018994897959183672,
+      "loss": 0.1564,
+      "step": 8194
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018993367346938772,
+      "loss": 0.1844,
+      "step": 8195
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018991836734693878,
+      "loss": 0.1173,
+      "step": 8196
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018990306122448978,
+      "loss": 0.1629,
+      "step": 8197
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018988775510204078,
+      "loss": 0.2135,
+      "step": 8198
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018987244897959181,
+      "loss": 0.1471,
+      "step": 8199
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018985714285714284,
+      "loss": 0.1446,
+      "step": 8200
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018984183673469388,
+      "loss": 0.1982,
+      "step": 8201
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018982653061224488,
+      "loss": 0.0968,
+      "step": 8202
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018981122448979588,
+      "loss": 0.1313,
+      "step": 8203
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00018979591836734694,
+      "loss": 0.3085,
+      "step": 8204
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00018978061224489794,
+      "loss": 0.2494,
+      "step": 8205
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00018976530612244895,
+      "loss": 0.203,
+      "step": 8206
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00018974999999999998,
+      "loss": 0.1155,
+      "step": 8207
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.000189734693877551,
+      "loss": 0.1523,
+      "step": 8208
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00018971938775510204,
+      "loss": 0.116,
+      "step": 8209
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00018970408163265304,
+      "loss": 0.2636,
+      "step": 8210
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00018968877551020405,
+      "loss": 0.1674,
+      "step": 8211
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.0001896734693877551,
+      "loss": 0.2587,
+      "step": 8212
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.0001896581632653061,
+      "loss": 0.0859,
+      "step": 8213
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.0001896428571428571,
+      "loss": 0.1471,
+      "step": 8214
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018962755102040814,
+      "loss": 0.0726,
+      "step": 8215
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018961224489795917,
+      "loss": 0.1477,
+      "step": 8216
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018959693877551017,
+      "loss": 0.1447,
+      "step": 8217
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.0001895816326530612,
+      "loss": 0.0973,
+      "step": 8218
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018956632653061224,
+      "loss": 0.1232,
+      "step": 8219
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018955102040816327,
+      "loss": 0.1752,
+      "step": 8220
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018953571428571427,
+      "loss": 0.1294,
+      "step": 8221
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018952040816326527,
+      "loss": 0.0789,
+      "step": 8222
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018950510204081633,
+      "loss": 0.1686,
+      "step": 8223
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018948979591836733,
+      "loss": 0.1569,
+      "step": 8224
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.00018947448979591834,
+      "loss": 0.09,
+      "step": 8225
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.00018945918367346937,
+      "loss": 0.1033,
+      "step": 8226
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0001894438775510204,
+      "loss": 0.1171,
+      "step": 8227
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.00018942857142857143,
+      "loss": 0.073,
+      "step": 8228
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.00018941326530612243,
+      "loss": 0.2042,
+      "step": 8229
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.00018939795918367344,
+      "loss": 0.1354,
+      "step": 8230
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0001893826530612245,
+      "loss": 0.0994,
+      "step": 8231
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0001893673469387755,
+      "loss": 0.1825,
+      "step": 8232
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0001893520408163265,
+      "loss": 0.1217,
+      "step": 8233
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.00018933673469387753,
+      "loss": 0.0956,
+      "step": 8234
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00018932142857142856,
+      "loss": 0.1512,
+      "step": 8235
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.0001893061224489796,
+      "loss": 0.2501,
+      "step": 8236
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.0001892908163265306,
+      "loss": 0.1889,
+      "step": 8237
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.0001892755102040816,
+      "loss": 0.0844,
+      "step": 8238
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00018926020408163266,
+      "loss": 0.0941,
+      "step": 8239
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00018924489795918366,
+      "loss": 0.3069,
+      "step": 8240
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00018922959183673466,
+      "loss": 0.1848,
+      "step": 8241
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.0001892142857142857,
+      "loss": 0.135,
+      "step": 8242
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00018919897959183672,
+      "loss": 0.0846,
+      "step": 8243
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00018918367346938776,
+      "loss": 0.1657,
+      "step": 8244
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00018916836734693876,
+      "loss": 0.0659,
+      "step": 8245
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00018915306122448976,
+      "loss": 0.0619,
+      "step": 8246
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00018913775510204082,
+      "loss": 0.2109,
+      "step": 8247
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00018912244897959182,
+      "loss": 0.1004,
+      "step": 8248
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00018910714285714283,
+      "loss": 0.0993,
+      "step": 8249
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00018909183673469386,
+      "loss": 0.0589,
+      "step": 8250
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.0001890765306122449,
+      "loss": 0.1428,
+      "step": 8251
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.0001890612244897959,
+      "loss": 0.1497,
+      "step": 8252
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00018904591836734692,
+      "loss": 0.1177,
+      "step": 8253
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00018903061224489793,
+      "loss": 0.1158,
+      "step": 8254
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00018901530612244898,
+      "loss": 0.1129,
+      "step": 8255
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00018899999999999999,
+      "loss": 0.0933,
+      "step": 8256
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.000188984693877551,
+      "loss": 0.2195,
+      "step": 8257
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00018896938775510202,
+      "loss": 0.0658,
+      "step": 8258
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00018895408163265305,
+      "loss": 0.1827,
+      "step": 8259
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00018893877551020405,
+      "loss": 0.136,
+      "step": 8260
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00018892346938775508,
+      "loss": 0.1262,
+      "step": 8261
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.0001889081632653061,
+      "loss": 0.0718,
+      "step": 8262
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00018889285714285715,
+      "loss": 0.0647,
+      "step": 8263
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00018887755102040815,
+      "loss": 0.0923,
+      "step": 8264
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00018886224489795915,
+      "loss": 0.2281,
+      "step": 8265
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.0001888469387755102,
+      "loss": 0.1254,
+      "step": 8266
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00018883163265306121,
+      "loss": 0.136,
+      "step": 8267
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00018881632653061222,
+      "loss": 0.2012,
+      "step": 8268
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00018880102040816325,
+      "loss": 0.0749,
+      "step": 8269
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00018878571428571428,
+      "loss": 0.1022,
+      "step": 8270
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.0001887704081632653,
+      "loss": 0.1287,
+      "step": 8271
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.0001887551020408163,
+      "loss": 0.2164,
+      "step": 8272
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00018873979591836732,
+      "loss": 0.1168,
+      "step": 8273
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00018872448979591837,
+      "loss": 0.1214,
+      "step": 8274
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00018870918367346938,
+      "loss": 0.0753,
+      "step": 8275
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00018869387755102038,
+      "loss": 0.0955,
+      "step": 8276
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0001886785714285714,
+      "loss": 0.1173,
+      "step": 8277
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018866326530612244,
+      "loss": 0.2384,
+      "step": 8278
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018864795918367344,
+      "loss": 0.1155,
+      "step": 8279
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018863265306122448,
+      "loss": 0.1134,
+      "step": 8280
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018861734693877548,
+      "loss": 0.1028,
+      "step": 8281
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018860204081632654,
+      "loss": 0.0737,
+      "step": 8282
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018858673469387754,
+      "loss": 0.0589,
+      "step": 8283
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018857142857142854,
+      "loss": 0.2128,
+      "step": 8284
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018855612244897957,
+      "loss": 0.2272,
+      "step": 8285
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0001885408163265306,
+      "loss": 0.1782,
+      "step": 8286
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.0001885255102040816,
+      "loss": 0.1608,
+      "step": 8287
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00018851020408163264,
+      "loss": 0.1246,
+      "step": 8288
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00018849489795918364,
+      "loss": 0.1314,
+      "step": 8289
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.0001884795918367347,
+      "loss": 0.0873,
+      "step": 8290
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.0001884642857142857,
+      "loss": 0.146,
+      "step": 8291
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.0001884489795918367,
+      "loss": 0.0933,
+      "step": 8292
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00018843367346938774,
+      "loss": 0.0652,
+      "step": 8293
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00018841836734693877,
+      "loss": 0.1838,
+      "step": 8294
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00018840306122448977,
+      "loss": 0.0799,
+      "step": 8295
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.0001883877551020408,
+      "loss": 0.0614,
+      "step": 8296
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.0001883724489795918,
+      "loss": 0.1523,
+      "step": 8297
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00018835714285714286,
+      "loss": 0.1837,
+      "step": 8298
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00018834183673469387,
+      "loss": 0.1603,
+      "step": 8299
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00018832653061224487,
+      "loss": 0.1428,
+      "step": 8300
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.0001883112244897959,
+      "loss": 0.1308,
+      "step": 8301
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00018829591836734693,
+      "loss": 0.112,
+      "step": 8302
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00018828061224489793,
+      "loss": 0.197,
+      "step": 8303
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00018826530612244896,
+      "loss": 0.1848,
+      "step": 8304
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00018824999999999997,
+      "loss": 0.0997,
+      "step": 8305
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00018823469387755103,
+      "loss": 0.077,
+      "step": 8306
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00018821938775510203,
+      "loss": 0.1046,
+      "step": 8307
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00018820408163265303,
+      "loss": 0.1005,
+      "step": 8308
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.0001881887755102041,
+      "loss": 0.1014,
+      "step": 8309
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.0001881734693877551,
+      "loss": 0.1299,
+      "step": 8310
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.0001881581632653061,
+      "loss": 0.1285,
+      "step": 8311
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00018814285714285713,
+      "loss": 0.0747,
+      "step": 8312
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00018812755102040816,
+      "loss": 0.0909,
+      "step": 8313
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00018811224489795916,
+      "loss": 0.0973,
+      "step": 8314
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.0001880969387755102,
+      "loss": 0.1141,
+      "step": 8315
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.0001880816326530612,
+      "loss": 0.2617,
+      "step": 8316
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00018806632653061225,
+      "loss": 0.1182,
+      "step": 8317
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018805102040816326,
+      "loss": 0.1602,
+      "step": 8318
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018803571428571426,
+      "loss": 0.1937,
+      "step": 8319
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.0001880204081632653,
+      "loss": 0.145,
+      "step": 8320
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018800510204081632,
+      "loss": 0.1059,
+      "step": 8321
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018798979591836732,
+      "loss": 0.2749,
+      "step": 8322
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018797448979591836,
+      "loss": 0.134,
+      "step": 8323
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018795918367346936,
+      "loss": 0.1227,
+      "step": 8324
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018794387755102042,
+      "loss": 0.0802,
+      "step": 8325
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018792857142857142,
+      "loss": 0.1168,
+      "step": 8326
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00018791326530612242,
+      "loss": 0.108,
+      "step": 8327
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018789795918367345,
+      "loss": 0.1717,
+      "step": 8328
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018788265306122448,
+      "loss": 0.1453,
+      "step": 8329
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.0001878673469387755,
+      "loss": 0.1994,
+      "step": 8330
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018785204081632652,
+      "loss": 0.1486,
+      "step": 8331
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018783673469387752,
+      "loss": 0.1022,
+      "step": 8332
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018782142857142858,
+      "loss": 0.0737,
+      "step": 8333
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018780612244897958,
+      "loss": 0.2271,
+      "step": 8334
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018779081632653059,
+      "loss": 0.1966,
+      "step": 8335
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.0001877755102040816,
+      "loss": 0.101,
+      "step": 8336
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018776020408163265,
+      "loss": 0.0818,
+      "step": 8337
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018774489795918365,
+      "loss": 0.1139,
+      "step": 8338
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018772959183673468,
+      "loss": 0.0643,
+      "step": 8339
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018771428571428568,
+      "loss": 0.1461,
+      "step": 8340
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018769897959183674,
+      "loss": 0.1284,
+      "step": 8341
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018768367346938775,
+      "loss": 0.0783,
+      "step": 8342
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018766836734693875,
+      "loss": 0.0837,
+      "step": 8343
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018765306122448975,
+      "loss": 0.1403,
+      "step": 8344
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.0001876377551020408,
+      "loss": 0.0739,
+      "step": 8345
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018762244897959181,
+      "loss": 0.1208,
+      "step": 8346
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018760714285714284,
+      "loss": 0.1721,
+      "step": 8347
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018759183673469385,
+      "loss": 0.0776,
+      "step": 8348
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.00018757653061224488,
+      "loss": 0.1731,
+      "step": 8349
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.0001875612244897959,
+      "loss": 0.084,
+      "step": 8350
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.0001875459183673469,
+      "loss": 0.1194,
+      "step": 8351
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.00018753061224489792,
+      "loss": 0.0496,
+      "step": 8352
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.00018751530612244897,
+      "loss": 0.1318,
+      "step": 8353
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.00018749999999999998,
+      "loss": 0.0818,
+      "step": 8354
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.000187484693877551,
+      "loss": 0.1329,
+      "step": 8355
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.00018746938775510204,
+      "loss": 0.176,
+      "step": 8356
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.00018745408163265304,
+      "loss": 0.07,
+      "step": 8357
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.00018743877551020407,
+      "loss": 0.0815,
+      "step": 8358
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00018742346938775508,
+      "loss": 0.2295,
+      "step": 8359
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00018740816326530613,
+      "loss": 0.115,
+      "step": 8360
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00018739285714285714,
+      "loss": 0.0983,
+      "step": 8361
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00018737755102040814,
+      "loss": 0.1862,
+      "step": 8362
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00018736224489795917,
+      "loss": 0.1525,
+      "step": 8363
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.0001873469387755102,
+      "loss": 0.2321,
+      "step": 8364
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.0001873316326530612,
+      "loss": 0.2963,
+      "step": 8365
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00018731632653061223,
+      "loss": 0.1981,
+      "step": 8366
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00018730102040816324,
+      "loss": 0.1851,
+      "step": 8367
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.0001872857142857143,
+      "loss": 0.0895,
+      "step": 8368
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0001872704081632653,
+      "loss": 0.1044,
+      "step": 8369
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0001872551020408163,
+      "loss": 0.1144,
+      "step": 8370
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0001872397959183673,
+      "loss": 0.1256,
+      "step": 8371
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00018722448979591836,
+      "loss": 0.2048,
+      "step": 8372
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00018720918367346937,
+      "loss": 0.1974,
+      "step": 8373
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0001871938775510204,
+      "loss": 0.0872,
+      "step": 8374
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0001871785714285714,
+      "loss": 0.0893,
+      "step": 8375
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00018716326530612243,
+      "loss": 0.0701,
+      "step": 8376
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00018714795918367346,
+      "loss": 0.0796,
+      "step": 8377
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00018713265306122447,
+      "loss": 0.2182,
+      "step": 8378
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00018711734693877547,
+      "loss": 0.1206,
+      "step": 8379
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.00018710204081632653,
+      "loss": 0.1685,
+      "step": 8380
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.00018708673469387753,
+      "loss": 0.0793,
+      "step": 8381
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.00018707142857142856,
+      "loss": 0.1802,
+      "step": 8382
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.00018705612244897956,
+      "loss": 0.0984,
+      "step": 8383
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0001870408163265306,
+      "loss": 0.2248,
+      "step": 8384
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.00018702551020408163,
+      "loss": 0.0983,
+      "step": 8385
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.00018701020408163263,
+      "loss": 0.0829,
+      "step": 8386
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.00018699489795918363,
+      "loss": 0.1429,
+      "step": 8387
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0001869795918367347,
+      "loss": 0.12,
+      "step": 8388
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0001869642857142857,
+      "loss": 0.1445,
+      "step": 8389
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00018694897959183672,
+      "loss": 0.1914,
+      "step": 8390
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00018693367346938773,
+      "loss": 0.2366,
+      "step": 8391
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00018691836734693876,
+      "loss": 0.0886,
+      "step": 8392
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0001869030612244898,
+      "loss": 0.0876,
+      "step": 8393
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0001868877551020408,
+      "loss": 0.1269,
+      "step": 8394
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0001868724489795918,
+      "loss": 0.1021,
+      "step": 8395
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00018685714285714285,
+      "loss": 0.1637,
+      "step": 8396
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00018684183673469386,
+      "loss": 0.2805,
+      "step": 8397
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00018682653061224486,
+      "loss": 0.1274,
+      "step": 8398
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0001868112244897959,
+      "loss": 0.1555,
+      "step": 8399
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018679591836734692,
+      "loss": 0.0676,
+      "step": 8400
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018678061224489795,
+      "loss": 0.0965,
+      "step": 8401
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018676530612244896,
+      "loss": 0.1044,
+      "step": 8402
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018675,
+      "loss": 0.174,
+      "step": 8403
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018673469387755102,
+      "loss": 0.1397,
+      "step": 8404
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018671938775510202,
+      "loss": 0.0766,
+      "step": 8405
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018670408163265302,
+      "loss": 0.0835,
+      "step": 8406
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018668877551020408,
+      "loss": 0.1082,
+      "step": 8407
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018667346938775508,
+      "loss": 0.1125,
+      "step": 8408
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018665816326530611,
+      "loss": 0.1941,
+      "step": 8409
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018664285714285712,
+      "loss": 0.099,
+      "step": 8410
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018662755102040815,
+      "loss": 0.101,
+      "step": 8411
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018661224489795918,
+      "loss": 0.0718,
+      "step": 8412
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018659693877551018,
+      "loss": 0.1842,
+      "step": 8413
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018658163265306119,
+      "loss": 0.0959,
+      "step": 8414
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018656632653061224,
+      "loss": 0.2288,
+      "step": 8415
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018655102040816325,
+      "loss": 0.2095,
+      "step": 8416
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018653571428571428,
+      "loss": 0.1979,
+      "step": 8417
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018652040816326528,
+      "loss": 0.0742,
+      "step": 8418
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.0001865051020408163,
+      "loss": 0.0618,
+      "step": 8419
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00018648979591836734,
+      "loss": 0.1087,
+      "step": 8420
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.00018647448979591835,
+      "loss": 0.1057,
+      "step": 8421
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.00018645918367346935,
+      "loss": 0.1384,
+      "step": 8422
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0001864438775510204,
+      "loss": 0.1259,
+      "step": 8423
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0001864285714285714,
+      "loss": 0.1695,
+      "step": 8424
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.00018641326530612244,
+      "loss": 0.1296,
+      "step": 8425
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.00018639795918367344,
+      "loss": 0.1554,
+      "step": 8426
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.00018638265306122447,
+      "loss": 0.0906,
+      "step": 8427
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0001863673469387755,
+      "loss": 0.0933,
+      "step": 8428
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0001863520408163265,
+      "loss": 0.1012,
+      "step": 8429
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0001863367346938775,
+      "loss": 0.144,
+      "step": 8430
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018632142857142857,
+      "loss": 0.1414,
+      "step": 8431
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018630612244897957,
+      "loss": 0.105,
+      "step": 8432
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018629081632653058,
+      "loss": 0.1115,
+      "step": 8433
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.0001862755102040816,
+      "loss": 0.1776,
+      "step": 8434
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018626020408163264,
+      "loss": 0.1226,
+      "step": 8435
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018624489795918367,
+      "loss": 0.0851,
+      "step": 8436
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018622959183673467,
+      "loss": 0.1216,
+      "step": 8437
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018621428571428568,
+      "loss": 0.0885,
+      "step": 8438
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018619897959183673,
+      "loss": 0.1625,
+      "step": 8439
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018618367346938774,
+      "loss": 0.1248,
+      "step": 8440
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00018616836734693874,
+      "loss": 0.1676,
+      "step": 8441
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00018615306122448977,
+      "loss": 0.0717,
+      "step": 8442
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.0001861377551020408,
+      "loss": 0.1533,
+      "step": 8443
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00018612244897959183,
+      "loss": 0.0821,
+      "step": 8444
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00018610714285714283,
+      "loss": 0.1462,
+      "step": 8445
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00018609183673469387,
+      "loss": 0.1219,
+      "step": 8446
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.0001860765306122449,
+      "loss": 0.249,
+      "step": 8447
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.0001860612244897959,
+      "loss": 0.1334,
+      "step": 8448
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.0001860459183673469,
+      "loss": 0.1187,
+      "step": 8449
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00018603061224489796,
+      "loss": 0.1085,
+      "step": 8450
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00018601530612244896,
+      "loss": 0.1189,
+      "step": 8451
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.000186,
+      "loss": 0.1418,
+      "step": 8452
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.000185984693877551,
+      "loss": 0.1994,
+      "step": 8453
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018596938775510203,
+      "loss": 0.184,
+      "step": 8454
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018595408163265306,
+      "loss": 0.1659,
+      "step": 8455
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018593877551020406,
+      "loss": 0.2805,
+      "step": 8456
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018592346938775507,
+      "loss": 0.0727,
+      "step": 8457
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018590816326530612,
+      "loss": 0.0696,
+      "step": 8458
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018589285714285713,
+      "loss": 0.2058,
+      "step": 8459
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018587755102040816,
+      "loss": 0.131,
+      "step": 8460
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018586224489795916,
+      "loss": 0.1634,
+      "step": 8461
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.0001858469387755102,
+      "loss": 0.112,
+      "step": 8462
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00018583163265306122,
+      "loss": 0.1026,
+      "step": 8463
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00018581632653061223,
+      "loss": 0.0876,
+      "step": 8464
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00018580102040816323,
+      "loss": 0.2073,
+      "step": 8465
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.0001857857142857143,
+      "loss": 0.1583,
+      "step": 8466
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.0001857704081632653,
+      "loss": 0.1205,
+      "step": 8467
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.0001857551020408163,
+      "loss": 0.1366,
+      "step": 8468
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00018573979591836732,
+      "loss": 0.1428,
+      "step": 8469
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00018572448979591835,
+      "loss": 0.1691,
+      "step": 8470
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00018570918367346939,
+      "loss": 0.2066,
+      "step": 8471
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.0001856938775510204,
+      "loss": 0.2069,
+      "step": 8472
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.0001856785714285714,
+      "loss": 0.1099,
+      "step": 8473
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00018566326530612245,
+      "loss": 0.1282,
+      "step": 8474
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00018564795918367345,
+      "loss": 0.0789,
+      "step": 8475
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00018563265306122446,
+      "loss": 0.1076,
+      "step": 8476
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.0001856173469387755,
+      "loss": 0.1596,
+      "step": 8477
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00018560204081632652,
+      "loss": 0.2649,
+      "step": 8478
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00018558673469387755,
+      "loss": 0.1055,
+      "step": 8479
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00018557142857142855,
+      "loss": 0.1466,
+      "step": 8480
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00018555612244897956,
+      "loss": 0.1199,
+      "step": 8481
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.0001855408163265306,
+      "loss": 0.1119,
+      "step": 8482
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018552551020408162,
+      "loss": 0.1049,
+      "step": 8483
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018551020408163262,
+      "loss": 0.1358,
+      "step": 8484
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018549489795918365,
+      "loss": 0.2095,
+      "step": 8485
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018547959183673468,
+      "loss": 0.0918,
+      "step": 8486
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.0001854642857142857,
+      "loss": 0.1257,
+      "step": 8487
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018544897959183671,
+      "loss": 0.0727,
+      "step": 8488
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018543367346938772,
+      "loss": 0.075,
+      "step": 8489
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018541836734693878,
+      "loss": 0.188,
+      "step": 8490
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018540306122448978,
+      "loss": 0.1406,
+      "step": 8491
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018538775510204078,
+      "loss": 0.1542,
+      "step": 8492
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018537244897959184,
+      "loss": 0.1228,
+      "step": 8493
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018535714285714284,
+      "loss": 0.1451,
+      "step": 8494
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018534183673469385,
+      "loss": 0.1222,
+      "step": 8495
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018532653061224488,
+      "loss": 0.0863,
+      "step": 8496
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.0001853112244897959,
+      "loss": 0.2114,
+      "step": 8497
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018529591836734694,
+      "loss": 0.1376,
+      "step": 8498
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018528061224489794,
+      "loss": 0.1711,
+      "step": 8499
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018526530612244895,
+      "loss": 0.1012,
+      "step": 8500
+    },
+    {
+      "epoch": 8.25,
+      "eval_cer": 0.26276716017556395,
+      "eval_loss": 0.9553415179252625,
+      "eval_runtime": 247.6749,
+      "eval_samples_per_second": 35.587,
+      "eval_steps_per_second": 4.449,
+      "eval_wer": 0.6787177712597084,
+      "step": 8500
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018525,
+      "loss": 0.0977,
+      "step": 8501
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.000185234693877551,
+      "loss": 0.067,
+      "step": 8502
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.000185219387755102,
+      "loss": 0.1977,
+      "step": 8503
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.00018520408163265304,
+      "loss": 0.1723,
+      "step": 8504
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.00018518877551020407,
+      "loss": 0.0952,
+      "step": 8505
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0001851734693877551,
+      "loss": 0.1678,
+      "step": 8506
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0001851581632653061,
+      "loss": 0.1324,
+      "step": 8507
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0001851428571428571,
+      "loss": 0.061,
+      "step": 8508
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.00018512755102040817,
+      "loss": 0.2341,
+      "step": 8509
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.00018511224489795917,
+      "loss": 0.1583,
+      "step": 8510
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.00018509693877551017,
+      "loss": 0.1421,
+      "step": 8511
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0001850816326530612,
+      "loss": 0.091,
+      "step": 8512
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018506632653061223,
+      "loss": 0.1734,
+      "step": 8513
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018505102040816326,
+      "loss": 0.1138,
+      "step": 8514
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018503571428571427,
+      "loss": 0.0966,
+      "step": 8515
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018502040816326527,
+      "loss": 0.1193,
+      "step": 8516
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018500510204081633,
+      "loss": 0.2232,
+      "step": 8517
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018498979591836733,
+      "loss": 0.1371,
+      "step": 8518
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018497448979591834,
+      "loss": 0.1279,
+      "step": 8519
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018495918367346937,
+      "loss": 0.0699,
+      "step": 8520
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.0001849438775510204,
+      "loss": 0.1412,
+      "step": 8521
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018492857142857143,
+      "loss": 0.1057,
+      "step": 8522
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00018491326530612243,
+      "loss": 0.1047,
+      "step": 8523
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00018489795918367343,
+      "loss": 0.0776,
+      "step": 8524
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0001848826530612245,
+      "loss": 0.1092,
+      "step": 8525
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0001848673469387755,
+      "loss": 0.072,
+      "step": 8526
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0001848520408163265,
+      "loss": 0.1285,
+      "step": 8527
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00018483673469387753,
+      "loss": 0.1035,
+      "step": 8528
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00018482142857142856,
+      "loss": 0.213,
+      "step": 8529
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00018480612244897956,
+      "loss": 0.1663,
+      "step": 8530
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0001847908163265306,
+      "loss": 0.0652,
+      "step": 8531
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0001847755102040816,
+      "loss": 0.1246,
+      "step": 8532
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00018476020408163266,
+      "loss": 0.082,
+      "step": 8533
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018474489795918366,
+      "loss": 0.2145,
+      "step": 8534
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018472959183673466,
+      "loss": 0.1133,
+      "step": 8535
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018471428571428572,
+      "loss": 0.1,
+      "step": 8536
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018469897959183672,
+      "loss": 0.121,
+      "step": 8537
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018468367346938773,
+      "loss": 0.1992,
+      "step": 8538
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018466836734693876,
+      "loss": 0.1539,
+      "step": 8539
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.0001846530612244898,
+      "loss": 0.1828,
+      "step": 8540
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018463775510204082,
+      "loss": 0.1743,
+      "step": 8541
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018462244897959182,
+      "loss": 0.1598,
+      "step": 8542
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018460714285714283,
+      "loss": 0.1203,
+      "step": 8543
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00018459183673469388,
+      "loss": 0.1077,
+      "step": 8544
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.0001845765306122449,
+      "loss": 0.1241,
+      "step": 8545
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.0001845612244897959,
+      "loss": 0.0766,
+      "step": 8546
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00018454591836734692,
+      "loss": 0.2081,
+      "step": 8547
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00018453061224489795,
+      "loss": 0.1306,
+      "step": 8548
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00018451530612244898,
+      "loss": 0.1029,
+      "step": 8549
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00018449999999999999,
+      "loss": 0.1423,
+      "step": 8550
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.000184484693877551,
+      "loss": 0.1358,
+      "step": 8551
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00018446938775510205,
+      "loss": 0.1424,
+      "step": 8552
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00018445408163265305,
+      "loss": 0.2238,
+      "step": 8553
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00018443877551020405,
+      "loss": 0.2619,
+      "step": 8554
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00018442346938775508,
+      "loss": 0.0852,
+      "step": 8555
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00018440816326530611,
+      "loss": 0.1234,
+      "step": 8556
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00018439285714285712,
+      "loss": 0.0827,
+      "step": 8557
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00018437755102040815,
+      "loss": 0.1438,
+      "step": 8558
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00018436224489795915,
+      "loss": 0.206,
+      "step": 8559
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.0001843469387755102,
+      "loss": 0.133,
+      "step": 8560
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.0001843316326530612,
+      "loss": 0.1259,
+      "step": 8561
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00018431632653061222,
+      "loss": 0.1539,
+      "step": 8562
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00018430102040816325,
+      "loss": 0.1144,
+      "step": 8563
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00018428571428571428,
+      "loss": 0.092,
+      "step": 8564
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00018427040816326528,
+      "loss": 0.0807,
+      "step": 8565
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.0001842551020408163,
+      "loss": 0.1919,
+      "step": 8566
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00018423979591836731,
+      "loss": 0.1475,
+      "step": 8567
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00018422448979591837,
+      "loss": 0.1094,
+      "step": 8568
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00018420918367346938,
+      "loss": 0.0739,
+      "step": 8569
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00018419387755102038,
+      "loss": 0.0738,
+      "step": 8570
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.0001841785714285714,
+      "loss": 0.0597,
+      "step": 8571
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00018416326530612244,
+      "loss": 0.2121,
+      "step": 8572
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00018414795918367344,
+      "loss": 0.1287,
+      "step": 8573
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00018413265306122447,
+      "loss": 0.1142,
+      "step": 8574
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00018411734693877548,
+      "loss": 0.1622,
+      "step": 8575
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00018410204081632654,
+      "loss": 0.1518,
+      "step": 8576
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00018408673469387754,
+      "loss": 0.1231,
+      "step": 8577
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00018407142857142854,
+      "loss": 0.1469,
+      "step": 8578
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00018405612244897957,
+      "loss": 0.1916,
+      "step": 8579
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.0001840408163265306,
+      "loss": 0.0842,
+      "step": 8580
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.0001840255102040816,
+      "loss": 0.089,
+      "step": 8581
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00018401020408163264,
+      "loss": 0.0678,
+      "step": 8582
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00018399489795918367,
+      "loss": 0.0782,
+      "step": 8583
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.0001839795918367347,
+      "loss": 0.2407,
+      "step": 8584
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.0001839642857142857,
+      "loss": 0.1691,
+      "step": 8585
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0001839489795918367,
+      "loss": 0.2386,
+      "step": 8586
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00018393367346938776,
+      "loss": 0.1268,
+      "step": 8587
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00018391836734693877,
+      "loss": 0.2051,
+      "step": 8588
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00018390306122448977,
+      "loss": 0.1012,
+      "step": 8589
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0001838877551020408,
+      "loss": 0.1208,
+      "step": 8590
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00018387244897959183,
+      "loss": 0.1372,
+      "step": 8591
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00018385714285714283,
+      "loss": 0.1817,
+      "step": 8592
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00018384183673469386,
+      "loss": 0.0871,
+      "step": 8593
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00018382653061224487,
+      "loss": 0.0627,
+      "step": 8594
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.00018381122448979593,
+      "loss": 0.1229,
+      "step": 8595
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00018379591836734693,
+      "loss": 0.1587,
+      "step": 8596
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00018378061224489793,
+      "loss": 0.1853,
+      "step": 8597
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00018376530612244896,
+      "loss": 0.0804,
+      "step": 8598
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00018375,
+      "loss": 0.138,
+      "step": 8599
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.000183734693877551,
+      "loss": 0.111,
+      "step": 8600
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00018371938775510203,
+      "loss": 0.0996,
+      "step": 8601
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00018370408163265303,
+      "loss": 0.1281,
+      "step": 8602
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.0001836887755102041,
+      "loss": 0.0943,
+      "step": 8603
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.0001836734693877551,
+      "loss": 0.1236,
+      "step": 8604
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.0001836581632653061,
+      "loss": 0.0872,
+      "step": 8605
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00018364285714285713,
+      "loss": 0.0782,
+      "step": 8606
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00018362755102040816,
+      "loss": 0.1101,
+      "step": 8607
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00018361224489795916,
+      "loss": 0.2244,
+      "step": 8608
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.0001835969387755102,
+      "loss": 0.0828,
+      "step": 8609
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.0001835816326530612,
+      "loss": 0.1055,
+      "step": 8610
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00018356632653061225,
+      "loss": 0.1077,
+      "step": 8611
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00018355102040816326,
+      "loss": 0.0962,
+      "step": 8612
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00018353571428571426,
+      "loss": 0.0833,
+      "step": 8613
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00018352040816326526,
+      "loss": 0.0607,
+      "step": 8614
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00018350510204081632,
+      "loss": 0.1958,
+      "step": 8615
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018348979591836732,
+      "loss": 0.1133,
+      "step": 8616
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018347448979591835,
+      "loss": 0.1217,
+      "step": 8617
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018345918367346936,
+      "loss": 0.1638,
+      "step": 8618
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018344387755102042,
+      "loss": 0.1177,
+      "step": 8619
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018342857142857142,
+      "loss": 0.1054,
+      "step": 8620
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018341326530612242,
+      "loss": 0.0802,
+      "step": 8621
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018339795918367343,
+      "loss": 0.1989,
+      "step": 8622
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018338265306122448,
+      "loss": 0.2186,
+      "step": 8623
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.0001833673469387755,
+      "loss": 0.1299,
+      "step": 8624
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018335204081632652,
+      "loss": 0.0851,
+      "step": 8625
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00018333673469387752,
+      "loss": 0.1146,
+      "step": 8626
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018332142857142855,
+      "loss": 0.064,
+      "step": 8627
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018330612244897958,
+      "loss": 0.2395,
+      "step": 8628
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018329081632653059,
+      "loss": 0.1888,
+      "step": 8629
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018327551020408164,
+      "loss": 0.113,
+      "step": 8630
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018326020408163265,
+      "loss": 0.1137,
+      "step": 8631
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018324489795918365,
+      "loss": 0.0765,
+      "step": 8632
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018322959183673468,
+      "loss": 0.1142,
+      "step": 8633
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.0001832142857142857,
+      "loss": 0.2089,
+      "step": 8634
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018319897959183671,
+      "loss": 0.1953,
+      "step": 8635
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00018318367346938774,
+      "loss": 0.0785,
+      "step": 8636
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.00018316836734693875,
+      "loss": 0.0644,
+      "step": 8637
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0001831530612244898,
+      "loss": 0.1294,
+      "step": 8638
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0001831377551020408,
+      "loss": 0.1629,
+      "step": 8639
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0001831224489795918,
+      "loss": 0.2464,
+      "step": 8640
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.00018310714285714284,
+      "loss": 0.1676,
+      "step": 8641
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.00018309183673469387,
+      "loss": 0.1264,
+      "step": 8642
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.00018307653061224488,
+      "loss": 0.0872,
+      "step": 8643
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0001830612244897959,
+      "loss": 0.0785,
+      "step": 8644
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0001830459183673469,
+      "loss": 0.0814,
+      "step": 8645
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.00018303061224489797,
+      "loss": 0.0632,
+      "step": 8646
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018301530612244897,
+      "loss": 0.1337,
+      "step": 8647
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018299999999999998,
+      "loss": 0.1412,
+      "step": 8648
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018298469387755098,
+      "loss": 0.1904,
+      "step": 8649
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018296938775510204,
+      "loss": 0.078,
+      "step": 8650
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018295408163265304,
+      "loss": 0.1089,
+      "step": 8651
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018293877551020407,
+      "loss": 0.1247,
+      "step": 8652
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018292346938775507,
+      "loss": 0.164,
+      "step": 8653
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.0001829081632653061,
+      "loss": 0.0801,
+      "step": 8654
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018289285714285714,
+      "loss": 0.0913,
+      "step": 8655
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018287755102040814,
+      "loss": 0.2084,
+      "step": 8656
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00018286224489795914,
+      "loss": 0.0706,
+      "step": 8657
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0001828469387755102,
+      "loss": 0.1317,
+      "step": 8658
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0001828316326530612,
+      "loss": 0.1467,
+      "step": 8659
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00018281632653061223,
+      "loss": 0.1855,
+      "step": 8660
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00018280102040816324,
+      "loss": 0.0979,
+      "step": 8661
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00018278571428571427,
+      "loss": 0.117,
+      "step": 8662
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0001827704081632653,
+      "loss": 0.1442,
+      "step": 8663
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0001827551020408163,
+      "loss": 0.0951,
+      "step": 8664
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0001827397959183673,
+      "loss": 0.1162,
+      "step": 8665
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00018272448979591836,
+      "loss": 0.1249,
+      "step": 8666
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00018270918367346937,
+      "loss": 0.1499,
+      "step": 8667
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.0001826938775510204,
+      "loss": 0.0993,
+      "step": 8668
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.0001826785714285714,
+      "loss": 0.0939,
+      "step": 8669
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00018266326530612243,
+      "loss": 0.1173,
+      "step": 8670
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00018264795918367346,
+      "loss": 0.1278,
+      "step": 8671
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00018263265306122447,
+      "loss": 0.1999,
+      "step": 8672
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00018261734693877552,
+      "loss": 0.1536,
+      "step": 8673
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00018260204081632653,
+      "loss": 0.126,
+      "step": 8674
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00018258673469387753,
+      "loss": 0.1206,
+      "step": 8675
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00018257142857142853,
+      "loss": 0.0896,
+      "step": 8676
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.0001825561224489796,
+      "loss": 0.1861,
+      "step": 8677
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0001825408163265306,
+      "loss": 0.1342,
+      "step": 8678
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00018252551020408162,
+      "loss": 0.0995,
+      "step": 8679
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00018251020408163263,
+      "loss": 0.2258,
+      "step": 8680
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00018249489795918369,
+      "loss": 0.107,
+      "step": 8681
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0001824795918367347,
+      "loss": 0.0683,
+      "step": 8682
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0001824642857142857,
+      "loss": 0.1133,
+      "step": 8683
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0001824489795918367,
+      "loss": 0.1341,
+      "step": 8684
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00018243367346938775,
+      "loss": 0.229,
+      "step": 8685
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00018241836734693876,
+      "loss": 0.1505,
+      "step": 8686
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0001824030612244898,
+      "loss": 0.0893,
+      "step": 8687
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0001823877551020408,
+      "loss": 0.1374,
+      "step": 8688
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018237244897959182,
+      "loss": 0.1077,
+      "step": 8689
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018235714285714285,
+      "loss": 0.1597,
+      "step": 8690
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018234183673469386,
+      "loss": 0.2481,
+      "step": 8691
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018232653061224486,
+      "loss": 0.1445,
+      "step": 8692
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018231122448979592,
+      "loss": 0.1359,
+      "step": 8693
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018229591836734692,
+      "loss": 0.0787,
+      "step": 8694
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018228061224489795,
+      "loss": 0.1469,
+      "step": 8695
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018226530612244895,
+      "loss": 0.1446,
+      "step": 8696
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018224999999999998,
+      "loss": 0.3022,
+      "step": 8697
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018223469387755102,
+      "loss": 0.1776,
+      "step": 8698
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018221938775510202,
+      "loss": 0.1141,
+      "step": 8699
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018220408163265302,
+      "loss": 0.2453,
+      "step": 8700
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018218877551020408,
+      "loss": 0.0831,
+      "step": 8701
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018217346938775508,
+      "loss": 0.0906,
+      "step": 8702
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018215816326530611,
+      "loss": 0.2105,
+      "step": 8703
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018214285714285712,
+      "loss": 0.1678,
+      "step": 8704
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018212755102040815,
+      "loss": 0.193,
+      "step": 8705
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018211224489795918,
+      "loss": 0.0889,
+      "step": 8706
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018209693877551018,
+      "loss": 0.1188,
+      "step": 8707
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00018208163265306119,
+      "loss": 0.0923,
+      "step": 8708
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00018206632653061224,
+      "loss": 0.2357,
+      "step": 8709
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00018205102040816325,
+      "loss": 0.2107,
+      "step": 8710
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00018203571428571425,
+      "loss": 0.1431,
+      "step": 8711
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00018202040816326528,
+      "loss": 0.081,
+      "step": 8712
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0001820051020408163,
+      "loss": 0.1343,
+      "step": 8713
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00018198979591836734,
+      "loss": 0.1087,
+      "step": 8714
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00018197448979591834,
+      "loss": 0.1416,
+      "step": 8715
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00018195918367346935,
+      "loss": 0.1894,
+      "step": 8716
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0001819438775510204,
+      "loss": 0.1409,
+      "step": 8717
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0001819285714285714,
+      "loss": 0.1589,
+      "step": 8718
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.0001819132653061224,
+      "loss": 0.0943,
+      "step": 8719
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018189795918367347,
+      "loss": 0.1128,
+      "step": 8720
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018188265306122447,
+      "loss": 0.0988,
+      "step": 8721
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.0001818673469387755,
+      "loss": 0.2096,
+      "step": 8722
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.0001818520408163265,
+      "loss": 0.1817,
+      "step": 8723
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018183673469387754,
+      "loss": 0.2217,
+      "step": 8724
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018182142857142857,
+      "loss": 0.0743,
+      "step": 8725
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018180612244897957,
+      "loss": 0.0955,
+      "step": 8726
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018179081632653058,
+      "loss": 0.0895,
+      "step": 8727
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018177551020408163,
+      "loss": 0.2598,
+      "step": 8728
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018176020408163264,
+      "loss": 0.1037,
+      "step": 8729
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.00018174489795918367,
+      "loss": 0.1383,
+      "step": 8730
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.00018172959183673467,
+      "loss": 0.1206,
+      "step": 8731
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0001817142857142857,
+      "loss": 0.0881,
+      "step": 8732
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.00018169897959183673,
+      "loss": 0.1439,
+      "step": 8733
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.00018168367346938774,
+      "loss": 0.2341,
+      "step": 8734
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.00018166836734693874,
+      "loss": 0.1657,
+      "step": 8735
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0001816530612244898,
+      "loss": 0.1055,
+      "step": 8736
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0001816377551020408,
+      "loss": 0.1066,
+      "step": 8737
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0001816224489795918,
+      "loss": 0.1332,
+      "step": 8738
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.00018160714285714283,
+      "loss": 0.112,
+      "step": 8739
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00018159183673469386,
+      "loss": 0.1018,
+      "step": 8740
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.0001815765306122449,
+      "loss": 0.1651,
+      "step": 8741
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.0001815612244897959,
+      "loss": 0.1067,
+      "step": 8742
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.0001815459183673469,
+      "loss": 0.1123,
+      "step": 8743
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00018153061224489796,
+      "loss": 0.0787,
+      "step": 8744
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00018151530612244896,
+      "loss": 0.1134,
+      "step": 8745
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00018149999999999997,
+      "loss": 0.1863,
+      "step": 8746
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.000181484693877551,
+      "loss": 0.2273,
+      "step": 8747
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00018146938775510203,
+      "loss": 0.0817,
+      "step": 8748
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00018145408163265306,
+      "loss": 0.1618,
+      "step": 8749
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018143877551020406,
+      "loss": 0.1286,
+      "step": 8750
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018142346938775507,
+      "loss": 0.1289,
+      "step": 8751
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018140816326530612,
+      "loss": 0.0804,
+      "step": 8752
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018139285714285713,
+      "loss": 0.2344,
+      "step": 8753
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018137755102040813,
+      "loss": 0.2483,
+      "step": 8754
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018136224489795916,
+      "loss": 0.1417,
+      "step": 8755
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.0001813469387755102,
+      "loss": 0.1103,
+      "step": 8756
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018133163265306122,
+      "loss": 0.0598,
+      "step": 8757
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018131632653061222,
+      "loss": 0.1615,
+      "step": 8758
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018130102040816323,
+      "loss": 0.2108,
+      "step": 8759
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018128571428571429,
+      "loss": 0.1157,
+      "step": 8760
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0001812704081632653,
+      "loss": 0.0991,
+      "step": 8761
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0001812551020408163,
+      "loss": 0.0965,
+      "step": 8762
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00018123979591836735,
+      "loss": 0.1159,
+      "step": 8763
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00018122448979591835,
+      "loss": 0.1281,
+      "step": 8764
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00018120918367346938,
+      "loss": 0.2183,
+      "step": 8765
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0001811938775510204,
+      "loss": 0.1993,
+      "step": 8766
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00018117857142857142,
+      "loss": 0.1993,
+      "step": 8767
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00018116326530612245,
+      "loss": 0.1138,
+      "step": 8768
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00018114795918367345,
+      "loss": 0.1618,
+      "step": 8769
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00018113265306122446,
+      "loss": 0.0956,
+      "step": 8770
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.0001811173469387755,
+      "loss": 0.1338,
+      "step": 8771
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00018110204081632652,
+      "loss": 0.1335,
+      "step": 8772
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00018108673469387752,
+      "loss": 0.0828,
+      "step": 8773
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00018107142857142855,
+      "loss": 0.1082,
+      "step": 8774
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00018105612244897958,
+      "loss": 0.1545,
+      "step": 8775
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.0001810408163265306,
+      "loss": 0.1226,
+      "step": 8776
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00018102551020408162,
+      "loss": 0.1073,
+      "step": 8777
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00018101020408163262,
+      "loss": 0.3437,
+      "step": 8778
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00018099489795918368,
+      "loss": 0.1139,
+      "step": 8779
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00018097959183673468,
+      "loss": 0.1228,
+      "step": 8780
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018096428571428568,
+      "loss": 0.1663,
+      "step": 8781
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018094897959183671,
+      "loss": 0.1196,
+      "step": 8782
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018093367346938774,
+      "loss": 0.0739,
+      "step": 8783
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018091836734693877,
+      "loss": 0.1603,
+      "step": 8784
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018090306122448978,
+      "loss": 0.1025,
+      "step": 8785
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018088775510204078,
+      "loss": 0.0974,
+      "step": 8786
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018087244897959184,
+      "loss": 0.1684,
+      "step": 8787
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018085714285714284,
+      "loss": 0.1494,
+      "step": 8788
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018084183673469385,
+      "loss": 0.1568,
+      "step": 8789
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018082653061224488,
+      "loss": 0.1515,
+      "step": 8790
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.0001808112244897959,
+      "loss": 0.1049,
+      "step": 8791
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.00018079591836734694,
+      "loss": 0.0897,
+      "step": 8792
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.00018078061224489794,
+      "loss": 0.0707,
+      "step": 8793
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.00018076530612244894,
+      "loss": 0.0789,
+      "step": 8794
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.00018075,
+      "loss": 0.1085,
+      "step": 8795
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.000180734693877551,
+      "loss": 0.1106,
+      "step": 8796
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.000180719387755102,
+      "loss": 0.2719,
+      "step": 8797
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.00018070408163265304,
+      "loss": 0.1097,
+      "step": 8798
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.00018068877551020407,
+      "loss": 0.0994,
+      "step": 8799
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.0001806734693877551,
+      "loss": 0.1777,
+      "step": 8800
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.0001806581632653061,
+      "loss": 0.1389,
+      "step": 8801
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.0001806428571428571,
+      "loss": 0.0861,
+      "step": 8802
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00018062755102040817,
+      "loss": 0.2467,
+      "step": 8803
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00018061224489795917,
+      "loss": 0.1024,
+      "step": 8804
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00018059693877551017,
+      "loss": 0.1096,
+      "step": 8805
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.0001805816326530612,
+      "loss": 0.0633,
+      "step": 8806
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00018056632653061223,
+      "loss": 0.1115,
+      "step": 8807
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00018055102040816324,
+      "loss": 0.0684,
+      "step": 8808
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00018053571428571427,
+      "loss": 0.2039,
+      "step": 8809
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.0001805204081632653,
+      "loss": 0.1643,
+      "step": 8810
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00018050510204081633,
+      "loss": 0.2452,
+      "step": 8811
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.00018048979591836733,
+      "loss": 0.1166,
+      "step": 8812
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.00018047448979591834,
+      "loss": 0.0998,
+      "step": 8813
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0001804591836734694,
+      "loss": 0.0904,
+      "step": 8814
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0001804438775510204,
+      "loss": 0.2053,
+      "step": 8815
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0001804285714285714,
+      "loss": 0.0971,
+      "step": 8816
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.00018041326530612243,
+      "loss": 0.1024,
+      "step": 8817
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.00018039795918367346,
+      "loss": 0.088,
+      "step": 8818
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0001803826530612245,
+      "loss": 0.1305,
+      "step": 8819
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0001803673469387755,
+      "loss": 0.0687,
+      "step": 8820
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0001803520408163265,
+      "loss": 0.1221,
+      "step": 8821
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018033673469387756,
+      "loss": 0.2293,
+      "step": 8822
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018032142857142856,
+      "loss": 0.1516,
+      "step": 8823
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018030612244897956,
+      "loss": 0.0791,
+      "step": 8824
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.0001802908163265306,
+      "loss": 0.0677,
+      "step": 8825
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018027551020408162,
+      "loss": 0.225,
+      "step": 8826
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018026020408163265,
+      "loss": 0.1106,
+      "step": 8827
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018024489795918366,
+      "loss": 0.1933,
+      "step": 8828
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018022959183673466,
+      "loss": 0.1039,
+      "step": 8829
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018021428571428572,
+      "loss": 0.1264,
+      "step": 8830
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018019897959183672,
+      "loss": 0.0979,
+      "step": 8831
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00018018367346938773,
+      "loss": 0.0604,
+      "step": 8832
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00018016836734693876,
+      "loss": 0.0971,
+      "step": 8833
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.0001801530612244898,
+      "loss": 0.2446,
+      "step": 8834
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.0001801377551020408,
+      "loss": 0.103,
+      "step": 8835
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00018012244897959182,
+      "loss": 0.115,
+      "step": 8836
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00018010714285714282,
+      "loss": 0.0773,
+      "step": 8837
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00018009183673469388,
+      "loss": 0.1111,
+      "step": 8838
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00018007653061224489,
+      "loss": 0.0716,
+      "step": 8839
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.0001800612244897959,
+      "loss": 0.2261,
+      "step": 8840
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00018004591836734692,
+      "loss": 0.201,
+      "step": 8841
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00018003061224489795,
+      "loss": 0.1605,
+      "step": 8842
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00018001530612244895,
+      "loss": 0.1135,
+      "step": 8843
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.1748,
+      "step": 8844
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.000179984693877551,
+      "loss": 0.1042,
+      "step": 8845
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00017996938775510205,
+      "loss": 0.0814,
+      "step": 8846
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00017995408163265305,
+      "loss": 0.1347,
+      "step": 8847
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00017993877551020405,
+      "loss": 0.1415,
+      "step": 8848
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00017992346938775508,
+      "loss": 0.1613,
+      "step": 8849
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.0001799081632653061,
+      "loss": 0.1335,
+      "step": 8850
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00017989285714285712,
+      "loss": 0.0688,
+      "step": 8851
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00017987755102040815,
+      "loss": 0.1265,
+      "step": 8852
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00017986224489795915,
+      "loss": 0.3203,
+      "step": 8853
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0001798469387755102,
+      "loss": 0.1614,
+      "step": 8854
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0001798316326530612,
+      "loss": 0.1348,
+      "step": 8855
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00017981632653061222,
+      "loss": 0.0977,
+      "step": 8856
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00017980102040816327,
+      "loss": 0.1168,
+      "step": 8857
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00017978571428571428,
+      "loss": 0.1043,
+      "step": 8858
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00017977040816326528,
+      "loss": 0.2434,
+      "step": 8859
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0001797551020408163,
+      "loss": 0.1055,
+      "step": 8860
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00017973979591836734,
+      "loss": 0.1792,
+      "step": 8861
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00017972448979591837,
+      "loss": 0.1185,
+      "step": 8862
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00017970918367346937,
+      "loss": 0.0809,
+      "step": 8863
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00017969387755102038,
+      "loss": 0.0822,
+      "step": 8864
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00017967857142857144,
+      "loss": 0.2957,
+      "step": 8865
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00017966326530612244,
+      "loss": 0.1701,
+      "step": 8866
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00017964795918367344,
+      "loss": 0.2134,
+      "step": 8867
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00017963265306122447,
+      "loss": 0.0728,
+      "step": 8868
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0001796173469387755,
+      "loss": 0.1241,
+      "step": 8869
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0001796020408163265,
+      "loss": 0.1099,
+      "step": 8870
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00017958673469387754,
+      "loss": 0.1384,
+      "step": 8871
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00017957142857142854,
+      "loss": 0.1731,
+      "step": 8872
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0001795561224489796,
+      "loss": 0.0832,
+      "step": 8873
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.0001795408163265306,
+      "loss": 0.1422,
+      "step": 8874
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.0001795255102040816,
+      "loss": 0.1482,
+      "step": 8875
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00017951020408163264,
+      "loss": 0.1298,
+      "step": 8876
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00017949489795918367,
+      "loss": 0.0769,
+      "step": 8877
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00017947959183673467,
+      "loss": 0.1785,
+      "step": 8878
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.0001794642857142857,
+      "loss": 0.1414,
+      "step": 8879
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.0001794489795918367,
+      "loss": 0.1274,
+      "step": 8880
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00017943367346938776,
+      "loss": 0.0866,
+      "step": 8881
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00017941836734693877,
+      "loss": 0.0751,
+      "step": 8882
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00017940306122448977,
+      "loss": 0.0939,
+      "step": 8883
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.0001793877551020408,
+      "loss": 0.1114,
+      "step": 8884
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017937244897959183,
+      "loss": 0.1054,
+      "step": 8885
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017935714285714283,
+      "loss": 0.1245,
+      "step": 8886
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017934183673469386,
+      "loss": 0.1041,
+      "step": 8887
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017932653061224487,
+      "loss": 0.0839,
+      "step": 8888
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017931122448979593,
+      "loss": 0.0604,
+      "step": 8889
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017929591836734693,
+      "loss": 0.1542,
+      "step": 8890
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017928061224489793,
+      "loss": 0.1509,
+      "step": 8891
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017926530612244894,
+      "loss": 0.2273,
+      "step": 8892
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017925,
+      "loss": 0.1154,
+      "step": 8893
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.000179234693877551,
+      "loss": 0.1909,
+      "step": 8894
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00017921938775510203,
+      "loss": 0.176,
+      "step": 8895
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00017920408163265303,
+      "loss": 0.0601,
+      "step": 8896
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00017918877551020406,
+      "loss": 0.1395,
+      "step": 8897
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.0001791734693877551,
+      "loss": 0.211,
+      "step": 8898
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.0001791581632653061,
+      "loss": 0.152,
+      "step": 8899
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00017914285714285715,
+      "loss": 0.1394,
+      "step": 8900
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00017912755102040816,
+      "loss": 0.121,
+      "step": 8901
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00017911224489795916,
+      "loss": 0.0629,
+      "step": 8902
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.0001790969387755102,
+      "loss": 0.1327,
+      "step": 8903
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00017908163265306122,
+      "loss": 0.1452,
+      "step": 8904
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017906632653061222,
+      "loss": 0.1553,
+      "step": 8905
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017905102040816325,
+      "loss": 0.0812,
+      "step": 8906
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017903571428571426,
+      "loss": 0.0741,
+      "step": 8907
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017902040816326532,
+      "loss": 0.1139,
+      "step": 8908
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017900510204081632,
+      "loss": 0.2462,
+      "step": 8909
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017898979591836732,
+      "loss": 0.1013,
+      "step": 8910
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017897448979591835,
+      "loss": 0.0817,
+      "step": 8911
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017895918367346938,
+      "loss": 0.1238,
+      "step": 8912
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.0001789438775510204,
+      "loss": 0.1323,
+      "step": 8913
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017892857142857142,
+      "loss": 0.1402,
+      "step": 8914
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017891326530612242,
+      "loss": 0.1225,
+      "step": 8915
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017889795918367348,
+      "loss": 0.1685,
+      "step": 8916
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017888265306122448,
+      "loss": 0.2034,
+      "step": 8917
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017886734693877549,
+      "loss": 0.1565,
+      "step": 8918
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017885204081632652,
+      "loss": 0.1485,
+      "step": 8919
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017883673469387755,
+      "loss": 0.0888,
+      "step": 8920
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017882142857142855,
+      "loss": 0.0719,
+      "step": 8921
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017880612244897958,
+      "loss": 0.1738,
+      "step": 8922
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017879081632653058,
+      "loss": 0.0904,
+      "step": 8923
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017877551020408164,
+      "loss": 0.1221,
+      "step": 8924
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00017876020408163265,
+      "loss": 0.2418,
+      "step": 8925
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00017874489795918365,
+      "loss": 0.0604,
+      "step": 8926
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00017872959183673465,
+      "loss": 0.1896,
+      "step": 8927
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.0001787142857142857,
+      "loss": 0.2136,
+      "step": 8928
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.0001786989795918367,
+      "loss": 0.1029,
+      "step": 8929
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00017868367346938774,
+      "loss": 0.0932,
+      "step": 8930
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00017866836734693875,
+      "loss": 0.1079,
+      "step": 8931
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00017865306122448978,
+      "loss": 0.1036,
+      "step": 8932
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.0001786377551020408,
+      "loss": 0.1316,
+      "step": 8933
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.0001786224489795918,
+      "loss": 0.194,
+      "step": 8934
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00017860714285714282,
+      "loss": 0.1656,
+      "step": 8935
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00017859183673469387,
+      "loss": 0.1331,
+      "step": 8936
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00017857653061224488,
+      "loss": 0.0905,
+      "step": 8937
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.0001785612244897959,
+      "loss": 0.2486,
+      "step": 8938
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.0001785459183673469,
+      "loss": 0.1077,
+      "step": 8939
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00017853061224489794,
+      "loss": 0.1801,
+      "step": 8940
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00017851530612244897,
+      "loss": 0.2635,
+      "step": 8941
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00017849999999999997,
+      "loss": 0.1309,
+      "step": 8942
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00017848469387755098,
+      "loss": 0.1206,
+      "step": 8943
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00017846938775510204,
+      "loss": 0.2448,
+      "step": 8944
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00017845408163265304,
+      "loss": 0.1631,
+      "step": 8945
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.00017843877551020407,
+      "loss": 0.1987,
+      "step": 8946
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0001784234693877551,
+      "loss": 0.2346,
+      "step": 8947
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0001784081632653061,
+      "loss": 0.1522,
+      "step": 8948
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.00017839285714285713,
+      "loss": 0.0709,
+      "step": 8949
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.00017837755102040814,
+      "loss": 0.1936,
+      "step": 8950
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0001783622448979592,
+      "loss": 0.097,
+      "step": 8951
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0001783469387755102,
+      "loss": 0.1538,
+      "step": 8952
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0001783316326530612,
+      "loss": 0.1522,
+      "step": 8953
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0001783163265306122,
+      "loss": 0.1455,
+      "step": 8954
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.00017830102040816326,
+      "loss": 0.1483,
+      "step": 8955
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017828571428571427,
+      "loss": 0.1596,
+      "step": 8956
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.0001782704081632653,
+      "loss": 0.2062,
+      "step": 8957
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.0001782551020408163,
+      "loss": 0.0803,
+      "step": 8958
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017823979591836736,
+      "loss": 0.1904,
+      "step": 8959
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017822448979591836,
+      "loss": 0.1691,
+      "step": 8960
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017820918367346937,
+      "loss": 0.0995,
+      "step": 8961
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017819387755102037,
+      "loss": 0.1055,
+      "step": 8962
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017817857142857143,
+      "loss": 0.1316,
+      "step": 8963
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017816326530612243,
+      "loss": 0.0739,
+      "step": 8964
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017814795918367346,
+      "loss": 0.2043,
+      "step": 8965
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017813265306122446,
+      "loss": 0.1536,
+      "step": 8966
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0001781173469387755,
+      "loss": 0.1448,
+      "step": 8967
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00017810204081632653,
+      "loss": 0.0942,
+      "step": 8968
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00017808673469387753,
+      "loss": 0.1216,
+      "step": 8969
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00017807142857142853,
+      "loss": 0.1524,
+      "step": 8970
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0001780561224489796,
+      "loss": 0.151,
+      "step": 8971
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0001780408163265306,
+      "loss": 0.2603,
+      "step": 8972
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00017802551020408162,
+      "loss": 0.1392,
+      "step": 8973
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00017801020408163263,
+      "loss": 0.1191,
+      "step": 8974
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00017799489795918366,
+      "loss": 0.0943,
+      "step": 8975
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0001779795918367347,
+      "loss": 0.1549,
+      "step": 8976
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0001779642857142857,
+      "loss": 0.0847,
+      "step": 8977
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0001779489795918367,
+      "loss": 0.166,
+      "step": 8978
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00017793367346938775,
+      "loss": 0.1754,
+      "step": 8979
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00017791836734693876,
+      "loss": 0.1332,
+      "step": 8980
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0001779030612244898,
+      "loss": 0.0849,
+      "step": 8981
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0001778877551020408,
+      "loss": 0.0727,
+      "step": 8982
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00017787244897959182,
+      "loss": 0.0864,
+      "step": 8983
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00017785714285714285,
+      "loss": 0.1551,
+      "step": 8984
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00017784183673469385,
+      "loss": 0.1215,
+      "step": 8985
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00017782653061224486,
+      "loss": 0.0907,
+      "step": 8986
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017781122448979592,
+      "loss": 0.1324,
+      "step": 8987
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017779591836734692,
+      "loss": 0.1149,
+      "step": 8988
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017778061224489792,
+      "loss": 0.0801,
+      "step": 8989
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017776530612244898,
+      "loss": 0.119,
+      "step": 8990
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017774999999999998,
+      "loss": 0.1083,
+      "step": 8991
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017773469387755101,
+      "loss": 0.0927,
+      "step": 8992
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017771938775510202,
+      "loss": 0.1013,
+      "step": 8993
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017770408163265305,
+      "loss": 0.1591,
+      "step": 8994
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017768877551020408,
+      "loss": 0.1113,
+      "step": 8995
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017767346938775508,
+      "loss": 0.0808,
+      "step": 8996
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00017765816326530609,
+      "loss": 0.2492,
+      "step": 8997
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00017764285714285714,
+      "loss": 0.2152,
+      "step": 8998
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00017762755102040815,
+      "loss": 0.0812,
+      "step": 8999
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00017761224489795918,
+      "loss": 0.1018,
+      "step": 9000
+    },
+    {
+      "epoch": 8.74,
+      "eval_cer": 0.25109161754351456,
+      "eval_loss": 0.9455049633979797,
+      "eval_runtime": 240.8935,
+      "eval_samples_per_second": 36.589,
+      "eval_steps_per_second": 4.575,
+      "eval_wer": 0.6445473420558283,
+      "step": 9000
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00017759693877551018,
+      "loss": 0.1245,
+      "step": 9001
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.0001775816326530612,
+      "loss": 0.1614,
+      "step": 9002
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00017756632653061224,
+      "loss": 0.2481,
+      "step": 9003
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00017755102040816325,
+      "loss": 0.1298,
+      "step": 9004
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00017753571428571425,
+      "loss": 0.1228,
+      "step": 9005
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.0001775204081632653,
+      "loss": 0.1165,
+      "step": 9006
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.0001775051020408163,
+      "loss": 0.0776,
+      "step": 9007
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00017748979591836734,
+      "loss": 0.0588,
+      "step": 9008
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00017747448979591834,
+      "loss": 0.1614,
+      "step": 9009
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00017745918367346937,
+      "loss": 0.2454,
+      "step": 9010
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0001774438775510204,
+      "loss": 0.1106,
+      "step": 9011
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0001774285714285714,
+      "loss": 0.157,
+      "step": 9012
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0001774132653061224,
+      "loss": 0.1257,
+      "step": 9013
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00017739795918367347,
+      "loss": 0.1178,
+      "step": 9014
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00017738265306122447,
+      "loss": 0.18,
+      "step": 9015
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00017736734693877548,
+      "loss": 0.2066,
+      "step": 9016
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0001773520408163265,
+      "loss": 0.129,
+      "step": 9017
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017733673469387754,
+      "loss": 0.1602,
+      "step": 9018
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017732142857142857,
+      "loss": 0.2789,
+      "step": 9019
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017730612244897957,
+      "loss": 0.0792,
+      "step": 9020
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017729081632653057,
+      "loss": 0.0761,
+      "step": 9021
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017727551020408163,
+      "loss": 0.1163,
+      "step": 9022
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017726020408163264,
+      "loss": 0.1636,
+      "step": 9023
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017724489795918364,
+      "loss": 0.0861,
+      "step": 9024
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017722959183673467,
+      "loss": 0.1147,
+      "step": 9025
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.0001772142857142857,
+      "loss": 0.1161,
+      "step": 9026
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00017719897959183673,
+      "loss": 0.1524,
+      "step": 9027
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00017718367346938773,
+      "loss": 0.2136,
+      "step": 9028
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00017716836734693874,
+      "loss": 0.129,
+      "step": 9029
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.0001771530612244898,
+      "loss": 0.0891,
+      "step": 9030
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.0001771377551020408,
+      "loss": 0.1489,
+      "step": 9031
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.0001771224489795918,
+      "loss": 0.0887,
+      "step": 9032
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00017710714285714283,
+      "loss": 0.1092,
+      "step": 9033
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00017709183673469386,
+      "loss": 0.1953,
+      "step": 9034
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.0001770765306122449,
+      "loss": 0.1553,
+      "step": 9035
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.0001770612244897959,
+      "loss": 0.1257,
+      "step": 9036
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00017704591836734693,
+      "loss": 0.1847,
+      "step": 9037
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00017703061224489796,
+      "loss": 0.0904,
+      "step": 9038
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017701530612244896,
+      "loss": 0.0657,
+      "step": 9039
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017699999999999997,
+      "loss": 0.2319,
+      "step": 9040
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017698469387755102,
+      "loss": 0.1868,
+      "step": 9041
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017696938775510203,
+      "loss": 0.1364,
+      "step": 9042
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017695408163265306,
+      "loss": 0.1241,
+      "step": 9043
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017693877551020406,
+      "loss": 0.1374,
+      "step": 9044
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.0001769234693877551,
+      "loss": 0.0824,
+      "step": 9045
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017690816326530612,
+      "loss": 0.1542,
+      "step": 9046
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017689285714285713,
+      "loss": 0.2095,
+      "step": 9047
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017687755102040813,
+      "loss": 0.2119,
+      "step": 9048
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00017686224489795919,
+      "loss": 0.1085,
+      "step": 9049
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.0001768469387755102,
+      "loss": 0.0913,
+      "step": 9050
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.0001768316326530612,
+      "loss": 0.0878,
+      "step": 9051
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00017681632653061222,
+      "loss": 0.1636,
+      "step": 9052
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00017680102040816325,
+      "loss": 0.1036,
+      "step": 9053
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00017678571428571428,
+      "loss": 0.1404,
+      "step": 9054
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.0001767704081632653,
+      "loss": 0.1458,
+      "step": 9055
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.0001767551020408163,
+      "loss": 0.1594,
+      "step": 9056
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00017673979591836735,
+      "loss": 0.1551,
+      "step": 9057
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00017672448979591835,
+      "loss": 0.1405,
+      "step": 9058
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017670918367346936,
+      "loss": 0.2212,
+      "step": 9059
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.0001766938775510204,
+      "loss": 0.1295,
+      "step": 9060
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017667857142857142,
+      "loss": 0.1616,
+      "step": 9061
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017666326530612245,
+      "loss": 0.1334,
+      "step": 9062
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017664795918367345,
+      "loss": 0.1941,
+      "step": 9063
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017663265306122445,
+      "loss": 0.0674,
+      "step": 9064
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.0001766173469387755,
+      "loss": 0.2198,
+      "step": 9065
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017660204081632652,
+      "loss": 0.1193,
+      "step": 9066
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017658673469387752,
+      "loss": 0.1818,
+      "step": 9067
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017657142857142855,
+      "loss": 0.0943,
+      "step": 9068
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017655612244897958,
+      "loss": 0.1303,
+      "step": 9069
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.0001765408163265306,
+      "loss": 0.1355,
+      "step": 9070
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017652551020408161,
+      "loss": 0.0842,
+      "step": 9071
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017651020408163262,
+      "loss": 0.2635,
+      "step": 9072
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017649489795918368,
+      "loss": 0.3516,
+      "step": 9073
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017647959183673468,
+      "loss": 0.0913,
+      "step": 9074
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017646428571428568,
+      "loss": 0.1264,
+      "step": 9075
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.0001764489795918367,
+      "loss": 0.0757,
+      "step": 9076
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017643367346938774,
+      "loss": 0.1077,
+      "step": 9077
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017641836734693877,
+      "loss": 0.1211,
+      "step": 9078
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017640306122448978,
+      "loss": 0.199,
+      "step": 9079
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.0001763877551020408,
+      "loss": 0.1925,
+      "step": 9080
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00017637244897959184,
+      "loss": 0.1122,
+      "step": 9081
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00017635714285714284,
+      "loss": 0.0675,
+      "step": 9082
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00017634183673469385,
+      "loss": 0.0931,
+      "step": 9083
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.0001763265306122449,
+      "loss": 0.1224,
+      "step": 9084
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.0001763112244897959,
+      "loss": 0.1997,
+      "step": 9085
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.0001762959183673469,
+      "loss": 0.1799,
+      "step": 9086
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00017628061224489794,
+      "loss": 0.0836,
+      "step": 9087
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00017626530612244897,
+      "loss": 0.0875,
+      "step": 9088
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00017625,
+      "loss": 0.1496,
+      "step": 9089
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.000176234693877551,
+      "loss": 0.2068,
+      "step": 9090
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.000176219387755102,
+      "loss": 0.3373,
+      "step": 9091
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017620408163265307,
+      "loss": 0.1226,
+      "step": 9092
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017618877551020407,
+      "loss": 0.0721,
+      "step": 9093
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017617346938775507,
+      "loss": 0.0772,
+      "step": 9094
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.0001761581632653061,
+      "loss": 0.0827,
+      "step": 9095
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017614285714285713,
+      "loss": 0.1307,
+      "step": 9096
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017612755102040816,
+      "loss": 0.1955,
+      "step": 9097
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017611224489795917,
+      "loss": 0.1365,
+      "step": 9098
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017609693877551017,
+      "loss": 0.1246,
+      "step": 9099
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017608163265306123,
+      "loss": 0.1369,
+      "step": 9100
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.00017606632653061223,
+      "loss": 0.0618,
+      "step": 9101
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.00017605102040816324,
+      "loss": 0.1831,
+      "step": 9102
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.00017603571428571427,
+      "loss": 0.2864,
+      "step": 9103
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0001760204081632653,
+      "loss": 0.0901,
+      "step": 9104
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.00017600510204081633,
+      "loss": 0.1179,
+      "step": 9105
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.00017598979591836733,
+      "loss": 0.1241,
+      "step": 9106
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.00017597448979591833,
+      "loss": 0.0763,
+      "step": 9107
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0001759591836734694,
+      "loss": 0.0741,
+      "step": 9108
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0001759438775510204,
+      "loss": 0.1312,
+      "step": 9109
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0001759285714285714,
+      "loss": 0.1537,
+      "step": 9110
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017591326530612243,
+      "loss": 0.1956,
+      "step": 9111
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017589795918367346,
+      "loss": 0.1598,
+      "step": 9112
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017588265306122446,
+      "loss": 0.2265,
+      "step": 9113
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.0001758673469387755,
+      "loss": 0.0779,
+      "step": 9114
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.0001758520408163265,
+      "loss": 0.1269,
+      "step": 9115
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017583673469387756,
+      "loss": 0.2305,
+      "step": 9116
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017582142857142856,
+      "loss": 0.1102,
+      "step": 9117
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017580612244897956,
+      "loss": 0.1215,
+      "step": 9118
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.0001757908163265306,
+      "loss": 0.1141,
+      "step": 9119
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017577551020408162,
+      "loss": 0.0655,
+      "step": 9120
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017576020408163263,
+      "loss": 0.1777,
+      "step": 9121
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017574489795918366,
+      "loss": 0.1552,
+      "step": 9122
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017572959183673466,
+      "loss": 0.1839,
+      "step": 9123
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017571428571428572,
+      "loss": 0.1185,
+      "step": 9124
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017569897959183672,
+      "loss": 0.0825,
+      "step": 9125
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017568367346938773,
+      "loss": 0.1603,
+      "step": 9126
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017566836734693878,
+      "loss": 0.1455,
+      "step": 9127
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017565306122448979,
+      "loss": 0.1596,
+      "step": 9128
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.0001756377551020408,
+      "loss": 0.1621,
+      "step": 9129
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017562244897959182,
+      "loss": 0.107,
+      "step": 9130
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017560714285714285,
+      "loss": 0.1187,
+      "step": 9131
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017559183673469388,
+      "loss": 0.1058,
+      "step": 9132
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017557653061224488,
+      "loss": 0.1457,
+      "step": 9133
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.0001755612244897959,
+      "loss": 0.1415,
+      "step": 9134
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017554591836734695,
+      "loss": 0.1695,
+      "step": 9135
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017553061224489795,
+      "loss": 0.2218,
+      "step": 9136
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017551530612244895,
+      "loss": 0.0905,
+      "step": 9137
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017549999999999998,
+      "loss": 0.0956,
+      "step": 9138
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017548469387755101,
+      "loss": 0.1346,
+      "step": 9139
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017546938775510204,
+      "loss": 0.167,
+      "step": 9140
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00017545408163265305,
+      "loss": 0.1897,
+      "step": 9141
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00017543877551020405,
+      "loss": 0.1394,
+      "step": 9142
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.0001754234693877551,
+      "loss": 0.165,
+      "step": 9143
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.0001754081632653061,
+      "loss": 0.1444,
+      "step": 9144
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00017539285714285712,
+      "loss": 0.0798,
+      "step": 9145
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00017537755102040815,
+      "loss": 0.1488,
+      "step": 9146
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00017536224489795918,
+      "loss": 0.1566,
+      "step": 9147
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00017534693877551018,
+      "loss": 0.1749,
+      "step": 9148
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.0001753316326530612,
+      "loss": 0.1016,
+      "step": 9149
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00017531632653061221,
+      "loss": 0.1135,
+      "step": 9150
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00017530102040816327,
+      "loss": 0.1515,
+      "step": 9151
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017528571428571428,
+      "loss": 0.1419,
+      "step": 9152
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017527040816326528,
+      "loss": 0.1902,
+      "step": 9153
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.0001752551020408163,
+      "loss": 0.2241,
+      "step": 9154
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017523979591836734,
+      "loss": 0.1252,
+      "step": 9155
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017522448979591834,
+      "loss": 0.1411,
+      "step": 9156
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017520918367346937,
+      "loss": 0.0987,
+      "step": 9157
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017519387755102038,
+      "loss": 0.0907,
+      "step": 9158
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017517857142857144,
+      "loss": 0.1715,
+      "step": 9159
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017516326530612244,
+      "loss": 0.1858,
+      "step": 9160
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017514795918367344,
+      "loss": 0.1495,
+      "step": 9161
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00017513265306122447,
+      "loss": 0.0801,
+      "step": 9162
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0001751173469387755,
+      "loss": 0.1139,
+      "step": 9163
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0001751020408163265,
+      "loss": 0.1277,
+      "step": 9164
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00017508673469387754,
+      "loss": 0.2341,
+      "step": 9165
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00017507142857142854,
+      "loss": 0.2319,
+      "step": 9166
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0001750561224489796,
+      "loss": 0.1149,
+      "step": 9167
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0001750408163265306,
+      "loss": 0.073,
+      "step": 9168
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0001750255102040816,
+      "loss": 0.0729,
+      "step": 9169
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0001750102040816326,
+      "loss": 0.0769,
+      "step": 9170
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00017499489795918367,
+      "loss": 0.1308,
+      "step": 9171
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00017497959183673467,
+      "loss": 0.1821,
+      "step": 9172
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.0001749642857142857,
+      "loss": 0.144,
+      "step": 9173
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017494897959183673,
+      "loss": 0.1656,
+      "step": 9174
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017493367346938773,
+      "loss": 0.1794,
+      "step": 9175
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017491836734693876,
+      "loss": 0.1045,
+      "step": 9176
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017490306122448977,
+      "loss": 0.0776,
+      "step": 9177
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017488775510204083,
+      "loss": 0.1634,
+      "step": 9178
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017487244897959183,
+      "loss": 0.1554,
+      "step": 9179
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017485714285714283,
+      "loss": 0.1577,
+      "step": 9180
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017484183673469386,
+      "loss": 0.0964,
+      "step": 9181
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.0001748265306122449,
+      "loss": 0.0794,
+      "step": 9182
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.0001748112244897959,
+      "loss": 0.104,
+      "step": 9183
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00017479591836734693,
+      "loss": 0.106,
+      "step": 9184
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00017478061224489793,
+      "loss": 0.1154,
+      "step": 9185
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.000174765306122449,
+      "loss": 0.1447,
+      "step": 9186
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00017475,
+      "loss": 0.1698,
+      "step": 9187
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.000174734693877551,
+      "loss": 0.1584,
+      "step": 9188
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00017471938775510203,
+      "loss": 0.1621,
+      "step": 9189
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00017470408163265306,
+      "loss": 0.2379,
+      "step": 9190
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00017468877551020406,
+      "loss": 0.1651,
+      "step": 9191
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.0001746734693877551,
+      "loss": 0.0987,
+      "step": 9192
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.0001746581632653061,
+      "loss": 0.1024,
+      "step": 9193
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017464285714285715,
+      "loss": 0.1575,
+      "step": 9194
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017462755102040816,
+      "loss": 0.1153,
+      "step": 9195
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017461224489795916,
+      "loss": 0.175,
+      "step": 9196
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017459693877551016,
+      "loss": 0.3241,
+      "step": 9197
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017458163265306122,
+      "loss": 0.2004,
+      "step": 9198
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017456632653061222,
+      "loss": 0.0974,
+      "step": 9199
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017455102040816325,
+      "loss": 0.0739,
+      "step": 9200
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017453571428571426,
+      "loss": 0.0623,
+      "step": 9201
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017452040816326531,
+      "loss": 0.1415,
+      "step": 9202
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017450510204081632,
+      "loss": 0.2829,
+      "step": 9203
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017448979591836732,
+      "loss": 0.1051,
+      "step": 9204
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017447448979591833,
+      "loss": 0.0938,
+      "step": 9205
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017445918367346938,
+      "loss": 0.0555,
+      "step": 9206
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017444387755102039,
+      "loss": 0.1389,
+      "step": 9207
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017442857142857142,
+      "loss": 0.0937,
+      "step": 9208
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017441326530612242,
+      "loss": 0.1233,
+      "step": 9209
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017439795918367345,
+      "loss": 0.1816,
+      "step": 9210
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017438265306122448,
+      "loss": 0.1704,
+      "step": 9211
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017436734693877548,
+      "loss": 0.1349,
+      "step": 9212
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.0001743520408163265,
+      "loss": 0.1164,
+      "step": 9213
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.00017433673469387755,
+      "loss": 0.1027,
+      "step": 9214
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.00017432142857142855,
+      "loss": 0.1183,
+      "step": 9215
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.00017430612244897958,
+      "loss": 0.1843,
+      "step": 9216
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0001742908163265306,
+      "loss": 0.1311,
+      "step": 9217
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.00017427551020408161,
+      "loss": 0.1179,
+      "step": 9218
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.00017426020408163264,
+      "loss": 0.0851,
+      "step": 9219
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.00017424489795918365,
+      "loss": 0.1445,
+      "step": 9220
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0001742295918367347,
+      "loss": 0.0718,
+      "step": 9221
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0001742142857142857,
+      "loss": 0.2249,
+      "step": 9222
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0001741989795918367,
+      "loss": 0.1773,
+      "step": 9223
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017418367346938774,
+      "loss": 0.1342,
+      "step": 9224
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017416836734693877,
+      "loss": 0.1387,
+      "step": 9225
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017415306122448978,
+      "loss": 0.1098,
+      "step": 9226
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.0001741377551020408,
+      "loss": 0.0776,
+      "step": 9227
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.0001741224489795918,
+      "loss": 0.2363,
+      "step": 9228
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017410714285714287,
+      "loss": 0.1006,
+      "step": 9229
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017409183673469387,
+      "loss": 0.1301,
+      "step": 9230
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017407653061224488,
+      "loss": 0.0783,
+      "step": 9231
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017406122448979588,
+      "loss": 0.1843,
+      "step": 9232
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017404591836734694,
+      "loss": 0.0787,
+      "step": 9233
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017403061224489794,
+      "loss": 0.2196,
+      "step": 9234
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017401530612244897,
+      "loss": 0.1844,
+      "step": 9235
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017399999999999997,
+      "loss": 0.0865,
+      "step": 9236
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.000173984693877551,
+      "loss": 0.129,
+      "step": 9237
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017396938775510204,
+      "loss": 0.1079,
+      "step": 9238
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017395408163265304,
+      "loss": 0.0767,
+      "step": 9239
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017393877551020404,
+      "loss": 0.1104,
+      "step": 9240
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.0001739234693877551,
+      "loss": 0.2546,
+      "step": 9241
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.0001739081632653061,
+      "loss": 0.0949,
+      "step": 9242
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017389285714285713,
+      "loss": 0.0807,
+      "step": 9243
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017387755102040814,
+      "loss": 0.1085,
+      "step": 9244
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00017386224489795917,
+      "loss": 0.0717,
+      "step": 9245
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0001738469387755102,
+      "loss": 0.0887,
+      "step": 9246
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0001738316326530612,
+      "loss": 0.1818,
+      "step": 9247
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0001738163265306122,
+      "loss": 0.1453,
+      "step": 9248
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00017380102040816326,
+      "loss": 0.0897,
+      "step": 9249
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00017378571428571427,
+      "loss": 0.1242,
+      "step": 9250
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0001737704081632653,
+      "loss": 0.1062,
+      "step": 9251
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0001737551020408163,
+      "loss": 0.0689,
+      "step": 9252
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00017373979591836733,
+      "loss": 0.2431,
+      "step": 9253
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00017372448979591836,
+      "loss": 0.1523,
+      "step": 9254
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017370918367346936,
+      "loss": 0.1155,
+      "step": 9255
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017369387755102037,
+      "loss": 0.112,
+      "step": 9256
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017367857142857143,
+      "loss": 0.0861,
+      "step": 9257
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017366326530612243,
+      "loss": 0.1509,
+      "step": 9258
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017364795918367346,
+      "loss": 0.1188,
+      "step": 9259
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017363265306122446,
+      "loss": 0.0954,
+      "step": 9260
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.0001736173469387755,
+      "loss": 0.175,
+      "step": 9261
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017360204081632652,
+      "loss": 0.0676,
+      "step": 9262
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017358673469387753,
+      "loss": 0.0698,
+      "step": 9263
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017357142857142859,
+      "loss": 0.172,
+      "step": 9264
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0001735561224489796,
+      "loss": 0.2152,
+      "step": 9265
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0001735408163265306,
+      "loss": 0.1549,
+      "step": 9266
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0001735255102040816,
+      "loss": 0.1023,
+      "step": 9267
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00017351020408163265,
+      "loss": 0.1017,
+      "step": 9268
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00017349489795918366,
+      "loss": 0.0974,
+      "step": 9269
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0001734795918367347,
+      "loss": 0.1509,
+      "step": 9270
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0001734642857142857,
+      "loss": 0.1239,
+      "step": 9271
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00017344897959183672,
+      "loss": 0.087,
+      "step": 9272
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00017343367346938775,
+      "loss": 0.0808,
+      "step": 9273
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00017341836734693876,
+      "loss": 0.117,
+      "step": 9274
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00017340306122448976,
+      "loss": 0.1441,
+      "step": 9275
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017338775510204082,
+      "loss": 0.1012,
+      "step": 9276
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017337244897959182,
+      "loss": 0.1399,
+      "step": 9277
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017335714285714285,
+      "loss": 0.0821,
+      "step": 9278
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017334183673469385,
+      "loss": 0.08,
+      "step": 9279
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017332653061224488,
+      "loss": 0.1154,
+      "step": 9280
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017331122448979591,
+      "loss": 0.0817,
+      "step": 9281
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017329591836734692,
+      "loss": 0.1018,
+      "step": 9282
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017328061224489792,
+      "loss": 0.2238,
+      "step": 9283
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017326530612244898,
+      "loss": 0.1586,
+      "step": 9284
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00017324999999999998,
+      "loss": 0.11,
+      "step": 9285
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.000173234693877551,
+      "loss": 0.0623,
+      "step": 9286
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017321938775510202,
+      "loss": 0.1036,
+      "step": 9287
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017320408163265305,
+      "loss": 0.0532,
+      "step": 9288
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017318877551020408,
+      "loss": 0.1566,
+      "step": 9289
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017317346938775508,
+      "loss": 0.1673,
+      "step": 9290
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017315816326530608,
+      "loss": 0.125,
+      "step": 9291
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017314285714285714,
+      "loss": 0.0856,
+      "step": 9292
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017312755102040815,
+      "loss": 0.1633,
+      "step": 9293
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017311224489795915,
+      "loss": 0.1297,
+      "step": 9294
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017309693877551018,
+      "loss": 0.1382,
+      "step": 9295
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.0001730816326530612,
+      "loss": 0.2242,
+      "step": 9296
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.00017306632653061224,
+      "loss": 0.0842,
+      "step": 9297
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.00017305102040816324,
+      "loss": 0.1141,
+      "step": 9298
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.00017303571428571425,
+      "loss": 0.1373,
+      "step": 9299
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.0001730204081632653,
+      "loss": 0.1025,
+      "step": 9300
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.0001730051020408163,
+      "loss": 0.1173,
+      "step": 9301
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.0001729897959183673,
+      "loss": 0.1269,
+      "step": 9302
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.00017297448979591834,
+      "loss": 0.0942,
+      "step": 9303
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.00017295918367346937,
+      "loss": 0.1393,
+      "step": 9304
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.0001729438775510204,
+      "loss": 0.0713,
+      "step": 9305
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.0001729285714285714,
+      "loss": 0.1185,
+      "step": 9306
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017291326530612244,
+      "loss": 0.143,
+      "step": 9307
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017289795918367347,
+      "loss": 0.185,
+      "step": 9308
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017288265306122447,
+      "loss": 0.1311,
+      "step": 9309
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017286734693877548,
+      "loss": 0.1036,
+      "step": 9310
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017285204081632653,
+      "loss": 0.1211,
+      "step": 9311
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017283673469387754,
+      "loss": 0.0523,
+      "step": 9312
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017282142857142857,
+      "loss": 0.0649,
+      "step": 9313
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017280612244897957,
+      "loss": 0.2879,
+      "step": 9314
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.0001727908163265306,
+      "loss": 0.1726,
+      "step": 9315
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017277551020408163,
+      "loss": 0.0749,
+      "step": 9316
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00017276020408163264,
+      "loss": 0.0484,
+      "step": 9317
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00017274489795918364,
+      "loss": 0.059,
+      "step": 9318
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0001727295918367347,
+      "loss": 0.1146,
+      "step": 9319
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0001727142857142857,
+      "loss": 0.1855,
+      "step": 9320
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00017269897959183673,
+      "loss": 0.1898,
+      "step": 9321
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00017268367346938773,
+      "loss": 0.1692,
+      "step": 9322
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00017266836734693876,
+      "loss": 0.0756,
+      "step": 9323
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0001726530612244898,
+      "loss": 0.0969,
+      "step": 9324
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0001726377551020408,
+      "loss": 0.0682,
+      "step": 9325
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0001726224489795918,
+      "loss": 0.1445,
+      "step": 9326
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017260714285714286,
+      "loss": 0.1409,
+      "step": 9327
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017259183673469386,
+      "loss": 0.2343,
+      "step": 9328
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017257653061224487,
+      "loss": 0.1174,
+      "step": 9329
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.0001725612244897959,
+      "loss": 0.0746,
+      "step": 9330
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017254591836734693,
+      "loss": 0.0917,
+      "step": 9331
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017253061224489796,
+      "loss": 0.0676,
+      "step": 9332
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017251530612244896,
+      "loss": 0.1476,
+      "step": 9333
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017249999999999996,
+      "loss": 0.172,
+      "step": 9334
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017248469387755102,
+      "loss": 0.1181,
+      "step": 9335
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017246938775510203,
+      "loss": 0.1036,
+      "step": 9336
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00017245408163265303,
+      "loss": 0.1202,
+      "step": 9337
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00017243877551020406,
+      "loss": 0.0535,
+      "step": 9338
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.0001724234693877551,
+      "loss": 0.1418,
+      "step": 9339
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00017240816326530612,
+      "loss": 0.1335,
+      "step": 9340
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00017239285714285712,
+      "loss": 0.1857,
+      "step": 9341
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00017237755102040813,
+      "loss": 0.0736,
+      "step": 9342
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00017236224489795919,
+      "loss": 0.0749,
+      "step": 9343
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.0001723469387755102,
+      "loss": 0.1339,
+      "step": 9344
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.0001723316326530612,
+      "loss": 0.1811,
+      "step": 9345
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00017231632653061222,
+      "loss": 0.2537,
+      "step": 9346
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00017230102040816325,
+      "loss": 0.2588,
+      "step": 9347
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00017228571428571428,
+      "loss": 0.1613,
+      "step": 9348
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.0001722704081632653,
+      "loss": 0.0671,
+      "step": 9349
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.0001722551020408163,
+      "loss": 0.0983,
+      "step": 9350
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00017223979591836735,
+      "loss": 0.0583,
+      "step": 9351
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00017222448979591835,
+      "loss": 0.2057,
+      "step": 9352
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00017220918367346936,
+      "loss": 0.0763,
+      "step": 9353
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.0001721938775510204,
+      "loss": 0.1051,
+      "step": 9354
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00017217857142857142,
+      "loss": 0.2982,
+      "step": 9355
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00017216326530612242,
+      "loss": 0.0649,
+      "step": 9356
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00017214795918367345,
+      "loss": 0.1156,
+      "step": 9357
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017214795918367345,
+      "loss": 0.1835,
+      "step": 9358
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017213265306122448,
+      "loss": 0.0753,
+      "step": 9359
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.0001721173469387755,
+      "loss": 0.0762,
+      "step": 9360
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017210204081632651,
+      "loss": 0.1499,
+      "step": 9361
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017208673469387752,
+      "loss": 0.1378,
+      "step": 9362
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017207142857142858,
+      "loss": 0.1466,
+      "step": 9363
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017205612244897958,
+      "loss": 0.1442,
+      "step": 9364
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017204081632653058,
+      "loss": 0.0884,
+      "step": 9365
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.0001720255102040816,
+      "loss": 0.1476,
+      "step": 9366
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017201020408163264,
+      "loss": 0.1391,
+      "step": 9367
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017199489795918367,
+      "loss": 0.1162,
+      "step": 9368
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017197959183673468,
+      "loss": 0.097,
+      "step": 9369
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017196428571428568,
+      "loss": 0.3951,
+      "step": 9370
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017194897959183674,
+      "loss": 0.1627,
+      "step": 9371
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017193367346938774,
+      "loss": 0.111,
+      "step": 9372
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017191836734693875,
+      "loss": 0.1296,
+      "step": 9373
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017190306122448978,
+      "loss": 0.0802,
+      "step": 9374
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.0001718877551020408,
+      "loss": 0.1169,
+      "step": 9375
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017187244897959184,
+      "loss": 0.0848,
+      "step": 9376
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017185714285714284,
+      "loss": 0.1786,
+      "step": 9377
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017184183673469384,
+      "loss": 0.1625,
+      "step": 9378
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.0001718265306122449,
+      "loss": 0.1632,
+      "step": 9379
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.0001718112244897959,
+      "loss": 0.1048,
+      "step": 9380
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.0001717959183673469,
+      "loss": 0.1299,
+      "step": 9381
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00017178061224489794,
+      "loss": 0.0991,
+      "step": 9382
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00017176530612244897,
+      "loss": 0.1539,
+      "step": 9383
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00017175,
+      "loss": 0.0918,
+      "step": 9384
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.000171734693877551,
+      "loss": 0.1442,
+      "step": 9385
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.000171719387755102,
+      "loss": 0.0851,
+      "step": 9386
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00017170408163265307,
+      "loss": 0.1511,
+      "step": 9387
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00017168877551020407,
+      "loss": 0.0721,
+      "step": 9388
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017167346938775507,
+      "loss": 0.2076,
+      "step": 9389
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.0001716581632653061,
+      "loss": 0.112,
+      "step": 9390
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017164285714285713,
+      "loss": 0.1408,
+      "step": 9391
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017162755102040814,
+      "loss": 0.116,
+      "step": 9392
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017161224489795917,
+      "loss": 0.0862,
+      "step": 9393
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017159693877551017,
+      "loss": 0.0608,
+      "step": 9394
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017158163265306123,
+      "loss": 0.1606,
+      "step": 9395
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017156632653061223,
+      "loss": 0.1686,
+      "step": 9396
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017155102040816324,
+      "loss": 0.0895,
+      "step": 9397
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00017153571428571427,
+      "loss": 0.1558,
+      "step": 9398
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.0001715204081632653,
+      "loss": 0.0696,
+      "step": 9399
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.0001715051020408163,
+      "loss": 0.0727,
+      "step": 9400
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00017148979591836733,
+      "loss": 0.1067,
+      "step": 9401
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00017147448979591836,
+      "loss": 0.2366,
+      "step": 9402
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.0001714591836734694,
+      "loss": 0.0933,
+      "step": 9403
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.0001714438775510204,
+      "loss": 0.1169,
+      "step": 9404
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.0001714285714285714,
+      "loss": 0.1119,
+      "step": 9405
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00017141326530612246,
+      "loss": 0.0668,
+      "step": 9406
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00017139795918367346,
+      "loss": 0.0479,
+      "step": 9407
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00017138265306122446,
+      "loss": 0.1614,
+      "step": 9408
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.0001713673469387755,
+      "loss": 0.2333,
+      "step": 9409
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017135204081632652,
+      "loss": 0.0741,
+      "step": 9410
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017133673469387755,
+      "loss": 0.0667,
+      "step": 9411
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017132142857142856,
+      "loss": 0.1973,
+      "step": 9412
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017130612244897956,
+      "loss": 0.0984,
+      "step": 9413
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017129081632653062,
+      "loss": 0.1144,
+      "step": 9414
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017127551020408162,
+      "loss": 0.2589,
+      "step": 9415
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017126020408163263,
+      "loss": 0.1838,
+      "step": 9416
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017124489795918366,
+      "loss": 0.145,
+      "step": 9417
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.0001712295918367347,
+      "loss": 0.1338,
+      "step": 9418
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00017121428571428572,
+      "loss": 0.0892,
+      "step": 9419
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00017119897959183672,
+      "loss": 0.1011,
+      "step": 9420
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00017118367346938772,
+      "loss": 0.0903,
+      "step": 9421
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00017116836734693878,
+      "loss": 0.1455,
+      "step": 9422
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00017115306122448979,
+      "loss": 0.1759,
+      "step": 9423
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.0001711377551020408,
+      "loss": 0.0607,
+      "step": 9424
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00017112244897959182,
+      "loss": 0.0959,
+      "step": 9425
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00017110714285714285,
+      "loss": 0.0682,
+      "step": 9426
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00017109183673469385,
+      "loss": 0.1522,
+      "step": 9427
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00017107653061224488,
+      "loss": 0.2332,
+      "step": 9428
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.0001710612244897959,
+      "loss": 0.1635,
+      "step": 9429
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00017104591836734694,
+      "loss": 0.088,
+      "step": 9430
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00017103061224489795,
+      "loss": 0.0578,
+      "step": 9431
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00017101530612244895,
+      "loss": 0.1835,
+      "step": 9432
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00017099999999999998,
+      "loss": 0.1689,
+      "step": 9433
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.000170984693877551,
+      "loss": 0.1736,
+      "step": 9434
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00017096938775510202,
+      "loss": 0.0789,
+      "step": 9435
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00017095408163265305,
+      "loss": 0.1992,
+      "step": 9436
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00017093877551020405,
+      "loss": 0.1844,
+      "step": 9437
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.0001709234693877551,
+      "loss": 0.1106,
+      "step": 9438
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.0001709081632653061,
+      "loss": 0.2592,
+      "step": 9439
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017089285714285711,
+      "loss": 0.1316,
+      "step": 9440
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017087755102040815,
+      "loss": 0.0681,
+      "step": 9441
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017086224489795918,
+      "loss": 0.0631,
+      "step": 9442
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017084693877551018,
+      "loss": 0.0724,
+      "step": 9443
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.0001708316326530612,
+      "loss": 0.1062,
+      "step": 9444
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017081632653061224,
+      "loss": 0.1574,
+      "step": 9445
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017080102040816327,
+      "loss": 0.1581,
+      "step": 9446
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017078571428571427,
+      "loss": 0.1032,
+      "step": 9447
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017077040816326528,
+      "loss": 0.1635,
+      "step": 9448
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017075510204081634,
+      "loss": 0.0795,
+      "step": 9449
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00017073979591836734,
+      "loss": 0.1398,
+      "step": 9450
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00017072448979591834,
+      "loss": 0.0996,
+      "step": 9451
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00017070918367346937,
+      "loss": 0.2176,
+      "step": 9452
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.0001706938775510204,
+      "loss": 0.1969,
+      "step": 9453
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.0001706785714285714,
+      "loss": 0.0654,
+      "step": 9454
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00017066326530612244,
+      "loss": 0.1194,
+      "step": 9455
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00017064795918367344,
+      "loss": 0.1085,
+      "step": 9456
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.0001706326530612245,
+      "loss": 0.0989,
+      "step": 9457
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.0001706173469387755,
+      "loss": 0.2249,
+      "step": 9458
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.0001706020408163265,
+      "loss": 0.0993,
+      "step": 9459
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00017058673469387754,
+      "loss": 0.1215,
+      "step": 9460
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.00017057142857142857,
+      "loss": 0.1203,
+      "step": 9461
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.00017055612244897957,
+      "loss": 0.0791,
+      "step": 9462
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.0001705408163265306,
+      "loss": 0.123,
+      "step": 9463
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.0001705255102040816,
+      "loss": 0.1162,
+      "step": 9464
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.00017051020408163266,
+      "loss": 0.1295,
+      "step": 9465
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.00017049489795918367,
+      "loss": 0.1447,
+      "step": 9466
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.00017047959183673467,
+      "loss": 0.1263,
+      "step": 9467
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.0001704642857142857,
+      "loss": 0.1274,
+      "step": 9468
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.00017044897959183673,
+      "loss": 0.0761,
+      "step": 9469
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.00017043367346938773,
+      "loss": 0.1577,
+      "step": 9470
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00017041836734693876,
+      "loss": 0.2013,
+      "step": 9471
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00017040306122448977,
+      "loss": 0.0808,
+      "step": 9472
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00017038775510204082,
+      "loss": 0.0848,
+      "step": 9473
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00017037244897959183,
+      "loss": 0.0894,
+      "step": 9474
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00017035714285714283,
+      "loss": 0.0856,
+      "step": 9475
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00017034183673469384,
+      "loss": 0.0747,
+      "step": 9476
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.0001703265306122449,
+      "loss": 0.1427,
+      "step": 9477
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.0001703112244897959,
+      "loss": 0.2665,
+      "step": 9478
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00017029591836734693,
+      "loss": 0.0939,
+      "step": 9479
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00017028061224489793,
+      "loss": 0.2025,
+      "step": 9480
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.000170265306122449,
+      "loss": 0.1098,
+      "step": 9481
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00017025,
+      "loss": 0.076,
+      "step": 9482
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.000170234693877551,
+      "loss": 0.1525,
+      "step": 9483
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.000170219387755102,
+      "loss": 0.1045,
+      "step": 9484
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00017020408163265306,
+      "loss": 0.0724,
+      "step": 9485
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00017018877551020406,
+      "loss": 0.1474,
+      "step": 9486
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.0001701734693877551,
+      "loss": 0.066,
+      "step": 9487
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.0001701581632653061,
+      "loss": 0.1107,
+      "step": 9488
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00017014285714285712,
+      "loss": 0.0752,
+      "step": 9489
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00017012755102040815,
+      "loss": 0.1447,
+      "step": 9490
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00017011224489795916,
+      "loss": 0.0664,
+      "step": 9491
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00017009693877551022,
+      "loss": 0.0803,
+      "step": 9492
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00017008163265306122,
+      "loss": 0.1202,
+      "step": 9493
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00017006632653061222,
+      "loss": 0.1072,
+      "step": 9494
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00017005102040816325,
+      "loss": 0.1747,
+      "step": 9495
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00017003571428571428,
+      "loss": 0.2226,
+      "step": 9496
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.0001700204081632653,
+      "loss": 0.112,
+      "step": 9497
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00017000510204081632,
+      "loss": 0.1083,
+      "step": 9498
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00016998979591836732,
+      "loss": 0.0812,
+      "step": 9499
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00016997448979591838,
+      "loss": 0.1148,
+      "step": 9500
+    },
+    {
+      "epoch": 9.22,
+      "eval_cer": 0.2628515826604919,
+      "eval_loss": 1.0206079483032227,
+      "eval_runtime": 234.3962,
+      "eval_samples_per_second": 37.603,
+      "eval_steps_per_second": 4.701,
+      "eval_wer": 0.67245053774072,
+      "step": 9500
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00016995918367346938,
+      "loss": 0.131,
+      "step": 9501
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016994387755102039,
+      "loss": 0.1864,
+      "step": 9502
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016992857142857142,
+      "loss": 0.1189,
+      "step": 9503
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016991326530612245,
+      "loss": 0.0807,
+      "step": 9504
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016989795918367345,
+      "loss": 0.1213,
+      "step": 9505
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016988265306122448,
+      "loss": 0.1167,
+      "step": 9506
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016986734693877548,
+      "loss": 0.1296,
+      "step": 9507
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016985204081632654,
+      "loss": 0.1648,
+      "step": 9508
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016983673469387754,
+      "loss": 0.1438,
+      "step": 9509
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016982142857142855,
+      "loss": 0.0759,
+      "step": 9510
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016980612244897955,
+      "loss": 0.1404,
+      "step": 9511
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.0001697908163265306,
+      "loss": 0.1753,
+      "step": 9512
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.0001697755102040816,
+      "loss": 0.1232,
+      "step": 9513
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00016976020408163264,
+      "loss": 0.1501,
+      "step": 9514
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00016974489795918365,
+      "loss": 0.1495,
+      "step": 9515
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00016972959183673468,
+      "loss": 0.0739,
+      "step": 9516
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.0001697142857142857,
+      "loss": 0.137,
+      "step": 9517
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.0001696989795918367,
+      "loss": 0.0643,
+      "step": 9518
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00016968367346938771,
+      "loss": 0.1126,
+      "step": 9519
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00016966836734693877,
+      "loss": 0.1591,
+      "step": 9520
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00016965306122448978,
+      "loss": 0.2223,
+      "step": 9521
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.0001696377551020408,
+      "loss": 0.0925,
+      "step": 9522
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.0001696224489795918,
+      "loss": 0.1309,
+      "step": 9523
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016960714285714284,
+      "loss": 0.0933,
+      "step": 9524
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016959183673469387,
+      "loss": 0.0861,
+      "step": 9525
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016957653061224487,
+      "loss": 0.1166,
+      "step": 9526
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016956122448979588,
+      "loss": 0.1597,
+      "step": 9527
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016954591836734694,
+      "loss": 0.134,
+      "step": 9528
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016953061224489794,
+      "loss": 0.0605,
+      "step": 9529
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016951530612244897,
+      "loss": 0.0771,
+      "step": 9530
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016949999999999997,
+      "loss": 0.1391,
+      "step": 9531
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.000169484693877551,
+      "loss": 0.1482,
+      "step": 9532
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00016946938775510203,
+      "loss": 0.1874,
+      "step": 9533
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00016945408163265304,
+      "loss": 0.12,
+      "step": 9534
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.0001694387755102041,
+      "loss": 0.1201,
+      "step": 9535
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.0001694234693877551,
+      "loss": 0.1518,
+      "step": 9536
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.0001694081632653061,
+      "loss": 0.12,
+      "step": 9537
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00016939285714285713,
+      "loss": 0.1364,
+      "step": 9538
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00016937755102040816,
+      "loss": 0.132,
+      "step": 9539
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00016936224489795917,
+      "loss": 0.1418,
+      "step": 9540
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.0001693469387755102,
+      "loss": 0.0827,
+      "step": 9541
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.0001693316326530612,
+      "loss": 0.1405,
+      "step": 9542
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016931632653061226,
+      "loss": 0.0789,
+      "step": 9543
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016930102040816326,
+      "loss": 0.112,
+      "step": 9544
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016928571428571427,
+      "loss": 0.1123,
+      "step": 9545
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016927040816326527,
+      "loss": 0.1673,
+      "step": 9546
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016925510204081633,
+      "loss": 0.1378,
+      "step": 9547
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016923979591836733,
+      "loss": 0.1093,
+      "step": 9548
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016922448979591836,
+      "loss": 0.1481,
+      "step": 9549
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016920918367346936,
+      "loss": 0.1009,
+      "step": 9550
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.0001691938775510204,
+      "loss": 0.1177,
+      "step": 9551
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016917857142857142,
+      "loss": 0.1354,
+      "step": 9552
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016916326530612243,
+      "loss": 0.0877,
+      "step": 9553
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00016914795918367343,
+      "loss": 0.1146,
+      "step": 9554
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0001691326530612245,
+      "loss": 0.1919,
+      "step": 9555
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0001691173469387755,
+      "loss": 0.1029,
+      "step": 9556
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00016910204081632652,
+      "loss": 0.1303,
+      "step": 9557
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00016908673469387753,
+      "loss": 0.215,
+      "step": 9558
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00016907142857142856,
+      "loss": 0.1803,
+      "step": 9559
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0001690561224489796,
+      "loss": 0.0651,
+      "step": 9560
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0001690408163265306,
+      "loss": 0.2025,
+      "step": 9561
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0001690255102040816,
+      "loss": 0.1382,
+      "step": 9562
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00016901020408163265,
+      "loss": 0.1181,
+      "step": 9563
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016899489795918366,
+      "loss": 0.206,
+      "step": 9564
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016897959183673469,
+      "loss": 0.1247,
+      "step": 9565
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.0001689642857142857,
+      "loss": 0.1789,
+      "step": 9566
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016894897959183672,
+      "loss": 0.0809,
+      "step": 9567
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016893367346938775,
+      "loss": 0.1615,
+      "step": 9568
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016891836734693875,
+      "loss": 0.0678,
+      "step": 9569
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016890306122448976,
+      "loss": 0.1679,
+      "step": 9570
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016888775510204082,
+      "loss": 0.1624,
+      "step": 9571
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016887244897959182,
+      "loss": 0.16,
+      "step": 9572
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016885714285714282,
+      "loss": 0.1295,
+      "step": 9573
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016884183673469385,
+      "loss": 0.0821,
+      "step": 9574
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016882653061224488,
+      "loss": 0.0703,
+      "step": 9575
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016881122448979591,
+      "loss": 0.1608,
+      "step": 9576
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016879591836734692,
+      "loss": 0.1702,
+      "step": 9577
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016878061224489792,
+      "loss": 0.1296,
+      "step": 9578
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016876530612244898,
+      "loss": 0.0839,
+      "step": 9579
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016874999999999998,
+      "loss": 0.2008,
+      "step": 9580
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016873469387755099,
+      "loss": 0.1778,
+      "step": 9581
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016871938775510204,
+      "loss": 0.12,
+      "step": 9582
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016870408163265305,
+      "loss": 0.1695,
+      "step": 9583
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016868877551020408,
+      "loss": 0.0942,
+      "step": 9584
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016867346938775508,
+      "loss": 0.2886,
+      "step": 9585
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.0001686581632653061,
+      "loss": 0.1224,
+      "step": 9586
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016864285714285714,
+      "loss": 0.0787,
+      "step": 9587
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016862755102040814,
+      "loss": 0.114,
+      "step": 9588
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016861224489795915,
+      "loss": 0.2125,
+      "step": 9589
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.0001685969387755102,
+      "loss": 0.229,
+      "step": 9590
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.0001685816326530612,
+      "loss": 0.0821,
+      "step": 9591
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016856632653061224,
+      "loss": 0.1008,
+      "step": 9592
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016855102040816324,
+      "loss": 0.1348,
+      "step": 9593
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016853571428571427,
+      "loss": 0.1451,
+      "step": 9594
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.0001685204081632653,
+      "loss": 0.1172,
+      "step": 9595
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.0001685051020408163,
+      "loss": 0.1223,
+      "step": 9596
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.0001684897959183673,
+      "loss": 0.1712,
+      "step": 9597
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00016847448979591837,
+      "loss": 0.0815,
+      "step": 9598
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00016845918367346937,
+      "loss": 0.1154,
+      "step": 9599
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.0001684438775510204,
+      "loss": 0.0863,
+      "step": 9600
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.0001684285714285714,
+      "loss": 0.1646,
+      "step": 9601
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00016841326530612244,
+      "loss": 0.2373,
+      "step": 9602
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00016839795918367347,
+      "loss": 0.1221,
+      "step": 9603
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00016838265306122447,
+      "loss": 0.0833,
+      "step": 9604
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016836734693877547,
+      "loss": 0.1019,
+      "step": 9605
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016835204081632653,
+      "loss": 0.1195,
+      "step": 9606
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016833673469387754,
+      "loss": 0.1208,
+      "step": 9607
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016832142857142854,
+      "loss": 0.1488,
+      "step": 9608
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016830612244897957,
+      "loss": 0.1374,
+      "step": 9609
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.0001682908163265306,
+      "loss": 0.1287,
+      "step": 9610
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016827551020408163,
+      "loss": 0.1503,
+      "step": 9611
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016826020408163263,
+      "loss": 0.0639,
+      "step": 9612
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016824489795918364,
+      "loss": 0.066,
+      "step": 9613
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.0001682295918367347,
+      "loss": 0.1613,
+      "step": 9614
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.0001682142857142857,
+      "loss": 0.1737,
+      "step": 9615
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.0001681989795918367,
+      "loss": 0.1783,
+      "step": 9616
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00016818367346938773,
+      "loss": 0.2274,
+      "step": 9617
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00016816836734693876,
+      "loss": 0.0989,
+      "step": 9618
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.0001681530612244898,
+      "loss": 0.1352,
+      "step": 9619
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.0001681377551020408,
+      "loss": 0.1451,
+      "step": 9620
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.0001681224489795918,
+      "loss": 0.2206,
+      "step": 9621
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00016810714285714286,
+      "loss": 0.1933,
+      "step": 9622
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00016809183673469386,
+      "loss": 0.122,
+      "step": 9623
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00016807653061224487,
+      "loss": 0.1807,
+      "step": 9624
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.0001680612244897959,
+      "loss": 0.0495,
+      "step": 9625
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016804591836734693,
+      "loss": 0.1503,
+      "step": 9626
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016803061224489796,
+      "loss": 0.2605,
+      "step": 9627
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016801530612244896,
+      "loss": 0.1394,
+      "step": 9628
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.000168,
+      "loss": 0.1131,
+      "step": 9629
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016798469387755102,
+      "loss": 0.1729,
+      "step": 9630
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016796938775510202,
+      "loss": 0.1071,
+      "step": 9631
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016795408163265303,
+      "loss": 0.1207,
+      "step": 9632
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016793877551020409,
+      "loss": 0.2031,
+      "step": 9633
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.0001679234693877551,
+      "loss": 0.1868,
+      "step": 9634
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.0001679081632653061,
+      "loss": 0.1762,
+      "step": 9635
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016789285714285712,
+      "loss": 0.1131,
+      "step": 9636
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016787755102040815,
+      "loss": 0.1124,
+      "step": 9637
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016786224489795918,
+      "loss": 0.1065,
+      "step": 9638
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.0001678469387755102,
+      "loss": 0.0937,
+      "step": 9639
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.0001678316326530612,
+      "loss": 0.2122,
+      "step": 9640
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016781632653061225,
+      "loss": 0.0862,
+      "step": 9641
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016780102040816325,
+      "loss": 0.1154,
+      "step": 9642
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016778571428571426,
+      "loss": 0.1119,
+      "step": 9643
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016777040816326529,
+      "loss": 0.0681,
+      "step": 9644
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016775510204081632,
+      "loss": 0.1334,
+      "step": 9645
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016773979591836735,
+      "loss": 0.1192,
+      "step": 9646
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016772448979591835,
+      "loss": 0.1554,
+      "step": 9647
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016770918367346935,
+      "loss": 0.0867,
+      "step": 9648
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.0001676938775510204,
+      "loss": 0.2016,
+      "step": 9649
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016767857142857142,
+      "loss": 0.0821,
+      "step": 9650
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016766326530612242,
+      "loss": 0.055,
+      "step": 9651
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016764795918367345,
+      "loss": 0.1639,
+      "step": 9652
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016763265306122448,
+      "loss": 0.0809,
+      "step": 9653
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.0001676173469387755,
+      "loss": 0.168,
+      "step": 9654
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016760204081632651,
+      "loss": 0.1275,
+      "step": 9655
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016758673469387752,
+      "loss": 0.077,
+      "step": 9656
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016757142857142857,
+      "loss": 0.0872,
+      "step": 9657
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016755612244897958,
+      "loss": 0.2196,
+      "step": 9658
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016754081632653058,
+      "loss": 0.1626,
+      "step": 9659
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.0001675255102040816,
+      "loss": 0.1247,
+      "step": 9660
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016751020408163264,
+      "loss": 0.1095,
+      "step": 9661
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016749489795918367,
+      "loss": 0.0536,
+      "step": 9662
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016747959183673468,
+      "loss": 0.1083,
+      "step": 9663
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016746428571428568,
+      "loss": 0.1942,
+      "step": 9664
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016744897959183674,
+      "loss": 0.0922,
+      "step": 9665
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016743367346938774,
+      "loss": 0.0752,
+      "step": 9666
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00016741836734693874,
+      "loss": 0.0555,
+      "step": 9667
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00016740306122448978,
+      "loss": 0.1234,
+      "step": 9668
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0001673877551020408,
+      "loss": 0.1183,
+      "step": 9669
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0001673724489795918,
+      "loss": 0.1673,
+      "step": 9670
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00016735714285714284,
+      "loss": 0.1542,
+      "step": 9671
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00016734183673469387,
+      "loss": 0.0978,
+      "step": 9672
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0001673265306122449,
+      "loss": 0.1057,
+      "step": 9673
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0001673112244897959,
+      "loss": 0.1563,
+      "step": 9674
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0001672959183673469,
+      "loss": 0.0932,
+      "step": 9675
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00016728061224489797,
+      "loss": 0.148,
+      "step": 9676
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016726530612244897,
+      "loss": 0.1608,
+      "step": 9677
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016724999999999997,
+      "loss": 0.1177,
+      "step": 9678
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.000167234693877551,
+      "loss": 0.1053,
+      "step": 9679
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016721938775510203,
+      "loss": 0.1504,
+      "step": 9680
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016720408163265306,
+      "loss": 0.1143,
+      "step": 9681
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016718877551020407,
+      "loss": 0.1053,
+      "step": 9682
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016717346938775507,
+      "loss": 0.1985,
+      "step": 9683
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016715816326530613,
+      "loss": 0.1103,
+      "step": 9684
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016714285714285713,
+      "loss": 0.1089,
+      "step": 9685
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016712755102040814,
+      "loss": 0.1889,
+      "step": 9686
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016711224489795917,
+      "loss": 0.0795,
+      "step": 9687
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0001670969387755102,
+      "loss": 0.1015,
+      "step": 9688
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00016708163265306123,
+      "loss": 0.2786,
+      "step": 9689
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00016706632653061223,
+      "loss": 0.1738,
+      "step": 9690
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00016705102040816323,
+      "loss": 0.0828,
+      "step": 9691
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0001670357142857143,
+      "loss": 0.1322,
+      "step": 9692
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0001670204081632653,
+      "loss": 0.2036,
+      "step": 9693
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0001670051020408163,
+      "loss": 0.0683,
+      "step": 9694
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00016698979591836733,
+      "loss": 0.1956,
+      "step": 9695
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00016697448979591836,
+      "loss": 0.1863,
+      "step": 9696
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00016695918367346936,
+      "loss": 0.1329,
+      "step": 9697
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.0001669438775510204,
+      "loss": 0.0872,
+      "step": 9698
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.0001669285714285714,
+      "loss": 0.1617,
+      "step": 9699
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00016691326530612245,
+      "loss": 0.1019,
+      "step": 9700
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00016689795918367346,
+      "loss": 0.084,
+      "step": 9701
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00016688265306122446,
+      "loss": 0.1448,
+      "step": 9702
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.0001668673469387755,
+      "loss": 0.1205,
+      "step": 9703
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00016685204081632652,
+      "loss": 0.1102,
+      "step": 9704
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00016683673469387753,
+      "loss": 0.0945,
+      "step": 9705
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00016682142857142856,
+      "loss": 0.1254,
+      "step": 9706
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00016680612244897956,
+      "loss": 0.0709,
+      "step": 9707
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016679081632653062,
+      "loss": 0.2328,
+      "step": 9708
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016677551020408162,
+      "loss": 0.128,
+      "step": 9709
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016676020408163262,
+      "loss": 0.1042,
+      "step": 9710
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016674489795918366,
+      "loss": 0.1575,
+      "step": 9711
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016672959183673469,
+      "loss": 0.1094,
+      "step": 9712
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.0001667142857142857,
+      "loss": 0.0645,
+      "step": 9713
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016669897959183672,
+      "loss": 0.1661,
+      "step": 9714
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016668367346938772,
+      "loss": 0.2099,
+      "step": 9715
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016666836734693878,
+      "loss": 0.1733,
+      "step": 9716
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016665306122448978,
+      "loss": 0.1803,
+      "step": 9717
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.0001666377551020408,
+      "loss": 0.1153,
+      "step": 9718
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00016662244897959185,
+      "loss": 0.0469,
+      "step": 9719
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00016660714285714285,
+      "loss": 0.1739,
+      "step": 9720
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00016659183673469385,
+      "loss": 0.0998,
+      "step": 9721
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00016657653061224488,
+      "loss": 0.1146,
+      "step": 9722
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.0001665612244897959,
+      "loss": 0.0829,
+      "step": 9723
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00016654591836734694,
+      "loss": 0.1024,
+      "step": 9724
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00016653061224489795,
+      "loss": 0.0841,
+      "step": 9725
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00016651530612244895,
+      "loss": 0.07,
+      "step": 9726
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.0001665,
+      "loss": 0.238,
+      "step": 9727
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.000166484693877551,
+      "loss": 0.1694,
+      "step": 9728
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016646938775510202,
+      "loss": 0.1028,
+      "step": 9729
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016645408163265305,
+      "loss": 0.1062,
+      "step": 9730
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016643877551020408,
+      "loss": 0.1097,
+      "step": 9731
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016642346938775508,
+      "loss": 0.1364,
+      "step": 9732
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.0001664081632653061,
+      "loss": 0.2786,
+      "step": 9733
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016639285714285711,
+      "loss": 0.1625,
+      "step": 9734
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016637755102040817,
+      "loss": 0.1583,
+      "step": 9735
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016636224489795917,
+      "loss": 0.1413,
+      "step": 9736
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016634693877551018,
+      "loss": 0.0631,
+      "step": 9737
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.0001663316326530612,
+      "loss": 0.103,
+      "step": 9738
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016631632653061224,
+      "loss": 0.2235,
+      "step": 9739
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016630102040816324,
+      "loss": 0.1417,
+      "step": 9740
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016628571428571427,
+      "loss": 0.0958,
+      "step": 9741
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016627040816326528,
+      "loss": 0.1631,
+      "step": 9742
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016625510204081633,
+      "loss": 0.1125,
+      "step": 9743
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016623979591836734,
+      "loss": 0.0756,
+      "step": 9744
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016622448979591834,
+      "loss": 0.1675,
+      "step": 9745
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016620918367346937,
+      "loss": 0.1761,
+      "step": 9746
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.0001661938775510204,
+      "loss": 0.0881,
+      "step": 9747
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.0001661785714285714,
+      "loss": 0.0941,
+      "step": 9748
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00016616326530612244,
+      "loss": 0.1323,
+      "step": 9749
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00016614795918367344,
+      "loss": 0.1011,
+      "step": 9750
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0001661326530612245,
+      "loss": 0.1114,
+      "step": 9751
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0001661173469387755,
+      "loss": 0.1547,
+      "step": 9752
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0001661020408163265,
+      "loss": 0.1548,
+      "step": 9753
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0001660867346938775,
+      "loss": 0.1463,
+      "step": 9754
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00016607142857142857,
+      "loss": 0.17,
+      "step": 9755
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00016605612244897957,
+      "loss": 0.0803,
+      "step": 9756
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0001660408163265306,
+      "loss": 0.0731,
+      "step": 9757
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0001660255102040816,
+      "loss": 0.1943,
+      "step": 9758
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00016601020408163266,
+      "loss": 0.0736,
+      "step": 9759
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016599489795918366,
+      "loss": 0.1216,
+      "step": 9760
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016597959183673467,
+      "loss": 0.1576,
+      "step": 9761
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016596428571428573,
+      "loss": 0.2197,
+      "step": 9762
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016594897959183673,
+      "loss": 0.111,
+      "step": 9763
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016593367346938773,
+      "loss": 0.2042,
+      "step": 9764
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016591836734693876,
+      "loss": 0.1839,
+      "step": 9765
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.0001659030612244898,
+      "loss": 0.0902,
+      "step": 9766
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.0001658877551020408,
+      "loss": 0.154,
+      "step": 9767
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016587244897959183,
+      "loss": 0.106,
+      "step": 9768
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016585714285714283,
+      "loss": 0.123,
+      "step": 9769
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.0001658418367346939,
+      "loss": 0.109,
+      "step": 9770
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.0001658265306122449,
+      "loss": 0.1418,
+      "step": 9771
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.0001658112244897959,
+      "loss": 0.0746,
+      "step": 9772
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.00016579591836734693,
+      "loss": 0.0933,
+      "step": 9773
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.00016578061224489796,
+      "loss": 0.0908,
+      "step": 9774
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.00016576530612244896,
+      "loss": 0.1298,
+      "step": 9775
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.00016575,
+      "loss": 0.1176,
+      "step": 9776
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.000165734693877551,
+      "loss": 0.1351,
+      "step": 9777
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.00016571938775510205,
+      "loss": 0.1253,
+      "step": 9778
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.00016570408163265305,
+      "loss": 0.132,
+      "step": 9779
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016568877551020406,
+      "loss": 0.098,
+      "step": 9780
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.0001656734693877551,
+      "loss": 0.1412,
+      "step": 9781
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016565816326530612,
+      "loss": 0.0798,
+      "step": 9782
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016564285714285712,
+      "loss": 0.1493,
+      "step": 9783
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016562755102040815,
+      "loss": 0.0953,
+      "step": 9784
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016561224489795916,
+      "loss": 0.0818,
+      "step": 9785
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016559693877551021,
+      "loss": 0.1088,
+      "step": 9786
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016558163265306122,
+      "loss": 0.0741,
+      "step": 9787
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016556632653061222,
+      "loss": 0.1071,
+      "step": 9788
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016555102040816322,
+      "loss": 0.1342,
+      "step": 9789
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016553571428571428,
+      "loss": 0.1245,
+      "step": 9790
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016552040816326529,
+      "loss": 0.1093,
+      "step": 9791
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016550510204081632,
+      "loss": 0.1323,
+      "step": 9792
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016548979591836732,
+      "loss": 0.1432,
+      "step": 9793
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016547448979591835,
+      "loss": 0.0981,
+      "step": 9794
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016545918367346938,
+      "loss": 0.1064,
+      "step": 9795
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016544387755102038,
+      "loss": 0.2231,
+      "step": 9796
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.0001654285714285714,
+      "loss": 0.083,
+      "step": 9797
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016541326530612245,
+      "loss": 0.1244,
+      "step": 9798
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016539795918367345,
+      "loss": 0.142,
+      "step": 9799
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016538265306122448,
+      "loss": 0.1487,
+      "step": 9800
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00016536734693877548,
+      "loss": 0.1777,
+      "step": 9801
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.0001653520408163265,
+      "loss": 0.1848,
+      "step": 9802
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00016533673469387754,
+      "loss": 0.1194,
+      "step": 9803
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00016532142857142855,
+      "loss": 0.0669,
+      "step": 9804
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00016530612244897955,
+      "loss": 0.1389,
+      "step": 9805
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.0001652908163265306,
+      "loss": 0.0871,
+      "step": 9806
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.0001652755102040816,
+      "loss": 0.1831,
+      "step": 9807
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00016526020408163264,
+      "loss": 0.2081,
+      "step": 9808
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00016524489795918367,
+      "loss": 0.1361,
+      "step": 9809
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00016522959183673468,
+      "loss": 0.1347,
+      "step": 9810
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.0001652142857142857,
+      "loss": 0.0831,
+      "step": 9811
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.0001651989795918367,
+      "loss": 0.0686,
+      "step": 9812
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016518367346938777,
+      "loss": 0.1165,
+      "step": 9813
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016516836734693877,
+      "loss": 0.1955,
+      "step": 9814
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016515306122448977,
+      "loss": 0.1293,
+      "step": 9815
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016513775510204078,
+      "loss": 0.1055,
+      "step": 9816
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016512244897959184,
+      "loss": 0.1461,
+      "step": 9817
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016510714285714284,
+      "loss": 0.1058,
+      "step": 9818
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016509183673469387,
+      "loss": 0.1394,
+      "step": 9819
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016507653061224487,
+      "loss": 0.1611,
+      "step": 9820
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00016506122448979593,
+      "loss": 0.1942,
+      "step": 9821
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00016504591836734693,
+      "loss": 0.1137,
+      "step": 9822
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00016503061224489794,
+      "loss": 0.1047,
+      "step": 9823
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00016501530612244894,
+      "loss": 0.114,
+      "step": 9824
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.000165,
+      "loss": 0.0708,
+      "step": 9825
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.000164984693877551,
+      "loss": 0.0752,
+      "step": 9826
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00016496938775510203,
+      "loss": 0.1275,
+      "step": 9827
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00016495408163265304,
+      "loss": 0.1148,
+      "step": 9828
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00016493877551020407,
+      "loss": 0.0682,
+      "step": 9829
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.0001649234693877551,
+      "loss": 0.151,
+      "step": 9830
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.0001649081632653061,
+      "loss": 0.0947,
+      "step": 9831
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.0001648928571428571,
+      "loss": 0.1251,
+      "step": 9832
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00016487755102040816,
+      "loss": 0.2084,
+      "step": 9833
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00016486224489795917,
+      "loss": 0.1505,
+      "step": 9834
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.0001648469387755102,
+      "loss": 0.0661,
+      "step": 9835
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.0001648316326530612,
+      "loss": 0.0834,
+      "step": 9836
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00016481632653061223,
+      "loss": 0.1585,
+      "step": 9837
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00016480102040816326,
+      "loss": 0.1963,
+      "step": 9838
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00016478571428571426,
+      "loss": 0.1406,
+      "step": 9839
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00016477040816326527,
+      "loss": 0.1065,
+      "step": 9840
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00016475510204081633,
+      "loss": 0.1317,
+      "step": 9841
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.00016473979591836733,
+      "loss": 0.0823,
+      "step": 9842
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.00016472448979591836,
+      "loss": 0.1307,
+      "step": 9843
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.00016470918367346936,
+      "loss": 0.0863,
+      "step": 9844
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.0001646938775510204,
+      "loss": 0.1487,
+      "step": 9845
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.00016467857142857142,
+      "loss": 0.2156,
+      "step": 9846
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.00016466326530612243,
+      "loss": 0.1009,
+      "step": 9847
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.00016464795918367343,
+      "loss": 0.0741,
+      "step": 9848
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.0001646326530612245,
+      "loss": 0.0825,
+      "step": 9849
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.0001646173469387755,
+      "loss": 0.1685,
+      "step": 9850
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.0001646020408163265,
+      "loss": 0.2056,
+      "step": 9851
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016458673469387753,
+      "loss": 0.1248,
+      "step": 9852
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016457142857142856,
+      "loss": 0.2257,
+      "step": 9853
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.0001645561224489796,
+      "loss": 0.0748,
+      "step": 9854
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.0001645408163265306,
+      "loss": 0.1593,
+      "step": 9855
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016452551020408162,
+      "loss": 0.0593,
+      "step": 9856
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016451020408163265,
+      "loss": 0.0745,
+      "step": 9857
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016449489795918365,
+      "loss": 0.0734,
+      "step": 9858
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016447959183673466,
+      "loss": 0.1251,
+      "step": 9859
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016446428571428572,
+      "loss": 0.1115,
+      "step": 9860
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016444897959183672,
+      "loss": 0.0565,
+      "step": 9861
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00016443367346938775,
+      "loss": 0.0695,
+      "step": 9862
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016441836734693875,
+      "loss": 0.0553,
+      "step": 9863
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016440306122448978,
+      "loss": 0.183,
+      "step": 9864
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016438775510204081,
+      "loss": 0.1071,
+      "step": 9865
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016437244897959182,
+      "loss": 0.1661,
+      "step": 9866
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016435714285714282,
+      "loss": 0.0785,
+      "step": 9867
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016434183673469388,
+      "loss": 0.1463,
+      "step": 9868
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016432653061224488,
+      "loss": 0.1099,
+      "step": 9869
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.0001643112244897959,
+      "loss": 0.1706,
+      "step": 9870
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016429591836734692,
+      "loss": 0.1576,
+      "step": 9871
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016428061224489795,
+      "loss": 0.1099,
+      "step": 9872
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016426530612244898,
+      "loss": 0.1662,
+      "step": 9873
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016424999999999998,
+      "loss": 0.1066,
+      "step": 9874
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016423469387755098,
+      "loss": 0.101,
+      "step": 9875
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016421938775510204,
+      "loss": 0.0941,
+      "step": 9876
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016420408163265305,
+      "loss": 0.1295,
+      "step": 9877
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016418877551020408,
+      "loss": 0.2136,
+      "step": 9878
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016417346938775508,
+      "loss": 0.0852,
+      "step": 9879
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.0001641581632653061,
+      "loss": 0.0776,
+      "step": 9880
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016414285714285714,
+      "loss": 0.0673,
+      "step": 9881
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016412755102040814,
+      "loss": 0.0867,
+      "step": 9882
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00016411224489795915,
+      "loss": 0.2903,
+      "step": 9883
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0001640969387755102,
+      "loss": 0.1228,
+      "step": 9884
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0001640816326530612,
+      "loss": 0.1344,
+      "step": 9885
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0001640663265306122,
+      "loss": 0.1994,
+      "step": 9886
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00016405102040816324,
+      "loss": 0.0702,
+      "step": 9887
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00016403571428571427,
+      "loss": 0.0898,
+      "step": 9888
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0001640204081632653,
+      "loss": 0.2651,
+      "step": 9889
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0001640051020408163,
+      "loss": 0.1626,
+      "step": 9890
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0001639897959183673,
+      "loss": 0.0577,
+      "step": 9891
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00016397448979591837,
+      "loss": 0.0966,
+      "step": 9892
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00016395918367346937,
+      "loss": 0.2218,
+      "step": 9893
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00016394387755102037,
+      "loss": 0.1136,
+      "step": 9894
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.0001639285714285714,
+      "loss": 0.2418,
+      "step": 9895
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00016391326530612244,
+      "loss": 0.2237,
+      "step": 9896
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00016389795918367347,
+      "loss": 0.117,
+      "step": 9897
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00016388265306122447,
+      "loss": 0.0979,
+      "step": 9898
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.0001638673469387755,
+      "loss": 0.0682,
+      "step": 9899
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00016385204081632653,
+      "loss": 0.0667,
+      "step": 9900
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00016383673469387753,
+      "loss": 0.1199,
+      "step": 9901
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00016382142857142854,
+      "loss": 0.1904,
+      "step": 9902
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.0001638061224489796,
+      "loss": 0.0854,
+      "step": 9903
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.0001637908163265306,
+      "loss": 0.1214,
+      "step": 9904
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00016377551020408163,
+      "loss": 0.1185,
+      "step": 9905
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00016376020408163263,
+      "loss": 0.1432,
+      "step": 9906
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00016374489795918366,
+      "loss": 0.0603,
+      "step": 9907
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.0001637295918367347,
+      "loss": 0.1369,
+      "step": 9908
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.0001637142857142857,
+      "loss": 0.1276,
+      "step": 9909
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.0001636989795918367,
+      "loss": 0.0827,
+      "step": 9910
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00016368367346938776,
+      "loss": 0.076,
+      "step": 9911
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00016366836734693876,
+      "loss": 0.0728,
+      "step": 9912
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00016365306122448977,
+      "loss": 0.0661,
+      "step": 9913
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.0001636377551020408,
+      "loss": 0.1186,
+      "step": 9914
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016362244897959183,
+      "loss": 0.1832,
+      "step": 9915
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016360714285714286,
+      "loss": 0.0744,
+      "step": 9916
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016359183673469386,
+      "loss": 0.1134,
+      "step": 9917
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016357653061224486,
+      "loss": 0.1484,
+      "step": 9918
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016356122448979592,
+      "loss": 0.0987,
+      "step": 9919
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016354591836734693,
+      "loss": 0.0864,
+      "step": 9920
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016353061224489793,
+      "loss": 0.1331,
+      "step": 9921
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016351530612244896,
+      "loss": 0.1707,
+      "step": 9922
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.0001635,
+      "loss": 0.1192,
+      "step": 9923
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00016348469387755102,
+      "loss": 0.0753,
+      "step": 9924
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00016346938775510202,
+      "loss": 0.113,
+      "step": 9925
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00016345408163265303,
+      "loss": 0.1529,
+      "step": 9926
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00016343877551020408,
+      "loss": 0.2054,
+      "step": 9927
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0001634234693877551,
+      "loss": 0.167,
+      "step": 9928
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0001634081632653061,
+      "loss": 0.0949,
+      "step": 9929
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00016339285714285712,
+      "loss": 0.0831,
+      "step": 9930
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00016337755102040815,
+      "loss": 0.0914,
+      "step": 9931
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00016336224489795918,
+      "loss": 0.1263,
+      "step": 9932
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0001633469387755102,
+      "loss": 0.1806,
+      "step": 9933
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0001633316326530612,
+      "loss": 0.1,
+      "step": 9934
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016331632653061225,
+      "loss": 0.1298,
+      "step": 9935
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016330102040816325,
+      "loss": 0.1006,
+      "step": 9936
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016328571428571425,
+      "loss": 0.0616,
+      "step": 9937
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016327040816326529,
+      "loss": 0.0943,
+      "step": 9938
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016325510204081632,
+      "loss": 0.2878,
+      "step": 9939
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016323979591836735,
+      "loss": 0.1206,
+      "step": 9940
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016322448979591835,
+      "loss": 0.1138,
+      "step": 9941
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016320918367346935,
+      "loss": 0.0687,
+      "step": 9942
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.0001631938775510204,
+      "loss": 0.1113,
+      "step": 9943
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00016317857142857141,
+      "loss": 0.0973,
+      "step": 9944
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016316326530612242,
+      "loss": 0.161,
+      "step": 9945
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016314795918367348,
+      "loss": 0.1611,
+      "step": 9946
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016313265306122448,
+      "loss": 0.1938,
+      "step": 9947
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016311734693877548,
+      "loss": 0.0884,
+      "step": 9948
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.0001631020408163265,
+      "loss": 0.109,
+      "step": 9949
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016308673469387754,
+      "loss": 0.082,
+      "step": 9950
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016307142857142857,
+      "loss": 0.1228,
+      "step": 9951
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016305612244897958,
+      "loss": 0.1882,
+      "step": 9952
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016304081632653058,
+      "loss": 0.1188,
+      "step": 9953
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00016302551020408164,
+      "loss": 0.0988,
+      "step": 9954
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00016301020408163264,
+      "loss": 0.0774,
+      "step": 9955
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00016299489795918365,
+      "loss": 0.0809,
+      "step": 9956
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00016297959183673468,
+      "loss": 0.1119,
+      "step": 9957
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0001629642857142857,
+      "loss": 0.1442,
+      "step": 9958
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00016294897959183674,
+      "loss": 0.1094,
+      "step": 9959
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00016293367346938774,
+      "loss": 0.0701,
+      "step": 9960
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00016291836734693874,
+      "loss": 0.1202,
+      "step": 9961
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0001629030612244898,
+      "loss": 0.1393,
+      "step": 9962
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0001628877551020408,
+      "loss": 0.0681,
+      "step": 9963
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0001628724489795918,
+      "loss": 0.1482,
+      "step": 9964
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00016285714285714284,
+      "loss": 0.1271,
+      "step": 9965
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00016284183673469387,
+      "loss": 0.1086,
+      "step": 9966
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.0001628265306122449,
+      "loss": 0.112,
+      "step": 9967
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.0001628112244897959,
+      "loss": 0.0723,
+      "step": 9968
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.0001627959183673469,
+      "loss": 0.1273,
+      "step": 9969
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00016278061224489796,
+      "loss": 0.1939,
+      "step": 9970
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00016276530612244897,
+      "loss": 0.0772,
+      "step": 9971
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00016274999999999997,
+      "loss": 0.1154,
+      "step": 9972
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.000162734693877551,
+      "loss": 0.1295,
+      "step": 9973
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00016271938775510203,
+      "loss": 0.0778,
+      "step": 9974
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00016270408163265304,
+      "loss": 0.0643,
+      "step": 9975
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00016268877551020407,
+      "loss": 0.1744,
+      "step": 9976
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00016267346938775507,
+      "loss": 0.1912,
+      "step": 9977
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00016265816326530613,
+      "loss": 0.3573,
+      "step": 9978
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00016264285714285713,
+      "loss": 0.0514,
+      "step": 9979
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00016262755102040813,
+      "loss": 0.1113,
+      "step": 9980
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00016261224489795917,
+      "loss": 0.0487,
+      "step": 9981
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.0001625969387755102,
+      "loss": 0.0591,
+      "step": 9982
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.0001625816326530612,
+      "loss": 0.1357,
+      "step": 9983
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00016256632653061223,
+      "loss": 0.1698,
+      "step": 9984
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00016255102040816323,
+      "loss": 0.1294,
+      "step": 9985
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.0001625357142857143,
+      "loss": 0.1397,
+      "step": 9986
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.0001625204081632653,
+      "loss": 0.1281,
+      "step": 9987
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.0001625051020408163,
+      "loss": 0.0744,
+      "step": 9988
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00016248979591836736,
+      "loss": 0.1562,
+      "step": 9989
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00016247448979591836,
+      "loss": 0.2475,
+      "step": 9990
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00016245918367346936,
+      "loss": 0.1067,
+      "step": 9991
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.0001624438775510204,
+      "loss": 0.0894,
+      "step": 9992
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00016242857142857142,
+      "loss": 0.1283,
+      "step": 9993
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00016241326530612245,
+      "loss": 0.1175,
+      "step": 9994
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00016239795918367346,
+      "loss": 0.0862,
+      "step": 9995
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00016238265306122446,
+      "loss": 0.205,
+      "step": 9996
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016236734693877552,
+      "loss": 0.0866,
+      "step": 9997
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016235204081632652,
+      "loss": 0.0783,
+      "step": 9998
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016233673469387753,
+      "loss": 0.0706,
+      "step": 9999
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016232142857142856,
+      "loss": 0.0794,
+      "step": 10000
+    },
+    {
+      "epoch": 9.71,
+      "eval_cer": 0.2526173146164913,
+      "eval_loss": 0.9304736256599426,
+      "eval_runtime": 238.9224,
+      "eval_samples_per_second": 36.891,
+      "eval_steps_per_second": 4.612,
+      "eval_wer": 0.6547309099298025,
+      "step": 10000
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016230612244897959,
+      "loss": 0.1352,
+      "step": 10001
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016229081632653062,
+      "loss": 0.1419,
+      "step": 10002
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016227551020408162,
+      "loss": 0.0825,
+      "step": 10003
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016226020408163262,
+      "loss": 0.0973,
+      "step": 10004
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016224489795918368,
+      "loss": 0.1437,
+      "step": 10005
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00016222959183673468,
+      "loss": 0.0556,
+      "step": 10006
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0001622142857142857,
+      "loss": 0.1011,
+      "step": 10007
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00016219897959183672,
+      "loss": 0.1279,
+      "step": 10008
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00016218367346938775,
+      "loss": 0.1319,
+      "step": 10009
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00016216836734693875,
+      "loss": 0.1287,
+      "step": 10010
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00016215306122448978,
+      "loss": 0.1038,
+      "step": 10011
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0001621377551020408,
+      "loss": 0.2128,
+      "step": 10012
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00016212244897959184,
+      "loss": 0.1004,
+      "step": 10013
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00016210714285714285,
+      "loss": 0.1182,
+      "step": 10014
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00016209183673469385,
+      "loss": 0.1685,
+      "step": 10015
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00016207653061224488,
+      "loss": 0.1837,
+      "step": 10016
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.0001620612244897959,
+      "loss": 0.0588,
+      "step": 10017
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00016204591836734692,
+      "loss": 0.0676,
+      "step": 10018
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00016203061224489795,
+      "loss": 0.1037,
+      "step": 10019
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00016201530612244895,
+      "loss": 0.0516,
+      "step": 10020
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.000162,
+      "loss": 0.1736,
+      "step": 10021
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.000161984693877551,
+      "loss": 0.1743,
+      "step": 10022
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00016196938775510201,
+      "loss": 0.1981,
+      "step": 10023
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00016195408163265304,
+      "loss": 0.0669,
+      "step": 10024
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00016193877551020408,
+      "loss": 0.1084,
+      "step": 10025
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00016192346938775508,
+      "loss": 0.1389,
+      "step": 10026
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.0001619081632653061,
+      "loss": 0.3422,
+      "step": 10027
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.0001618928571428571,
+      "loss": 0.1071,
+      "step": 10028
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00016187755102040817,
+      "loss": 0.1108,
+      "step": 10029
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00016186224489795917,
+      "loss": 0.0888,
+      "step": 10030
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00016184693877551018,
+      "loss": 0.0867,
+      "step": 10031
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00016183163265306118,
+      "loss": 0.0728,
+      "step": 10032
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00016181632653061224,
+      "loss": 0.2087,
+      "step": 10033
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00016180102040816324,
+      "loss": 0.1338,
+      "step": 10034
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00016178571428571427,
+      "loss": 0.0661,
+      "step": 10035
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.0001617704081632653,
+      "loss": 0.0802,
+      "step": 10036
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00016175510204081633,
+      "loss": 0.1439,
+      "step": 10037
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.00016173979591836734,
+      "loss": 0.1462,
+      "step": 10038
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.00016172448979591834,
+      "loss": 0.3567,
+      "step": 10039
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0001617091836734694,
+      "loss": 0.0836,
+      "step": 10040
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0001616938775510204,
+      "loss": 0.1079,
+      "step": 10041
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0001616785714285714,
+      "loss": 0.075,
+      "step": 10042
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.00016166326530612244,
+      "loss": 0.0527,
+      "step": 10043
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.00016164795918367347,
+      "loss": 0.1363,
+      "step": 10044
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.00016163265306122447,
+      "loss": 0.1464,
+      "step": 10045
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0001616173469387755,
+      "loss": 0.1243,
+      "step": 10046
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0001616020408163265,
+      "loss": 0.1487,
+      "step": 10047
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016158673469387756,
+      "loss": 0.1593,
+      "step": 10048
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016157142857142856,
+      "loss": 0.0718,
+      "step": 10049
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016155612244897957,
+      "loss": 0.0771,
+      "step": 10050
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.0001615408163265306,
+      "loss": 0.0911,
+      "step": 10051
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016152551020408163,
+      "loss": 0.1741,
+      "step": 10052
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016151020408163263,
+      "loss": 0.1204,
+      "step": 10053
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016149489795918366,
+      "loss": 0.1554,
+      "step": 10054
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016147959183673467,
+      "loss": 0.127,
+      "step": 10055
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016146428571428572,
+      "loss": 0.0967,
+      "step": 10056
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00016144897959183673,
+      "loss": 0.069,
+      "step": 10057
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.00016143367346938773,
+      "loss": 0.3383,
+      "step": 10058
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.00016141836734693876,
+      "loss": 0.1166,
+      "step": 10059
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0001614030612244898,
+      "loss": 0.128,
+      "step": 10060
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0001613877551020408,
+      "loss": 0.1029,
+      "step": 10061
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.00016137244897959183,
+      "loss": 0.0614,
+      "step": 10062
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.00016135714285714283,
+      "loss": 0.0904,
+      "step": 10063
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0001613418367346939,
+      "loss": 0.3271,
+      "step": 10064
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0001613265306122449,
+      "loss": 0.1804,
+      "step": 10065
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0001613112244897959,
+      "loss": 0.191,
+      "step": 10066
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0001612959183673469,
+      "loss": 0.1414,
+      "step": 10067
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.00016128061224489796,
+      "loss": 0.0501,
+      "step": 10068
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016126530612244896,
+      "loss": 0.1106,
+      "step": 10069
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016125,
+      "loss": 0.2449,
+      "step": 10070
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.000161234693877551,
+      "loss": 0.1335,
+      "step": 10071
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016121938775510202,
+      "loss": 0.1253,
+      "step": 10072
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016120408163265305,
+      "loss": 0.1126,
+      "step": 10073
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016118877551020406,
+      "loss": 0.0822,
+      "step": 10074
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016117346938775506,
+      "loss": 0.1108,
+      "step": 10075
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016115816326530612,
+      "loss": 0.1711,
+      "step": 10076
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016114285714285712,
+      "loss": 0.2229,
+      "step": 10077
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00016112755102040815,
+      "loss": 0.0902,
+      "step": 10078
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016111224489795916,
+      "loss": 0.0747,
+      "step": 10079
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016109693877551019,
+      "loss": 0.075,
+      "step": 10080
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016108163265306122,
+      "loss": 0.0461,
+      "step": 10081
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016106632653061222,
+      "loss": 0.0914,
+      "step": 10082
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016105102040816328,
+      "loss": 0.2901,
+      "step": 10083
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016103571428571428,
+      "loss": 0.1219,
+      "step": 10084
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016102040816326528,
+      "loss": 0.1162,
+      "step": 10085
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016100510204081632,
+      "loss": 0.1292,
+      "step": 10086
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016098979591836735,
+      "loss": 0.1127,
+      "step": 10087
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00016097448979591835,
+      "loss": 0.1406,
+      "step": 10088
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00016095918367346938,
+      "loss": 0.1113,
+      "step": 10089
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00016094387755102038,
+      "loss": 0.0944,
+      "step": 10090
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00016092857142857144,
+      "loss": 0.1349,
+      "step": 10091
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00016091326530612244,
+      "loss": 0.1127,
+      "step": 10092
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00016089795918367345,
+      "loss": 0.0713,
+      "step": 10093
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00016088265306122445,
+      "loss": 0.147,
+      "step": 10094
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.0001608673469387755,
+      "loss": 0.1146,
+      "step": 10095
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.0001608520408163265,
+      "loss": 0.0766,
+      "step": 10096
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00016083673469387754,
+      "loss": 0.0636,
+      "step": 10097
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00016082142857142855,
+      "loss": 0.0714,
+      "step": 10098
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.0001608061224489796,
+      "loss": 0.11,
+      "step": 10099
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0001607908163265306,
+      "loss": 0.0754,
+      "step": 10100
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0001607755102040816,
+      "loss": 0.1013,
+      "step": 10101
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00016076020408163261,
+      "loss": 0.1636,
+      "step": 10102
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00016074489795918367,
+      "loss": 0.0888,
+      "step": 10103
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00016072959183673468,
+      "loss": 0.117,
+      "step": 10104
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0001607142857142857,
+      "loss": 0.1501,
+      "step": 10105
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0001606989795918367,
+      "loss": 0.0663,
+      "step": 10106
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00016068367346938774,
+      "loss": 0.0505,
+      "step": 10107
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00016066836734693877,
+      "loss": 0.107,
+      "step": 10108
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00016065306122448977,
+      "loss": 0.1587,
+      "step": 10109
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00016063775510204078,
+      "loss": 0.0701,
+      "step": 10110
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00016062244897959184,
+      "loss": 0.1034,
+      "step": 10111
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00016060714285714284,
+      "loss": 0.1037,
+      "step": 10112
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00016059183673469387,
+      "loss": 0.1423,
+      "step": 10113
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00016057653061224487,
+      "loss": 0.1783,
+      "step": 10114
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.0001605612244897959,
+      "loss": 0.1798,
+      "step": 10115
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00016054591836734693,
+      "loss": 0.0868,
+      "step": 10116
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00016053061224489794,
+      "loss": 0.1267,
+      "step": 10117
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00016051530612244894,
+      "loss": 0.0545,
+      "step": 10118
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.0001605,
+      "loss": 0.0625,
+      "step": 10119
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.000160484693877551,
+      "loss": 0.0882,
+      "step": 10120
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00016046938775510203,
+      "loss": 0.1909,
+      "step": 10121
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00016045408163265304,
+      "loss": 0.1123,
+      "step": 10122
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00016043877551020407,
+      "loss": 0.1228,
+      "step": 10123
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.0001604234693877551,
+      "loss": 0.0788,
+      "step": 10124
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.0001604081632653061,
+      "loss": 0.1802,
+      "step": 10125
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00016039285714285716,
+      "loss": 0.1016,
+      "step": 10126
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00016037755102040816,
+      "loss": 0.266,
+      "step": 10127
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00016036224489795916,
+      "loss": 0.1638,
+      "step": 10128
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00016034693877551017,
+      "loss": 0.1013,
+      "step": 10129
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00016033163265306123,
+      "loss": 0.0875,
+      "step": 10130
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00016031632653061223,
+      "loss": 0.1128,
+      "step": 10131
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00016030102040816326,
+      "loss": 0.0591,
+      "step": 10132
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00016028571428571426,
+      "loss": 0.1229,
+      "step": 10133
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.0001602704081632653,
+      "loss": 0.2152,
+      "step": 10134
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00016025510204081632,
+      "loss": 0.1763,
+      "step": 10135
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00016023979591836733,
+      "loss": 0.0812,
+      "step": 10136
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00016022448979591833,
+      "loss": 0.1006,
+      "step": 10137
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.0001602091836734694,
+      "loss": 0.1127,
+      "step": 10138
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.0001601938775510204,
+      "loss": 0.1962,
+      "step": 10139
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00016017857142857142,
+      "loss": 0.1059,
+      "step": 10140
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00016016326530612243,
+      "loss": 0.0879,
+      "step": 10141
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00016014795918367346,
+      "loss": 0.1841,
+      "step": 10142
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0001601326530612245,
+      "loss": 0.1011,
+      "step": 10143
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0001601173469387755,
+      "loss": 0.1088,
+      "step": 10144
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0001601020408163265,
+      "loss": 0.0954,
+      "step": 10145
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00016008673469387755,
+      "loss": 0.1439,
+      "step": 10146
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00016007142857142856,
+      "loss": 0.083,
+      "step": 10147
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00016005612244897959,
+      "loss": 0.0673,
+      "step": 10148
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0001600408163265306,
+      "loss": 0.1852,
+      "step": 10149
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00016002551020408162,
+      "loss": 0.1314,
+      "step": 10150
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00016001020408163265,
+      "loss": 0.0552,
+      "step": 10151
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015999489795918365,
+      "loss": 0.1006,
+      "step": 10152
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015997959183673466,
+      "loss": 0.1937,
+      "step": 10153
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015996428571428571,
+      "loss": 0.1849,
+      "step": 10154
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015994897959183672,
+      "loss": 0.0745,
+      "step": 10155
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015993367346938772,
+      "loss": 0.0448,
+      "step": 10156
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015991836734693875,
+      "loss": 0.0986,
+      "step": 10157
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015990306122448978,
+      "loss": 0.1994,
+      "step": 10158
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.0001598877551020408,
+      "loss": 0.1279,
+      "step": 10159
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015987244897959182,
+      "loss": 0.121,
+      "step": 10160
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015985714285714282,
+      "loss": 0.0594,
+      "step": 10161
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015984183673469388,
+      "loss": 0.0909,
+      "step": 10162
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015982653061224488,
+      "loss": 0.097,
+      "step": 10163
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015981122448979588,
+      "loss": 0.1828,
+      "step": 10164
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015979591836734692,
+      "loss": 0.1244,
+      "step": 10165
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015978061224489795,
+      "loss": 0.1472,
+      "step": 10166
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015976530612244898,
+      "loss": 0.1078,
+      "step": 10167
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015974999999999998,
+      "loss": 0.07,
+      "step": 10168
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015973469387755098,
+      "loss": 0.1264,
+      "step": 10169
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015971938775510204,
+      "loss": 0.1983,
+      "step": 10170
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015970408163265304,
+      "loss": 0.2174,
+      "step": 10171
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.00015968877551020405,
+      "loss": 0.1204,
+      "step": 10172
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0001596734693877551,
+      "loss": 0.0842,
+      "step": 10173
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0001596581632653061,
+      "loss": 0.0883,
+      "step": 10174
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.00015964285714285714,
+      "loss": 0.0553,
+      "step": 10175
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.00015962755102040814,
+      "loss": 0.2215,
+      "step": 10176
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.00015961224489795917,
+      "loss": 0.1469,
+      "step": 10177
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0001595969387755102,
+      "loss": 0.0961,
+      "step": 10178
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0001595816326530612,
+      "loss": 0.1135,
+      "step": 10179
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0001595663265306122,
+      "loss": 0.067,
+      "step": 10180
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.00015955102040816327,
+      "loss": 0.1095,
+      "step": 10181
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00015953571428571427,
+      "loss": 0.0705,
+      "step": 10182
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.0001595204081632653,
+      "loss": 0.2541,
+      "step": 10183
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.0001595051020408163,
+      "loss": 0.0787,
+      "step": 10184
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00015948979591836734,
+      "loss": 0.1305,
+      "step": 10185
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00015947448979591837,
+      "loss": 0.1041,
+      "step": 10186
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00015945918367346937,
+      "loss": 0.1628,
+      "step": 10187
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00015944387755102037,
+      "loss": 0.1315,
+      "step": 10188
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00015942857142857143,
+      "loss": 0.0972,
+      "step": 10189
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00015941326530612244,
+      "loss": 0.0841,
+      "step": 10190
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00015939795918367344,
+      "loss": 0.1182,
+      "step": 10191
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00015938265306122447,
+      "loss": 0.1113,
+      "step": 10192
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0001593673469387755,
+      "loss": 0.1161,
+      "step": 10193
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00015935204081632653,
+      "loss": 0.0618,
+      "step": 10194
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00015933673469387753,
+      "loss": 0.1925,
+      "step": 10195
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00015932142857142854,
+      "loss": 0.2145,
+      "step": 10196
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0001593061224489796,
+      "loss": 0.2151,
+      "step": 10197
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0001592908163265306,
+      "loss": 0.1439,
+      "step": 10198
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0001592755102040816,
+      "loss": 0.0917,
+      "step": 10199
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00015926020408163263,
+      "loss": 0.1041,
+      "step": 10200
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00015924489795918366,
+      "loss": 0.2698,
+      "step": 10201
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0001592295918367347,
+      "loss": 0.3396,
+      "step": 10202
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.0001592142857142857,
+      "loss": 0.1534,
+      "step": 10203
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.0001591989795918367,
+      "loss": 0.2064,
+      "step": 10204
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00015918367346938776,
+      "loss": 0.0661,
+      "step": 10205
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00015916836734693876,
+      "loss": 0.1202,
+      "step": 10206
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00015915306122448976,
+      "loss": 0.0794,
+      "step": 10207
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.0001591377551020408,
+      "loss": 0.1535,
+      "step": 10208
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00015912244897959183,
+      "loss": 0.119,
+      "step": 10209
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00015910714285714286,
+      "loss": 0.0853,
+      "step": 10210
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00015909183673469386,
+      "loss": 0.1453,
+      "step": 10211
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00015907653061224486,
+      "loss": 0.0658,
+      "step": 10212
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00015906122448979592,
+      "loss": 0.1064,
+      "step": 10213
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00015904591836734692,
+      "loss": 0.1712,
+      "step": 10214
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00015903061224489793,
+      "loss": 0.1074,
+      "step": 10215
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00015901530612244899,
+      "loss": 0.0995,
+      "step": 10216
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.000159,
+      "loss": 0.1018,
+      "step": 10217
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00015898469387755102,
+      "loss": 0.0747,
+      "step": 10218
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00015896938775510202,
+      "loss": 0.1122,
+      "step": 10219
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00015895408163265305,
+      "loss": 0.1518,
+      "step": 10220
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00015893877551020408,
+      "loss": 0.1512,
+      "step": 10221
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.0001589234693877551,
+      "loss": 0.0859,
+      "step": 10222
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.0001589081632653061,
+      "loss": 0.0701,
+      "step": 10223
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015889285714285715,
+      "loss": 0.127,
+      "step": 10224
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015887755102040815,
+      "loss": 0.0686,
+      "step": 10225
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015886224489795916,
+      "loss": 0.1239,
+      "step": 10226
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015884693877551019,
+      "loss": 0.198,
+      "step": 10227
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015883163265306122,
+      "loss": 0.1314,
+      "step": 10228
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015881632653061225,
+      "loss": 0.1366,
+      "step": 10229
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015880102040816325,
+      "loss": 0.1632,
+      "step": 10230
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015878571428571425,
+      "loss": 0.0615,
+      "step": 10231
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.0001587704081632653,
+      "loss": 0.0939,
+      "step": 10232
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00015875510204081631,
+      "loss": 0.1084,
+      "step": 10233
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00015873979591836732,
+      "loss": 0.1653,
+      "step": 10234
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00015872448979591835,
+      "loss": 0.2187,
+      "step": 10235
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00015870918367346938,
+      "loss": 0.1317,
+      "step": 10236
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0001586938775510204,
+      "loss": 0.067,
+      "step": 10237
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0001586785714285714,
+      "loss": 0.0634,
+      "step": 10238
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00015866326530612242,
+      "loss": 0.1489,
+      "step": 10239
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00015864795918367347,
+      "loss": 0.0911,
+      "step": 10240
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00015863265306122448,
+      "loss": 0.0895,
+      "step": 10241
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00015861734693877548,
+      "loss": 0.1072,
+      "step": 10242
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0001586020408163265,
+      "loss": 0.0953,
+      "step": 10243
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015858673469387754,
+      "loss": 0.0905,
+      "step": 10244
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015857142857142857,
+      "loss": 0.0977,
+      "step": 10245
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015855612244897958,
+      "loss": 0.0898,
+      "step": 10246
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015854081632653058,
+      "loss": 0.1504,
+      "step": 10247
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015852551020408164,
+      "loss": 0.0814,
+      "step": 10248
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015851020408163264,
+      "loss": 0.0746,
+      "step": 10249
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015849489795918364,
+      "loss": 0.0647,
+      "step": 10250
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015847959183673467,
+      "loss": 0.0803,
+      "step": 10251
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.0001584642857142857,
+      "loss": 0.1419,
+      "step": 10252
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.0001584489795918367,
+      "loss": 0.1304,
+      "step": 10253
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00015843367346938774,
+      "loss": 0.141,
+      "step": 10254
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00015841836734693874,
+      "loss": 0.1118,
+      "step": 10255
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.0001584030612244898,
+      "loss": 0.1141,
+      "step": 10256
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.0001583877551020408,
+      "loss": 0.0722,
+      "step": 10257
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.0001583724489795918,
+      "loss": 0.1614,
+      "step": 10258
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00015835714285714284,
+      "loss": 0.1697,
+      "step": 10259
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00015834183673469387,
+      "loss": 0.2474,
+      "step": 10260
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00015832653061224487,
+      "loss": 0.0975,
+      "step": 10261
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.0001583112244897959,
+      "loss": 0.1151,
+      "step": 10262
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00015829591836734693,
+      "loss": 0.0499,
+      "step": 10263
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015828061224489796,
+      "loss": 0.1928,
+      "step": 10264
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015826530612244897,
+      "loss": 0.0929,
+      "step": 10265
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015824999999999997,
+      "loss": 0.0834,
+      "step": 10266
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015823469387755103,
+      "loss": 0.1597,
+      "step": 10267
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015821938775510203,
+      "loss": 0.1173,
+      "step": 10268
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015820408163265304,
+      "loss": 0.0749,
+      "step": 10269
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015818877551020407,
+      "loss": 0.1621,
+      "step": 10270
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.0001581734693877551,
+      "loss": 0.1214,
+      "step": 10271
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015815816326530613,
+      "loss": 0.1026,
+      "step": 10272
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015814285714285713,
+      "loss": 0.1523,
+      "step": 10273
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00015812755102040813,
+      "loss": 0.1146,
+      "step": 10274
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.0001581122448979592,
+      "loss": 0.0833,
+      "step": 10275
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.0001580969387755102,
+      "loss": 0.1016,
+      "step": 10276
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.0001580816326530612,
+      "loss": 0.1748,
+      "step": 10277
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.00015806632653061223,
+      "loss": 0.1409,
+      "step": 10278
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.00015805102040816326,
+      "loss": 0.1145,
+      "step": 10279
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.0001580357142857143,
+      "loss": 0.111,
+      "step": 10280
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.0001580204081632653,
+      "loss": 0.1126,
+      "step": 10281
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.0001580051020408163,
+      "loss": 0.1579,
+      "step": 10282
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.00015798979591836735,
+      "loss": 0.2345,
+      "step": 10283
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.00015797448979591836,
+      "loss": 0.2398,
+      "step": 10284
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015795918367346936,
+      "loss": 0.1136,
+      "step": 10285
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.0001579438775510204,
+      "loss": 0.1577,
+      "step": 10286
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015792857142857142,
+      "loss": 0.2226,
+      "step": 10287
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015791326530612243,
+      "loss": 0.1109,
+      "step": 10288
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015789795918367346,
+      "loss": 0.2407,
+      "step": 10289
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015788265306122446,
+      "loss": 0.1013,
+      "step": 10290
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015786734693877552,
+      "loss": 0.07,
+      "step": 10291
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015785204081632652,
+      "loss": 0.0806,
+      "step": 10292
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015783673469387752,
+      "loss": 0.1753,
+      "step": 10293
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015782142857142855,
+      "loss": 0.0849,
+      "step": 10294
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015780612244897959,
+      "loss": 0.2371,
+      "step": 10295
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.0001577908163265306,
+      "loss": 0.1222,
+      "step": 10296
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015777551020408162,
+      "loss": 0.1224,
+      "step": 10297
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015776020408163262,
+      "loss": 0.0683,
+      "step": 10298
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015774489795918368,
+      "loss": 0.074,
+      "step": 10299
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015772959183673468,
+      "loss": 0.1366,
+      "step": 10300
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.0001577142857142857,
+      "loss": 0.1559,
+      "step": 10301
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015769897959183672,
+      "loss": 0.0734,
+      "step": 10302
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015768367346938775,
+      "loss": 0.112,
+      "step": 10303
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015766836734693875,
+      "loss": 0.0548,
+      "step": 10304
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015765306122448978,
+      "loss": 0.134,
+      "step": 10305
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.0001576377551020408,
+      "loss": 0.1389,
+      "step": 10306
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00015762244897959184,
+      "loss": 0.0691,
+      "step": 10307
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00015760714285714285,
+      "loss": 0.073,
+      "step": 10308
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00015759183673469385,
+      "loss": 0.1384,
+      "step": 10309
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.0001575765306122449,
+      "loss": 0.0702,
+      "step": 10310
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.0001575612244897959,
+      "loss": 0.104,
+      "step": 10311
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00015754591836734691,
+      "loss": 0.081,
+      "step": 10312
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00015753061224489795,
+      "loss": 0.1832,
+      "step": 10313
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00015751530612244898,
+      "loss": 0.1102,
+      "step": 10314
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00015749999999999998,
+      "loss": 0.14,
+      "step": 10315
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.000157484693877551,
+      "loss": 0.1153,
+      "step": 10316
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.000157469387755102,
+      "loss": 0.0723,
+      "step": 10317
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00015745408163265307,
+      "loss": 0.1473,
+      "step": 10318
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00015743877551020407,
+      "loss": 0.3497,
+      "step": 10319
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00015742346938775508,
+      "loss": 0.1749,
+      "step": 10320
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.0001574081632653061,
+      "loss": 0.1818,
+      "step": 10321
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00015739285714285714,
+      "loss": 0.1081,
+      "step": 10322
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00015737755102040814,
+      "loss": 0.1535,
+      "step": 10323
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00015736224489795917,
+      "loss": 0.0762,
+      "step": 10324
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00015734693877551018,
+      "loss": 0.2046,
+      "step": 10325
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.00015733163265306123,
+      "loss": 0.1147,
+      "step": 10326
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.00015731632653061224,
+      "loss": 0.0986,
+      "step": 10327
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.00015730102040816324,
+      "loss": 0.1215,
+      "step": 10328
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.00015728571428571427,
+      "loss": 0.164,
+      "step": 10329
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0001572704081632653,
+      "loss": 0.1206,
+      "step": 10330
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0001572551020408163,
+      "loss": 0.0433,
+      "step": 10331
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.00015723979591836734,
+      "loss": 0.1214,
+      "step": 10332
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.00015722448979591834,
+      "loss": 0.0981,
+      "step": 10333
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0001572091836734694,
+      "loss": 0.0657,
+      "step": 10334
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0001571938775510204,
+      "loss": 0.1198,
+      "step": 10335
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0001571785714285714,
+      "loss": 0.0656,
+      "step": 10336
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015716326530612243,
+      "loss": 0.0963,
+      "step": 10337
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015714795918367347,
+      "loss": 0.1717,
+      "step": 10338
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015713265306122447,
+      "loss": 0.1567,
+      "step": 10339
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.0001571173469387755,
+      "loss": 0.1051,
+      "step": 10340
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.0001571020408163265,
+      "loss": 0.1037,
+      "step": 10341
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015708673469387756,
+      "loss": 0.0602,
+      "step": 10342
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015707142857142856,
+      "loss": 0.0567,
+      "step": 10343
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015705612244897957,
+      "loss": 0.16,
+      "step": 10344
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015704081632653057,
+      "loss": 0.0851,
+      "step": 10345
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015702551020408163,
+      "loss": 0.1003,
+      "step": 10346
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00015701020408163263,
+      "loss": 0.1141,
+      "step": 10347
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00015699489795918366,
+      "loss": 0.058,
+      "step": 10348
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00015697959183673467,
+      "loss": 0.0639,
+      "step": 10349
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.0001569642857142857,
+      "loss": 0.1845,
+      "step": 10350
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00015694897959183673,
+      "loss": 0.1299,
+      "step": 10351
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00015693367346938773,
+      "loss": 0.1327,
+      "step": 10352
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.0001569183673469388,
+      "loss": 0.1352,
+      "step": 10353
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.0001569030612244898,
+      "loss": 0.0832,
+      "step": 10354
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.0001568877551020408,
+      "loss": 0.0489,
+      "step": 10355
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00015687244897959183,
+      "loss": 0.0947,
+      "step": 10356
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00015685714285714286,
+      "loss": 0.1157,
+      "step": 10357
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00015684183673469386,
+      "loss": 0.1374,
+      "step": 10358
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.0001568265306122449,
+      "loss": 0.1637,
+      "step": 10359
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.0001568112244897959,
+      "loss": 0.0458,
+      "step": 10360
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00015679591836734695,
+      "loss": 0.054,
+      "step": 10361
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00015678061224489795,
+      "loss": 0.1079,
+      "step": 10362
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00015676530612244896,
+      "loss": 0.2612,
+      "step": 10363
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00015675,
+      "loss": 0.1374,
+      "step": 10364
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00015673469387755102,
+      "loss": 0.1053,
+      "step": 10365
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00015671938775510202,
+      "loss": 0.1292,
+      "step": 10366
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015670408163265305,
+      "loss": 0.0859,
+      "step": 10367
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015668877551020406,
+      "loss": 0.0438,
+      "step": 10368
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015667346938775511,
+      "loss": 0.1595,
+      "step": 10369
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015665816326530612,
+      "loss": 0.0704,
+      "step": 10370
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015664285714285712,
+      "loss": 0.1448,
+      "step": 10371
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015662755102040812,
+      "loss": 0.0977,
+      "step": 10372
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015661224489795918,
+      "loss": 0.0994,
+      "step": 10373
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015659693877551019,
+      "loss": 0.0823,
+      "step": 10374
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015658163265306122,
+      "loss": 0.2481,
+      "step": 10375
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015656632653061222,
+      "loss": 0.1731,
+      "step": 10376
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00015655102040816328,
+      "loss": 0.1499,
+      "step": 10377
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015653571428571428,
+      "loss": 0.0966,
+      "step": 10378
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015652040816326528,
+      "loss": 0.0563,
+      "step": 10379
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.0001565051020408163,
+      "loss": 0.0503,
+      "step": 10380
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015648979591836734,
+      "loss": 0.096,
+      "step": 10381
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015647448979591835,
+      "loss": 0.1475,
+      "step": 10382
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015645918367346938,
+      "loss": 0.2256,
+      "step": 10383
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015644387755102038,
+      "loss": 0.1003,
+      "step": 10384
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.0001564285714285714,
+      "loss": 0.1708,
+      "step": 10385
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015641326530612244,
+      "loss": 0.121,
+      "step": 10386
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015639795918367345,
+      "loss": 0.0936,
+      "step": 10387
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.00015638265306122445,
+      "loss": 0.1194,
+      "step": 10388
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0001563673469387755,
+      "loss": 0.1968,
+      "step": 10389
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0001563520408163265,
+      "loss": 0.1729,
+      "step": 10390
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.00015633673469387754,
+      "loss": 0.1141,
+      "step": 10391
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.00015632142857142855,
+      "loss": 0.0956,
+      "step": 10392
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.00015630612244897958,
+      "loss": 0.0876,
+      "step": 10393
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0001562908163265306,
+      "loss": 0.293,
+      "step": 10394
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0001562755102040816,
+      "loss": 0.133,
+      "step": 10395
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0001562602040816326,
+      "loss": 0.1465,
+      "step": 10396
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.00015624489795918367,
+      "loss": 0.1068,
+      "step": 10397
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015622959183673467,
+      "loss": 0.0499,
+      "step": 10398
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.0001562142857142857,
+      "loss": 0.0681,
+      "step": 10399
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015619897959183674,
+      "loss": 0.0416,
+      "step": 10400
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015618367346938774,
+      "loss": 0.1457,
+      "step": 10401
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015616836734693877,
+      "loss": 0.0717,
+      "step": 10402
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015615306122448977,
+      "loss": 0.1374,
+      "step": 10403
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015613775510204083,
+      "loss": 0.0753,
+      "step": 10404
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015612244897959183,
+      "loss": 0.1355,
+      "step": 10405
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015610714285714284,
+      "loss": 0.1773,
+      "step": 10406
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015609183673469384,
+      "loss": 0.0758,
+      "step": 10407
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.0001560765306122449,
+      "loss": 0.0895,
+      "step": 10408
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.0001560612244897959,
+      "loss": 0.0919,
+      "step": 10409
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00015604591836734693,
+      "loss": 0.085,
+      "step": 10410
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00015603061224489794,
+      "loss": 0.0477,
+      "step": 10411
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00015601530612244897,
+      "loss": 0.1224,
+      "step": 10412
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.000156,
+      "loss": 0.1552,
+      "step": 10413
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.000155984693877551,
+      "loss": 0.1684,
+      "step": 10414
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.000155969387755102,
+      "loss": 0.0606,
+      "step": 10415
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00015595408163265306,
+      "loss": 0.0617,
+      "step": 10416
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00015593877551020407,
+      "loss": 0.0646,
+      "step": 10417
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.0001559234693877551,
+      "loss": 0.0508,
+      "step": 10418
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.0001559081632653061,
+      "loss": 0.1137,
+      "step": 10419
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00015589285714285713,
+      "loss": 0.1231,
+      "step": 10420
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00015587755102040816,
+      "loss": 0.1007,
+      "step": 10421
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00015586224489795916,
+      "loss": 0.073,
+      "step": 10422
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00015584693877551017,
+      "loss": 0.1046,
+      "step": 10423
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00015583163265306122,
+      "loss": 0.0618,
+      "step": 10424
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00015581632653061223,
+      "loss": 0.0697,
+      "step": 10425
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00015580102040816326,
+      "loss": 0.1638,
+      "step": 10426
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00015578571428571426,
+      "loss": 0.1027,
+      "step": 10427
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.0001557704081632653,
+      "loss": 0.0663,
+      "step": 10428
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00015575510204081632,
+      "loss": 0.179,
+      "step": 10429
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00015573979591836733,
+      "loss": 0.0521,
+      "step": 10430
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00015572448979591833,
+      "loss": 0.1047,
+      "step": 10431
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.0001557091836734694,
+      "loss": 0.1301,
+      "step": 10432
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.0001556938775510204,
+      "loss": 0.0714,
+      "step": 10433
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.0001556785714285714,
+      "loss": 0.0581,
+      "step": 10434
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00015566326530612243,
+      "loss": 0.0951,
+      "step": 10435
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00015564795918367346,
+      "loss": 0.0422,
+      "step": 10436
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00015563265306122449,
+      "loss": 0.1625,
+      "step": 10437
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.0001556173469387755,
+      "loss": 0.2357,
+      "step": 10438
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.0001556020408163265,
+      "loss": 0.0732,
+      "step": 10439
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015558673469387755,
+      "loss": 0.1117,
+      "step": 10440
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015557142857142855,
+      "loss": 0.1,
+      "step": 10441
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015555612244897956,
+      "loss": 0.2309,
+      "step": 10442
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015554081632653062,
+      "loss": 0.0937,
+      "step": 10443
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015552551020408162,
+      "loss": 0.0999,
+      "step": 10444
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015551020408163265,
+      "loss": 0.1282,
+      "step": 10445
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015549489795918365,
+      "loss": 0.1902,
+      "step": 10446
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015547959183673468,
+      "loss": 0.0769,
+      "step": 10447
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015546428571428571,
+      "loss": 0.0691,
+      "step": 10448
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00015544897959183672,
+      "loss": 0.1411,
+      "step": 10449
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015543367346938772,
+      "loss": 0.1772,
+      "step": 10450
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015541836734693878,
+      "loss": 0.1024,
+      "step": 10451
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015540306122448978,
+      "loss": 0.0978,
+      "step": 10452
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.0001553877551020408,
+      "loss": 0.0681,
+      "step": 10453
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015537244897959182,
+      "loss": 0.0732,
+      "step": 10454
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015535714285714285,
+      "loss": 0.0696,
+      "step": 10455
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015534183673469388,
+      "loss": 0.0574,
+      "step": 10456
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015532653061224488,
+      "loss": 0.1205,
+      "step": 10457
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015531122448979588,
+      "loss": 0.1422,
+      "step": 10458
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015529591836734694,
+      "loss": 0.1467,
+      "step": 10459
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00015528061224489794,
+      "loss": 0.0939,
+      "step": 10460
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00015526530612244898,
+      "loss": 0.0722,
+      "step": 10461
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00015524999999999998,
+      "loss": 0.1083,
+      "step": 10462
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.000155234693877551,
+      "loss": 0.2467,
+      "step": 10463
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00015521938775510204,
+      "loss": 0.1144,
+      "step": 10464
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00015520408163265304,
+      "loss": 0.1037,
+      "step": 10465
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00015518877551020405,
+      "loss": 0.1311,
+      "step": 10466
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.0001551734693877551,
+      "loss": 0.0607,
+      "step": 10467
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.0001551581632653061,
+      "loss": 0.0757,
+      "step": 10468
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.0001551428571428571,
+      "loss": 0.2581,
+      "step": 10469
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00015512755102040814,
+      "loss": 0.1616,
+      "step": 10470
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00015511224489795917,
+      "loss": 0.0923,
+      "step": 10471
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0001550969387755102,
+      "loss": 0.0906,
+      "step": 10472
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0001550816326530612,
+      "loss": 0.0559,
+      "step": 10473
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0001550663265306122,
+      "loss": 0.0505,
+      "step": 10474
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00015505102040816327,
+      "loss": 0.1242,
+      "step": 10475
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00015503571428571427,
+      "loss": 0.1328,
+      "step": 10476
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00015502040816326527,
+      "loss": 0.1221,
+      "step": 10477
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0001550051020408163,
+      "loss": 0.0712,
+      "step": 10478
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00015498979591836734,
+      "loss": 0.1,
+      "step": 10479
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00015497448979591837,
+      "loss": 0.0538,
+      "step": 10480
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015495918367346937,
+      "loss": 0.0836,
+      "step": 10481
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015494387755102037,
+      "loss": 0.0743,
+      "step": 10482
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015492857142857143,
+      "loss": 0.2135,
+      "step": 10483
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015491326530612243,
+      "loss": 0.1493,
+      "step": 10484
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015489795918367344,
+      "loss": 0.1092,
+      "step": 10485
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015488265306122447,
+      "loss": 0.0639,
+      "step": 10486
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.0001548673469387755,
+      "loss": 0.1243,
+      "step": 10487
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015485204081632653,
+      "loss": 0.2042,
+      "step": 10488
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015483673469387753,
+      "loss": 0.1217,
+      "step": 10489
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00015482142857142856,
+      "loss": 0.148,
+      "step": 10490
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.0001548061224489796,
+      "loss": 0.1422,
+      "step": 10491
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.0001547908163265306,
+      "loss": 0.1283,
+      "step": 10492
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.0001547755102040816,
+      "loss": 0.0607,
+      "step": 10493
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00015476020408163266,
+      "loss": 0.2568,
+      "step": 10494
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00015474489795918366,
+      "loss": 0.1466,
+      "step": 10495
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.0001547295918367347,
+      "loss": 0.1534,
+      "step": 10496
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.0001547142857142857,
+      "loss": 0.0724,
+      "step": 10497
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00015469897959183673,
+      "loss": 0.0563,
+      "step": 10498
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00015468367346938776,
+      "loss": 0.0586,
+      "step": 10499
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00015466836734693876,
+      "loss": 0.2891,
+      "step": 10500
+    },
+    {
+      "epoch": 10.19,
+      "eval_cer": 0.2570412268936964,
+      "eval_loss": 1.042435646057129,
+      "eval_runtime": 252.7317,
+      "eval_samples_per_second": 34.875,
+      "eval_steps_per_second": 4.36,
+      "eval_wer": 0.6708521460194005,
+      "step": 10500
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015465306122448976,
+      "loss": 0.1634,
+      "step": 10501
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015463775510204082,
+      "loss": 0.1595,
+      "step": 10502
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015462244897959182,
+      "loss": 0.1121,
+      "step": 10503
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015460714285714283,
+      "loss": 0.062,
+      "step": 10504
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015459183673469386,
+      "loss": 0.0622,
+      "step": 10505
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.0001545765306122449,
+      "loss": 0.0629,
+      "step": 10506
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015456122448979592,
+      "loss": 0.1375,
+      "step": 10507
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015454591836734692,
+      "loss": 0.1435,
+      "step": 10508
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015453061224489793,
+      "loss": 0.1482,
+      "step": 10509
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015451530612244898,
+      "loss": 0.1298,
+      "step": 10510
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.0001545,
+      "loss": 0.1035,
+      "step": 10511
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.000154484693877551,
+      "loss": 0.1125,
+      "step": 10512
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00015446938775510202,
+      "loss": 0.2364,
+      "step": 10513
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00015445408163265305,
+      "loss": 0.2255,
+      "step": 10514
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00015443877551020408,
+      "loss": 0.1238,
+      "step": 10515
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00015442346938775509,
+      "loss": 0.0766,
+      "step": 10516
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.0001544081632653061,
+      "loss": 0.1063,
+      "step": 10517
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00015439285714285715,
+      "loss": 0.0936,
+      "step": 10518
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00015437755102040815,
+      "loss": 0.1914,
+      "step": 10519
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00015436224489795915,
+      "loss": 0.2235,
+      "step": 10520
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00015434693877551018,
+      "loss": 0.1031,
+      "step": 10521
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015433163265306122,
+      "loss": 0.0657,
+      "step": 10522
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015431632653061225,
+      "loss": 0.0757,
+      "step": 10523
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015430102040816325,
+      "loss": 0.0693,
+      "step": 10524
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015428571428571425,
+      "loss": 0.1968,
+      "step": 10525
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.0001542704081632653,
+      "loss": 0.237,
+      "step": 10526
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015425510204081631,
+      "loss": 0.1236,
+      "step": 10527
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015423979591836732,
+      "loss": 0.0514,
+      "step": 10528
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015422448979591835,
+      "loss": 0.0715,
+      "step": 10529
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015420918367346938,
+      "loss": 0.0901,
+      "step": 10530
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015419387755102038,
+      "loss": 0.1419,
+      "step": 10531
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.0001541785714285714,
+      "loss": 0.231,
+      "step": 10532
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015416326530612244,
+      "loss": 0.0669,
+      "step": 10533
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015414795918367347,
+      "loss": 0.139,
+      "step": 10534
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015413265306122448,
+      "loss": 0.0879,
+      "step": 10535
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015411734693877548,
+      "loss": 0.0581,
+      "step": 10536
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015410204081632654,
+      "loss": 0.076,
+      "step": 10537
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015408673469387754,
+      "loss": 0.2917,
+      "step": 10538
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015407142857142854,
+      "loss": 0.0847,
+      "step": 10539
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015405612244897958,
+      "loss": 0.0642,
+      "step": 10540
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.0001540408163265306,
+      "loss": 0.0736,
+      "step": 10541
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015402551020408164,
+      "loss": 0.056,
+      "step": 10542
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.00015401020408163264,
+      "loss": 0.0896,
+      "step": 10543
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.00015399489795918364,
+      "loss": 0.1777,
+      "step": 10544
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0001539795918367347,
+      "loss": 0.0833,
+      "step": 10545
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0001539642857142857,
+      "loss": 0.082,
+      "step": 10546
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0001539489795918367,
+      "loss": 0.1162,
+      "step": 10547
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.00015393367346938774,
+      "loss": 0.0547,
+      "step": 10548
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.00015391836734693877,
+      "loss": 0.1082,
+      "step": 10549
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0001539030612244898,
+      "loss": 0.1279,
+      "step": 10550
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0001538877551020408,
+      "loss": 0.0857,
+      "step": 10551
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0001538724489795918,
+      "loss": 0.0871,
+      "step": 10552
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015385714285714286,
+      "loss": 0.0769,
+      "step": 10553
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015384183673469387,
+      "loss": 0.1106,
+      "step": 10554
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015382653061224487,
+      "loss": 0.0752,
+      "step": 10555
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.0001538112244897959,
+      "loss": 0.0491,
+      "step": 10556
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015379591836734693,
+      "loss": 0.1711,
+      "step": 10557
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015378061224489796,
+      "loss": 0.0869,
+      "step": 10558
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015376530612244897,
+      "loss": 0.0974,
+      "step": 10559
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015374999999999997,
+      "loss": 0.067,
+      "step": 10560
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015373469387755103,
+      "loss": 0.1494,
+      "step": 10561
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015371938775510203,
+      "loss": 0.0839,
+      "step": 10562
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00015370408163265303,
+      "loss": 0.1746,
+      "step": 10563
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00015368877551020406,
+      "loss": 0.1585,
+      "step": 10564
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.0001536734693877551,
+      "loss": 0.0667,
+      "step": 10565
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.0001536581632653061,
+      "loss": 0.0684,
+      "step": 10566
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00015364285714285713,
+      "loss": 0.0899,
+      "step": 10567
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00015362755102040813,
+      "loss": 0.1664,
+      "step": 10568
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.0001536122448979592,
+      "loss": 0.0926,
+      "step": 10569
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.0001535969387755102,
+      "loss": 0.1999,
+      "step": 10570
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.0001535816326530612,
+      "loss": 0.1069,
+      "step": 10571
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00015356632653061223,
+      "loss": 0.0562,
+      "step": 10572
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015355102040816326,
+      "loss": 0.1327,
+      "step": 10573
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015353571428571426,
+      "loss": 0.1783,
+      "step": 10574
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.0001535204081632653,
+      "loss": 0.1033,
+      "step": 10575
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.0001535051020408163,
+      "loss": 0.1598,
+      "step": 10576
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015348979591836735,
+      "loss": 0.0828,
+      "step": 10577
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015347448979591836,
+      "loss": 0.1133,
+      "step": 10578
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015345918367346936,
+      "loss": 0.0513,
+      "step": 10579
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015344387755102042,
+      "loss": 0.0488,
+      "step": 10580
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015342857142857142,
+      "loss": 0.0883,
+      "step": 10581
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015341326530612242,
+      "loss": 0.122,
+      "step": 10582
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00015339795918367346,
+      "loss": 0.1122,
+      "step": 10583
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015338265306122449,
+      "loss": 0.1085,
+      "step": 10584
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015336734693877552,
+      "loss": 0.0778,
+      "step": 10585
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015335204081632652,
+      "loss": 0.1103,
+      "step": 10586
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015333673469387752,
+      "loss": 0.0675,
+      "step": 10587
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015332142857142858,
+      "loss": 0.123,
+      "step": 10588
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015330612244897958,
+      "loss": 0.0825,
+      "step": 10589
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.0001532908163265306,
+      "loss": 0.0669,
+      "step": 10590
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015327551020408162,
+      "loss": 0.0985,
+      "step": 10591
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015326020408163265,
+      "loss": 0.1678,
+      "step": 10592
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00015324489795918365,
+      "loss": 0.0652,
+      "step": 10593
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015322959183673468,
+      "loss": 0.0727,
+      "step": 10594
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015321428571428569,
+      "loss": 0.1919,
+      "step": 10595
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015319897959183674,
+      "loss": 0.1086,
+      "step": 10596
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015318367346938775,
+      "loss": 0.0813,
+      "step": 10597
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015316836734693875,
+      "loss": 0.1002,
+      "step": 10598
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015315306122448978,
+      "loss": 0.05,
+      "step": 10599
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.0001531377551020408,
+      "loss": 0.1245,
+      "step": 10600
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015312244897959182,
+      "loss": 0.084,
+      "step": 10601
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015310714285714285,
+      "loss": 0.2532,
+      "step": 10602
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00015309183673469385,
+      "loss": 0.0735,
+      "step": 10603
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.0001530765306122449,
+      "loss": 0.1112,
+      "step": 10604
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.0001530612244897959,
+      "loss": 0.12,
+      "step": 10605
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00015304591836734691,
+      "loss": 0.1308,
+      "step": 10606
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00015303061224489794,
+      "loss": 0.1559,
+      "step": 10607
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00015301530612244898,
+      "loss": 0.0852,
+      "step": 10608
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00015299999999999998,
+      "loss": 0.1434,
+      "step": 10609
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.000152984693877551,
+      "loss": 0.0685,
+      "step": 10610
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.000152969387755102,
+      "loss": 0.2033,
+      "step": 10611
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00015295408163265307,
+      "loss": 0.077,
+      "step": 10612
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00015293877551020407,
+      "loss": 0.2218,
+      "step": 10613
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00015292346938775508,
+      "loss": 0.1251,
+      "step": 10614
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015290816326530608,
+      "loss": 0.0893,
+      "step": 10615
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015289285714285714,
+      "loss": 0.1268,
+      "step": 10616
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015287755102040814,
+      "loss": 0.0951,
+      "step": 10617
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015286224489795917,
+      "loss": 0.0758,
+      "step": 10618
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015284693877551018,
+      "loss": 0.0983,
+      "step": 10619
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015283163265306123,
+      "loss": 0.1588,
+      "step": 10620
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015281632653061224,
+      "loss": 0.0597,
+      "step": 10621
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015280102040816324,
+      "loss": 0.11,
+      "step": 10622
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00015278571428571424,
+      "loss": 0.136,
+      "step": 10623
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.0001527704081632653,
+      "loss": 0.0574,
+      "step": 10624
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0001527551020408163,
+      "loss": 0.1592,
+      "step": 10625
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00015273979591836734,
+      "loss": 0.1497,
+      "step": 10626
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00015272448979591837,
+      "loss": 0.0688,
+      "step": 10627
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00015270918367346937,
+      "loss": 0.1404,
+      "step": 10628
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0001526938775510204,
+      "loss": 0.1603,
+      "step": 10629
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0001526785714285714,
+      "loss": 0.2104,
+      "step": 10630
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00015266326530612246,
+      "loss": 0.1689,
+      "step": 10631
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00015264795918367346,
+      "loss": 0.1956,
+      "step": 10632
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00015263265306122447,
+      "loss": 0.0879,
+      "step": 10633
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0001526173469387755,
+      "loss": 0.0734,
+      "step": 10634
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015260204081632653,
+      "loss": 0.1514,
+      "step": 10635
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015258673469387753,
+      "loss": 0.1031,
+      "step": 10636
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015257142857142856,
+      "loss": 0.0931,
+      "step": 10637
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015255612244897957,
+      "loss": 0.2343,
+      "step": 10638
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015254081632653062,
+      "loss": 0.1697,
+      "step": 10639
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015252551020408163,
+      "loss": 0.0565,
+      "step": 10640
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015251020408163263,
+      "loss": 0.1246,
+      "step": 10641
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015249489795918366,
+      "loss": 0.0459,
+      "step": 10642
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.0001524795918367347,
+      "loss": 0.096,
+      "step": 10643
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.0001524642857142857,
+      "loss": 0.261,
+      "step": 10644
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00015244897959183673,
+      "loss": 0.103,
+      "step": 10645
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00015243367346938773,
+      "loss": 0.0604,
+      "step": 10646
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0001524183673469388,
+      "loss": 0.1875,
+      "step": 10647
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0001524030612244898,
+      "loss": 0.1237,
+      "step": 10648
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0001523877551020408,
+      "loss": 0.0726,
+      "step": 10649
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0001523724489795918,
+      "loss": 0.1132,
+      "step": 10650
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00015235714285714285,
+      "loss": 0.1865,
+      "step": 10651
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00015234183673469386,
+      "loss": 0.1156,
+      "step": 10652
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0001523265306122449,
+      "loss": 0.1214,
+      "step": 10653
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0001523112244897959,
+      "loss": 0.0693,
+      "step": 10654
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00015229591836734692,
+      "loss": 0.1386,
+      "step": 10655
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015228061224489795,
+      "loss": 0.1742,
+      "step": 10656
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015226530612244896,
+      "loss": 0.2297,
+      "step": 10657
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015224999999999996,
+      "loss": 0.1209,
+      "step": 10658
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015223469387755102,
+      "loss": 0.1782,
+      "step": 10659
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015221938775510202,
+      "loss": 0.114,
+      "step": 10660
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015220408163265305,
+      "loss": 0.1199,
+      "step": 10661
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015218877551020406,
+      "loss": 0.1094,
+      "step": 10662
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015217346938775509,
+      "loss": 0.2262,
+      "step": 10663
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015215816326530612,
+      "loss": 0.1105,
+      "step": 10664
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00015214285714285712,
+      "loss": 0.0523,
+      "step": 10665
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015212755102040812,
+      "loss": 0.0957,
+      "step": 10666
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015211224489795918,
+      "loss": 0.0837,
+      "step": 10667
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015209693877551018,
+      "loss": 0.0484,
+      "step": 10668
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015208163265306121,
+      "loss": 0.1501,
+      "step": 10669
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015206632653061225,
+      "loss": 0.1286,
+      "step": 10670
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015205102040816325,
+      "loss": 0.107,
+      "step": 10671
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015203571428571428,
+      "loss": 0.0679,
+      "step": 10672
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015202040816326528,
+      "loss": 0.0824,
+      "step": 10673
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015200510204081634,
+      "loss": 0.0756,
+      "step": 10674
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00015198979591836734,
+      "loss": 0.075,
+      "step": 10675
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00015197448979591835,
+      "loss": 0.2475,
+      "step": 10676
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00015195918367346938,
+      "loss": 0.13,
+      "step": 10677
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0001519438775510204,
+      "loss": 0.1797,
+      "step": 10678
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0001519285714285714,
+      "loss": 0.0981,
+      "step": 10679
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00015191326530612244,
+      "loss": 0.0753,
+      "step": 10680
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00015189795918367345,
+      "loss": 0.0941,
+      "step": 10681
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0001518826530612245,
+      "loss": 0.1316,
+      "step": 10682
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0001518673469387755,
+      "loss": 0.1618,
+      "step": 10683
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0001518520408163265,
+      "loss": 0.1803,
+      "step": 10684
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00015183673469387751,
+      "loss": 0.0603,
+      "step": 10685
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00015182142857142857,
+      "loss": 0.0838,
+      "step": 10686
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00015180612244897958,
+      "loss": 0.0586,
+      "step": 10687
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.0001517908163265306,
+      "loss": 0.2212,
+      "step": 10688
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.0001517755102040816,
+      "loss": 0.1665,
+      "step": 10689
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00015176020408163264,
+      "loss": 0.0829,
+      "step": 10690
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00015174489795918367,
+      "loss": 0.1064,
+      "step": 10691
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00015172959183673467,
+      "loss": 0.1128,
+      "step": 10692
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00015171428571428568,
+      "loss": 0.0979,
+      "step": 10693
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00015169897959183673,
+      "loss": 0.099,
+      "step": 10694
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00015168367346938774,
+      "loss": 0.0793,
+      "step": 10695
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00015166836734693877,
+      "loss": 0.1186,
+      "step": 10696
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00015165306122448977,
+      "loss": 0.0407,
+      "step": 10697
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0001516377551020408,
+      "loss": 0.0744,
+      "step": 10698
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00015162244897959183,
+      "loss": 0.0652,
+      "step": 10699
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00015160714285714284,
+      "loss": 0.2237,
+      "step": 10700
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00015159183673469384,
+      "loss": 0.1238,
+      "step": 10701
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0001515765306122449,
+      "loss": 0.1391,
+      "step": 10702
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0001515612244897959,
+      "loss": 0.1703,
+      "step": 10703
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00015154591836734693,
+      "loss": 0.1436,
+      "step": 10704
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00015153061224489794,
+      "loss": 0.155,
+      "step": 10705
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00015151530612244897,
+      "loss": 0.145,
+      "step": 10706
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.0001515,
+      "loss": 0.1414,
+      "step": 10707
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.000151484693877551,
+      "loss": 0.1229,
+      "step": 10708
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.000151469387755102,
+      "loss": 0.0987,
+      "step": 10709
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00015145408163265306,
+      "loss": 0.1335,
+      "step": 10710
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00015143877551020406,
+      "loss": 0.1028,
+      "step": 10711
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00015142346938775507,
+      "loss": 0.0703,
+      "step": 10712
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.0001514081632653061,
+      "loss": 0.3134,
+      "step": 10713
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00015139285714285713,
+      "loss": 0.112,
+      "step": 10714
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00015137755102040816,
+      "loss": 0.0764,
+      "step": 10715
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00015136224489795916,
+      "loss": 0.0818,
+      "step": 10716
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00015134693877551022,
+      "loss": 0.0985,
+      "step": 10717
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00015133163265306122,
+      "loss": 0.0515,
+      "step": 10718
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00015131632653061223,
+      "loss": 0.1866,
+      "step": 10719
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00015130102040816323,
+      "loss": 0.1956,
+      "step": 10720
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.0001512857142857143,
+      "loss": 0.0657,
+      "step": 10721
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.0001512704081632653,
+      "loss": 0.0653,
+      "step": 10722
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00015125510204081632,
+      "loss": 0.1025,
+      "step": 10723
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00015123979591836733,
+      "loss": 0.053,
+      "step": 10724
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00015122448979591836,
+      "loss": 0.1078,
+      "step": 10725
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.0001512091836734694,
+      "loss": 0.0915,
+      "step": 10726
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.0001511938775510204,
+      "loss": 0.1144,
+      "step": 10727
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.0001511785714285714,
+      "loss": 0.1095,
+      "step": 10728
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00015116326530612245,
+      "loss": 0.185,
+      "step": 10729
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00015114795918367345,
+      "loss": 0.1203,
+      "step": 10730
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00015113265306122449,
+      "loss": 0.0637,
+      "step": 10731
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.0001511173469387755,
+      "loss": 0.1985,
+      "step": 10732
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00015110204081632652,
+      "loss": 0.1043,
+      "step": 10733
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00015108673469387755,
+      "loss": 0.0711,
+      "step": 10734
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00015107142857142855,
+      "loss": 0.2237,
+      "step": 10735
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00015105612244897956,
+      "loss": 0.1007,
+      "step": 10736
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00015104081632653061,
+      "loss": 0.1234,
+      "step": 10737
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015102551020408162,
+      "loss": 0.0865,
+      "step": 10738
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015101020408163265,
+      "loss": 0.0913,
+      "step": 10739
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015099489795918365,
+      "loss": 0.1019,
+      "step": 10740
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015097959183673468,
+      "loss": 0.1459,
+      "step": 10741
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.0001509642857142857,
+      "loss": 0.0478,
+      "step": 10742
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015094897959183672,
+      "loss": 0.0763,
+      "step": 10743
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015093367346938772,
+      "loss": 0.0587,
+      "step": 10744
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015091836734693878,
+      "loss": 0.092,
+      "step": 10745
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015090306122448978,
+      "loss": 0.0666,
+      "step": 10746
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015088775510204078,
+      "loss": 0.0967,
+      "step": 10747
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00015087244897959181,
+      "loss": 0.1038,
+      "step": 10748
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015085714285714285,
+      "loss": 0.0802,
+      "step": 10749
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015084183673469388,
+      "loss": 0.1477,
+      "step": 10750
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015082653061224488,
+      "loss": 0.0877,
+      "step": 10751
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015081122448979588,
+      "loss": 0.0564,
+      "step": 10752
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015079591836734694,
+      "loss": 0.1463,
+      "step": 10753
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015078061224489794,
+      "loss": 0.1728,
+      "step": 10754
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015076530612244895,
+      "loss": 0.1839,
+      "step": 10755
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015074999999999998,
+      "loss": 0.1514,
+      "step": 10756
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.000150734693877551,
+      "loss": 0.303,
+      "step": 10757
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00015071938775510204,
+      "loss": 0.1143,
+      "step": 10758
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00015070408163265304,
+      "loss": 0.1501,
+      "step": 10759
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00015068877551020407,
+      "loss": 0.1765,
+      "step": 10760
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.0001506734693877551,
+      "loss": 0.1095,
+      "step": 10761
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.0001506581632653061,
+      "loss": 0.0885,
+      "step": 10762
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.0001506428571428571,
+      "loss": 0.1688,
+      "step": 10763
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00015062755102040817,
+      "loss": 0.143,
+      "step": 10764
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00015061224489795917,
+      "loss": 0.091,
+      "step": 10765
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.0001505969387755102,
+      "loss": 0.1588,
+      "step": 10766
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.0001505816326530612,
+      "loss": 0.0693,
+      "step": 10767
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00015056632653061224,
+      "loss": 0.0659,
+      "step": 10768
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015055102040816327,
+      "loss": 0.2042,
+      "step": 10769
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015053571428571427,
+      "loss": 0.0753,
+      "step": 10770
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015052040816326527,
+      "loss": 0.1252,
+      "step": 10771
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015050510204081633,
+      "loss": 0.1575,
+      "step": 10772
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015048979591836733,
+      "loss": 0.0523,
+      "step": 10773
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015047448979591834,
+      "loss": 0.0683,
+      "step": 10774
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015045918367346937,
+      "loss": 0.1002,
+      "step": 10775
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.0001504438775510204,
+      "loss": 0.157,
+      "step": 10776
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015042857142857143,
+      "loss": 0.1139,
+      "step": 10777
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00015041326530612243,
+      "loss": 0.0768,
+      "step": 10778
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.00015039795918367344,
+      "loss": 0.1345,
+      "step": 10779
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0001503826530612245,
+      "loss": 0.0951,
+      "step": 10780
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0001503673469387755,
+      "loss": 0.1347,
+      "step": 10781
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0001503520408163265,
+      "loss": 0.2009,
+      "step": 10782
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.00015033673469387753,
+      "loss": 0.2034,
+      "step": 10783
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.00015032142857142856,
+      "loss": 0.0713,
+      "step": 10784
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0001503061224489796,
+      "loss": 0.0657,
+      "step": 10785
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0001502908163265306,
+      "loss": 0.1125,
+      "step": 10786
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0001502755102040816,
+      "loss": 0.1301,
+      "step": 10787
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.00015026020408163266,
+      "loss": 0.2178,
+      "step": 10788
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.00015024489795918366,
+      "loss": 0.1542,
+      "step": 10789
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015022959183673466,
+      "loss": 0.1468,
+      "step": 10790
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.0001502142857142857,
+      "loss": 0.1189,
+      "step": 10791
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015019897959183673,
+      "loss": 0.1162,
+      "step": 10792
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015018367346938776,
+      "loss": 0.0534,
+      "step": 10793
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015016836734693876,
+      "loss": 0.1393,
+      "step": 10794
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015015306122448976,
+      "loss": 0.077,
+      "step": 10795
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015013775510204082,
+      "loss": 0.0713,
+      "step": 10796
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015012244897959182,
+      "loss": 0.0692,
+      "step": 10797
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015010714285714283,
+      "loss": 0.0913,
+      "step": 10798
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00015009183673469386,
+      "loss": 0.1498,
+      "step": 10799
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.0001500765306122449,
+      "loss": 0.21,
+      "step": 10800
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00015006122448979592,
+      "loss": 0.086,
+      "step": 10801
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00015004591836734692,
+      "loss": 0.0993,
+      "step": 10802
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00015003061224489793,
+      "loss": 0.0791,
+      "step": 10803
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00015001530612244898,
+      "loss": 0.1013,
+      "step": 10804
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00015,
+      "loss": 0.1659,
+      "step": 10805
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00014998469387755102,
+      "loss": 0.1183,
+      "step": 10806
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00014996938775510202,
+      "loss": 0.1129,
+      "step": 10807
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00014995408163265305,
+      "loss": 0.2355,
+      "step": 10808
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00014993877551020405,
+      "loss": 0.0829,
+      "step": 10809
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014992346938775509,
+      "loss": 0.0874,
+      "step": 10810
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014990816326530612,
+      "loss": 0.0914,
+      "step": 10811
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014989285714285715,
+      "loss": 0.0668,
+      "step": 10812
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014987755102040815,
+      "loss": 0.1951,
+      "step": 10813
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014986224489795918,
+      "loss": 0.1453,
+      "step": 10814
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014984693877551018,
+      "loss": 0.103,
+      "step": 10815
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014983163265306121,
+      "loss": 0.1595,
+      "step": 10816
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014981632653061222,
+      "loss": 0.0488,
+      "step": 10817
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014980102040816325,
+      "loss": 0.1645,
+      "step": 10818
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00014978571428571428,
+      "loss": 0.1733,
+      "step": 10819
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0001497704081632653,
+      "loss": 0.0839,
+      "step": 10820
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.0001497551020408163,
+      "loss": 0.1101,
+      "step": 10821
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00014973979591836734,
+      "loss": 0.1573,
+      "step": 10822
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00014972448979591835,
+      "loss": 0.2233,
+      "step": 10823
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00014970918367346938,
+      "loss": 0.0928,
+      "step": 10824
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00014969387755102038,
+      "loss": 0.3968,
+      "step": 10825
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.0001496785714285714,
+      "loss": 0.2526,
+      "step": 10826
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00014966326530612244,
+      "loss": 0.0774,
+      "step": 10827
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00014964795918367347,
+      "loss": 0.1005,
+      "step": 10828
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00014963265306122448,
+      "loss": 0.1546,
+      "step": 10829
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.0001496173469387755,
+      "loss": 0.098,
+      "step": 10830
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.0001496020408163265,
+      "loss": 0.1267,
+      "step": 10831
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00014958673469387754,
+      "loss": 0.1956,
+      "step": 10832
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00014957142857142854,
+      "loss": 0.102,
+      "step": 10833
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00014955612244897957,
+      "loss": 0.1046,
+      "step": 10834
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.0001495408163265306,
+      "loss": 0.1255,
+      "step": 10835
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00014952551020408164,
+      "loss": 0.0826,
+      "step": 10836
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00014951020408163264,
+      "loss": 0.0593,
+      "step": 10837
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00014949489795918367,
+      "loss": 0.1561,
+      "step": 10838
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00014947959183673467,
+      "loss": 0.1027,
+      "step": 10839
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.0001494642857142857,
+      "loss": 0.0958,
+      "step": 10840
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.0001494489795918367,
+      "loss": 0.0526,
+      "step": 10841
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00014943367346938774,
+      "loss": 0.0573,
+      "step": 10842
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00014941836734693877,
+      "loss": 0.149,
+      "step": 10843
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00014940306122448977,
+      "loss": 0.1566,
+      "step": 10844
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.0001493877551020408,
+      "loss": 0.1498,
+      "step": 10845
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00014937244897959183,
+      "loss": 0.1106,
+      "step": 10846
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00014935714285714284,
+      "loss": 0.1025,
+      "step": 10847
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00014934183673469387,
+      "loss": 0.1036,
+      "step": 10848
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00014932653061224487,
+      "loss": 0.1277,
+      "step": 10849
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.0001493112244897959,
+      "loss": 0.1872,
+      "step": 10850
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00014929591836734693,
+      "loss": 0.1157,
+      "step": 10851
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00014928061224489793,
+      "loss": 0.1585,
+      "step": 10852
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00014926530612244897,
+      "loss": 0.1011,
+      "step": 10853
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00014925,
+      "loss": 0.1514,
+      "step": 10854
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00014923469387755103,
+      "loss": 0.0979,
+      "step": 10855
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00014921938775510203,
+      "loss": 0.1004,
+      "step": 10856
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00014920408163265306,
+      "loss": 0.1961,
+      "step": 10857
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00014918877551020406,
+      "loss": 0.1154,
+      "step": 10858
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.0001491734693877551,
+      "loss": 0.0838,
+      "step": 10859
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.0001491581632653061,
+      "loss": 0.1018,
+      "step": 10860
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00014914285714285713,
+      "loss": 0.1054,
+      "step": 10861
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00014912755102040816,
+      "loss": 0.1552,
+      "step": 10862
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.0001491122448979592,
+      "loss": 0.2209,
+      "step": 10863
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.0001490969387755102,
+      "loss": 0.1986,
+      "step": 10864
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00014908163265306122,
+      "loss": 0.1192,
+      "step": 10865
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00014906632653061223,
+      "loss": 0.0521,
+      "step": 10866
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00014905102040816326,
+      "loss": 0.0708,
+      "step": 10867
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00014903571428571426,
+      "loss": 0.052,
+      "step": 10868
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.0001490204081632653,
+      "loss": 0.1649,
+      "step": 10869
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00014900510204081632,
+      "loss": 0.1245,
+      "step": 10870
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00014898979591836733,
+      "loss": 0.0794,
+      "step": 10871
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014897448979591836,
+      "loss": 0.1015,
+      "step": 10872
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014895918367346939,
+      "loss": 0.0686,
+      "step": 10873
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.0001489438775510204,
+      "loss": 0.1703,
+      "step": 10874
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014892857142857142,
+      "loss": 0.1399,
+      "step": 10875
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014891326530612242,
+      "loss": 0.1663,
+      "step": 10876
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014889795918367345,
+      "loss": 0.1203,
+      "step": 10877
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014888265306122448,
+      "loss": 0.1209,
+      "step": 10878
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.0001488673469387755,
+      "loss": 0.0571,
+      "step": 10879
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014885204081632652,
+      "loss": 0.0715,
+      "step": 10880
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014883673469387755,
+      "loss": 0.113,
+      "step": 10881
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014882142857142855,
+      "loss": 0.1582,
+      "step": 10882
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014880612244897958,
+      "loss": 0.1484,
+      "step": 10883
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.0001487908163265306,
+      "loss": 0.201,
+      "step": 10884
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014877551020408162,
+      "loss": 0.068,
+      "step": 10885
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014876020408163262,
+      "loss": 0.1583,
+      "step": 10886
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014874489795918365,
+      "loss": 0.1238,
+      "step": 10887
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014872959183673468,
+      "loss": 0.116,
+      "step": 10888
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.0001487142857142857,
+      "loss": 0.0806,
+      "step": 10889
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014869897959183672,
+      "loss": 0.1092,
+      "step": 10890
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014868367346938775,
+      "loss": 0.1285,
+      "step": 10891
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014866836734693875,
+      "loss": 0.1869,
+      "step": 10892
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00014865306122448978,
+      "loss": 0.0473,
+      "step": 10893
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00014863775510204078,
+      "loss": 0.1161,
+      "step": 10894
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00014862244897959181,
+      "loss": 0.0947,
+      "step": 10895
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00014860714285714284,
+      "loss": 0.0891,
+      "step": 10896
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00014859183673469388,
+      "loss": 0.1455,
+      "step": 10897
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.0001485765306122449,
+      "loss": 0.1124,
+      "step": 10898
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.0001485612244897959,
+      "loss": 0.1186,
+      "step": 10899
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00014854591836734694,
+      "loss": 0.0956,
+      "step": 10900
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00014853061224489794,
+      "loss": 0.1661,
+      "step": 10901
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00014851530612244897,
+      "loss": 0.1165,
+      "step": 10902
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00014849999999999998,
+      "loss": 0.2466,
+      "step": 10903
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.000148484693877551,
+      "loss": 0.0514,
+      "step": 10904
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00014846938775510204,
+      "loss": 0.1809,
+      "step": 10905
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00014845408163265304,
+      "loss": 0.1203,
+      "step": 10906
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00014843877551020407,
+      "loss": 0.1708,
+      "step": 10907
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.0001484234693877551,
+      "loss": 0.0731,
+      "step": 10908
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.0001484081632653061,
+      "loss": 0.1144,
+      "step": 10909
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00014839285714285714,
+      "loss": 0.1209,
+      "step": 10910
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00014837755102040814,
+      "loss": 0.0884,
+      "step": 10911
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00014836224489795917,
+      "loss": 0.08,
+      "step": 10912
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00014834693877551017,
+      "loss": 0.187,
+      "step": 10913
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.0001483316326530612,
+      "loss": 0.1331,
+      "step": 10914
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00014831632653061224,
+      "loss": 0.0709,
+      "step": 10915
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00014830102040816327,
+      "loss": 0.113,
+      "step": 10916
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00014828571428571427,
+      "loss": 0.1769,
+      "step": 10917
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.0001482704081632653,
+      "loss": 0.0942,
+      "step": 10918
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.0001482551020408163,
+      "loss": 0.2486,
+      "step": 10919
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00014823979591836733,
+      "loss": 0.2078,
+      "step": 10920
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00014822448979591834,
+      "loss": 0.1005,
+      "step": 10921
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00014820918367346937,
+      "loss": 0.0587,
+      "step": 10922
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.0001481938775510204,
+      "loss": 0.0609,
+      "step": 10923
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00014817857142857143,
+      "loss": 0.0555,
+      "step": 10924
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00014816326530612243,
+      "loss": 0.0966,
+      "step": 10925
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00014814795918367346,
+      "loss": 0.0791,
+      "step": 10926
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00014813265306122447,
+      "loss": 0.1261,
+      "step": 10927
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0001481173469387755,
+      "loss": 0.0871,
+      "step": 10928
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0001481020408163265,
+      "loss": 0.1214,
+      "step": 10929
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00014808673469387753,
+      "loss": 0.0555,
+      "step": 10930
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00014807142857142856,
+      "loss": 0.0904,
+      "step": 10931
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0001480561224489796,
+      "loss": 0.1249,
+      "step": 10932
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0001480408163265306,
+      "loss": 0.2627,
+      "step": 10933
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00014802551020408163,
+      "loss": 0.0823,
+      "step": 10934
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00014801020408163263,
+      "loss": 0.053,
+      "step": 10935
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00014799489795918366,
+      "loss": 0.1596,
+      "step": 10936
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00014797959183673466,
+      "loss": 0.1031,
+      "step": 10937
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.0001479642857142857,
+      "loss": 0.1854,
+      "step": 10938
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00014794897959183672,
+      "loss": 0.1209,
+      "step": 10939
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00014793367346938776,
+      "loss": 0.1562,
+      "step": 10940
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00014791836734693876,
+      "loss": 0.0973,
+      "step": 10941
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.0001479030612244898,
+      "loss": 0.0825,
+      "step": 10942
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00014788775510204082,
+      "loss": 0.0398,
+      "step": 10943
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014787244897959182,
+      "loss": 0.1911,
+      "step": 10944
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014785714285714285,
+      "loss": 0.2285,
+      "step": 10945
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014784183673469386,
+      "loss": 0.1635,
+      "step": 10946
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.0001478265306122449,
+      "loss": 0.1708,
+      "step": 10947
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.0001478112244897959,
+      "loss": 0.0528,
+      "step": 10948
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014779591836734692,
+      "loss": 0.0974,
+      "step": 10949
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014778061224489795,
+      "loss": 0.209,
+      "step": 10950
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014776530612244898,
+      "loss": 0.1522,
+      "step": 10951
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014774999999999999,
+      "loss": 0.0631,
+      "step": 10952
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014773469387755102,
+      "loss": 0.0678,
+      "step": 10953
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00014771938775510202,
+      "loss": 0.1439,
+      "step": 10954
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014770408163265305,
+      "loss": 0.0446,
+      "step": 10955
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014768877551020405,
+      "loss": 0.0609,
+      "step": 10956
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014767346938775508,
+      "loss": 0.1528,
+      "step": 10957
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014765816326530612,
+      "loss": 0.0618,
+      "step": 10958
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014764285714285715,
+      "loss": 0.1455,
+      "step": 10959
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014762755102040815,
+      "loss": 0.1256,
+      "step": 10960
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014761224489795918,
+      "loss": 0.091,
+      "step": 10961
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014759693877551018,
+      "loss": 0.0847,
+      "step": 10962
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014758163265306121,
+      "loss": 0.1561,
+      "step": 10963
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014756632653061222,
+      "loss": 0.0887,
+      "step": 10964
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00014755102040816325,
+      "loss": 0.1571,
+      "step": 10965
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00014753571428571428,
+      "loss": 0.1433,
+      "step": 10966
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.0001475204081632653,
+      "loss": 0.0956,
+      "step": 10967
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.0001475051020408163,
+      "loss": 0.1622,
+      "step": 10968
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00014748979591836734,
+      "loss": 0.2006,
+      "step": 10969
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00014747448979591835,
+      "loss": 0.1421,
+      "step": 10970
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00014745918367346938,
+      "loss": 0.054,
+      "step": 10971
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00014744387755102038,
+      "loss": 0.1163,
+      "step": 10972
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.0001474285714285714,
+      "loss": 0.061,
+      "step": 10973
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00014741326530612244,
+      "loss": 0.0901,
+      "step": 10974
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00014739795918367347,
+      "loss": 0.1265,
+      "step": 10975
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00014738265306122448,
+      "loss": 0.1277,
+      "step": 10976
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0001473673469387755,
+      "loss": 0.1169,
+      "step": 10977
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0001473520408163265,
+      "loss": 0.1115,
+      "step": 10978
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00014733673469387754,
+      "loss": 0.1516,
+      "step": 10979
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00014732142857142854,
+      "loss": 0.0349,
+      "step": 10980
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00014730612244897957,
+      "loss": 0.1694,
+      "step": 10981
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0001472908163265306,
+      "loss": 0.2145,
+      "step": 10982
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0001472755102040816,
+      "loss": 0.1173,
+      "step": 10983
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00014726020408163264,
+      "loss": 0.059,
+      "step": 10984
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014724489795918367,
+      "loss": 0.1045,
+      "step": 10985
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014722959183673467,
+      "loss": 0.0682,
+      "step": 10986
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.0001472142857142857,
+      "loss": 0.0789,
+      "step": 10987
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014719897959183673,
+      "loss": 0.2011,
+      "step": 10988
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014718367346938774,
+      "loss": 0.139,
+      "step": 10989
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014716836734693877,
+      "loss": 0.1123,
+      "step": 10990
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014715306122448977,
+      "loss": 0.1133,
+      "step": 10991
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.0001471377551020408,
+      "loss": 0.1016,
+      "step": 10992
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014712244897959183,
+      "loss": 0.1185,
+      "step": 10993
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014710714285714286,
+      "loss": 0.1642,
+      "step": 10994
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00014709183673469387,
+      "loss": 0.2916,
+      "step": 10995
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.0001470765306122449,
+      "loss": 0.084,
+      "step": 10996
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.0001470612244897959,
+      "loss": 0.0703,
+      "step": 10997
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00014704591836734693,
+      "loss": 0.1123,
+      "step": 10998
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00014703061224489793,
+      "loss": 0.0671,
+      "step": 10999
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00014701530612244896,
+      "loss": 0.1665,
+      "step": 11000
+    },
+    {
+      "epoch": 10.68,
+      "eval_cer": 0.2507382615756717,
+      "eval_loss": 0.9760210514068604,
+      "eval_runtime": 236.8511,
+      "eval_samples_per_second": 37.213,
+      "eval_steps_per_second": 4.653,
+      "eval_wer": 0.6595974909094904,
+      "step": 11000
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.000147,
+      "loss": 0.14,
+      "step": 11001
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00014698469387755103,
+      "loss": 0.1143,
+      "step": 11002
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00014696938775510203,
+      "loss": 0.0439,
+      "step": 11003
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00014695408163265306,
+      "loss": 0.1115,
+      "step": 11004
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00014693877551020406,
+      "loss": 0.0886,
+      "step": 11005
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.0001469234693877551,
+      "loss": 0.1318,
+      "step": 11006
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.0001469081632653061,
+      "loss": 0.1235,
+      "step": 11007
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00014689285714285713,
+      "loss": 0.1558,
+      "step": 11008
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00014687755102040816,
+      "loss": 0.101,
+      "step": 11009
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00014686224489795916,
+      "loss": 0.0934,
+      "step": 11010
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.0001468469387755102,
+      "loss": 0.1001,
+      "step": 11011
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00014683163265306122,
+      "loss": 0.0619,
+      "step": 11012
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00014681632653061223,
+      "loss": 0.1308,
+      "step": 11013
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00014680102040816326,
+      "loss": 0.1368,
+      "step": 11014
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00014678571428571426,
+      "loss": 0.0776,
+      "step": 11015
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.0001467704081632653,
+      "loss": 0.1453,
+      "step": 11016
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014675510204081632,
+      "loss": 0.0655,
+      "step": 11017
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014673979591836732,
+      "loss": 0.0653,
+      "step": 11018
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014672448979591836,
+      "loss": 0.1386,
+      "step": 11019
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014670918367346939,
+      "loss": 0.0777,
+      "step": 11020
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.0001466938775510204,
+      "loss": 0.108,
+      "step": 11021
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014667857142857142,
+      "loss": 0.1449,
+      "step": 11022
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014666326530612242,
+      "loss": 0.0797,
+      "step": 11023
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014664795918367345,
+      "loss": 0.0485,
+      "step": 11024
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014663265306122446,
+      "loss": 0.0573,
+      "step": 11025
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.0001466173469387755,
+      "loss": 0.1992,
+      "step": 11026
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014660204081632652,
+      "loss": 0.1461,
+      "step": 11027
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014658673469387755,
+      "loss": 0.1196,
+      "step": 11028
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014657142857142855,
+      "loss": 0.0665,
+      "step": 11029
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014655612244897958,
+      "loss": 0.0655,
+      "step": 11030
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014654081632653059,
+      "loss": 0.1248,
+      "step": 11031
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014652551020408162,
+      "loss": 0.212,
+      "step": 11032
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014651020408163265,
+      "loss": 0.1983,
+      "step": 11033
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014649489795918365,
+      "loss": 0.0689,
+      "step": 11034
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00014647959183673468,
+      "loss": 0.1044,
+      "step": 11035
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.0001464642857142857,
+      "loss": 0.0995,
+      "step": 11036
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00014644897959183674,
+      "loss": 0.0674,
+      "step": 11037
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00014643367346938775,
+      "loss": 0.1359,
+      "step": 11038
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00014641836734693878,
+      "loss": 0.2052,
+      "step": 11039
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00014640306122448978,
+      "loss": 0.0764,
+      "step": 11040
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.0001463877551020408,
+      "loss": 0.1532,
+      "step": 11041
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00014637244897959181,
+      "loss": 0.1279,
+      "step": 11042
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00014635714285714284,
+      "loss": 0.1001,
+      "step": 11043
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00014634183673469388,
+      "loss": 0.2581,
+      "step": 11044
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00014632653061224488,
+      "loss": 0.1635,
+      "step": 11045
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.0001463112244897959,
+      "loss": 0.187,
+      "step": 11046
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00014629591836734694,
+      "loss": 0.0815,
+      "step": 11047
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00014628061224489794,
+      "loss": 0.1339,
+      "step": 11048
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00014626530612244897,
+      "loss": 0.045,
+      "step": 11049
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00014624999999999998,
+      "loss": 0.179,
+      "step": 11050
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.000146234693877551,
+      "loss": 0.1455,
+      "step": 11051
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.000146219387755102,
+      "loss": 0.0725,
+      "step": 11052
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00014620408163265304,
+      "loss": 0.064,
+      "step": 11053
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00014618877551020407,
+      "loss": 0.0874,
+      "step": 11054
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.0001461734693877551,
+      "loss": 0.1043,
+      "step": 11055
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.0001461581632653061,
+      "loss": 0.0986,
+      "step": 11056
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00014614285714285714,
+      "loss": 0.1091,
+      "step": 11057
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00014612755102040814,
+      "loss": 0.1273,
+      "step": 11058
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00014611224489795917,
+      "loss": 0.1307,
+      "step": 11059
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00014609693877551017,
+      "loss": 0.0696,
+      "step": 11060
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.0001460816326530612,
+      "loss": 0.0492,
+      "step": 11061
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00014606632653061224,
+      "loss": 0.0616,
+      "step": 11062
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00014605102040816327,
+      "loss": 0.2351,
+      "step": 11063
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00014603571428571427,
+      "loss": 0.1166,
+      "step": 11064
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.0001460204081632653,
+      "loss": 0.108,
+      "step": 11065
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.0001460051020408163,
+      "loss": 0.1403,
+      "step": 11066
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00014598979591836733,
+      "loss": 0.1084,
+      "step": 11067
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00014597448979591834,
+      "loss": 0.0626,
+      "step": 11068
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00014595918367346937,
+      "loss": 0.311,
+      "step": 11069
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.0001459438775510204,
+      "loss": 0.0852,
+      "step": 11070
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00014592857142857143,
+      "loss": 0.0774,
+      "step": 11071
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00014591326530612243,
+      "loss": 0.1585,
+      "step": 11072
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00014589795918367346,
+      "loss": 0.0454,
+      "step": 11073
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00014588265306122447,
+      "loss": 0.1039,
+      "step": 11074
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.0001458673469387755,
+      "loss": 0.1341,
+      "step": 11075
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.0001458520408163265,
+      "loss": 0.1214,
+      "step": 11076
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00014583673469387753,
+      "loss": 0.1332,
+      "step": 11077
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00014582142857142856,
+      "loss": 0.0814,
+      "step": 11078
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.0001458061224489796,
+      "loss": 0.0902,
+      "step": 11079
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.0001457908163265306,
+      "loss": 0.0913,
+      "step": 11080
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00014577551020408163,
+      "loss": 0.103,
+      "step": 11081
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00014576020408163266,
+      "loss": 0.1234,
+      "step": 11082
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00014574489795918366,
+      "loss": 0.1545,
+      "step": 11083
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.0001457295918367347,
+      "loss": 0.0692,
+      "step": 11084
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.0001457142857142857,
+      "loss": 0.0901,
+      "step": 11085
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00014569897959183672,
+      "loss": 0.1154,
+      "step": 11086
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00014568367346938773,
+      "loss": 0.1054,
+      "step": 11087
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00014566836734693876,
+      "loss": 0.2059,
+      "step": 11088
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.0001456530612244898,
+      "loss": 0.1665,
+      "step": 11089
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00014563775510204082,
+      "loss": 0.1618,
+      "step": 11090
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00014562244897959182,
+      "loss": 0.0924,
+      "step": 11091
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00014560714285714285,
+      "loss": 0.0613,
+      "step": 11092
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00014559183673469386,
+      "loss": 0.1413,
+      "step": 11093
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.0001455765306122449,
+      "loss": 0.1491,
+      "step": 11094
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.0001455612244897959,
+      "loss": 0.0708,
+      "step": 11095
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00014554591836734692,
+      "loss": 0.0787,
+      "step": 11096
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00014553061224489795,
+      "loss": 0.0695,
+      "step": 11097
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00014551530612244898,
+      "loss": 0.101,
+      "step": 11098
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014549999999999999,
+      "loss": 0.0649,
+      "step": 11099
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014548469387755102,
+      "loss": 0.2043,
+      "step": 11100
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014546938775510202,
+      "loss": 0.173,
+      "step": 11101
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014545408163265305,
+      "loss": 0.1012,
+      "step": 11102
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014543877551020405,
+      "loss": 0.0789,
+      "step": 11103
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014542346938775508,
+      "loss": 0.0554,
+      "step": 11104
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014540816326530611,
+      "loss": 0.06,
+      "step": 11105
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014539285714285715,
+      "loss": 0.1435,
+      "step": 11106
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014537755102040815,
+      "loss": 0.1368,
+      "step": 11107
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014536224489795918,
+      "loss": 0.1195,
+      "step": 11108
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014534693877551018,
+      "loss": 0.1816,
+      "step": 11109
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.0001453316326530612,
+      "loss": 0.0797,
+      "step": 11110
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014531632653061222,
+      "loss": 0.0507,
+      "step": 11111
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014530102040816325,
+      "loss": 0.0722,
+      "step": 11112
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014528571428571428,
+      "loss": 0.1581,
+      "step": 11113
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014527040816326528,
+      "loss": 0.098,
+      "step": 11114
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.0001452551020408163,
+      "loss": 0.0965,
+      "step": 11115
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014523979591836734,
+      "loss": 0.0738,
+      "step": 11116
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014522448979591835,
+      "loss": 0.1103,
+      "step": 11117
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014520918367346938,
+      "loss": 0.0916,
+      "step": 11118
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014519387755102038,
+      "loss": 0.167,
+      "step": 11119
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.0001451785714285714,
+      "loss": 0.1628,
+      "step": 11120
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014516326530612244,
+      "loss": 0.0848,
+      "step": 11121
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014514795918367344,
+      "loss": 0.0725,
+      "step": 11122
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014513265306122448,
+      "loss": 0.1103,
+      "step": 11123
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.0001451173469387755,
+      "loss": 0.0596,
+      "step": 11124
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014510204081632654,
+      "loss": 0.0816,
+      "step": 11125
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014508673469387754,
+      "loss": 0.137,
+      "step": 11126
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014507142857142857,
+      "loss": 0.1166,
+      "step": 11127
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014505612244897957,
+      "loss": 0.0907,
+      "step": 11128
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.0001450408163265306,
+      "loss": 0.0608,
+      "step": 11129
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.0001450255102040816,
+      "loss": 0.0426,
+      "step": 11130
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00014501020408163264,
+      "loss": 0.0979,
+      "step": 11131
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00014499489795918367,
+      "loss": 0.1216,
+      "step": 11132
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.0001449795918367347,
+      "loss": 0.1169,
+      "step": 11133
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.0001449642857142857,
+      "loss": 0.0824,
+      "step": 11134
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00014494897959183673,
+      "loss": 0.1986,
+      "step": 11135
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00014493367346938774,
+      "loss": 0.128,
+      "step": 11136
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00014491836734693877,
+      "loss": 0.0584,
+      "step": 11137
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00014490306122448977,
+      "loss": 0.3607,
+      "step": 11138
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.0001448877551020408,
+      "loss": 0.1805,
+      "step": 11139
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00014487244897959183,
+      "loss": 0.1403,
+      "step": 11140
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00014485714285714286,
+      "loss": 0.0945,
+      "step": 11141
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00014484183673469387,
+      "loss": 0.1367,
+      "step": 11142
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.0001448265306122449,
+      "loss": 0.1039,
+      "step": 11143
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.0001448112244897959,
+      "loss": 0.1351,
+      "step": 11144
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00014479591836734693,
+      "loss": 0.1649,
+      "step": 11145
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00014478061224489793,
+      "loss": 0.0821,
+      "step": 11146
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00014476530612244896,
+      "loss": 0.0713,
+      "step": 11147
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00014475,
+      "loss": 0.0934,
+      "step": 11148
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.000144734693877551,
+      "loss": 0.0684,
+      "step": 11149
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00014471938775510203,
+      "loss": 0.1451,
+      "step": 11150
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00014470408163265306,
+      "loss": 0.145,
+      "step": 11151
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00014468877551020406,
+      "loss": 0.1945,
+      "step": 11152
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0001446734693877551,
+      "loss": 0.1517,
+      "step": 11153
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0001446581632653061,
+      "loss": 0.075,
+      "step": 11154
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00014464285714285713,
+      "loss": 0.0892,
+      "step": 11155
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00014462755102040816,
+      "loss": 0.0727,
+      "step": 11156
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00014461224489795916,
+      "loss": 0.1864,
+      "step": 11157
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0001445969387755102,
+      "loss": 0.0776,
+      "step": 11158
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00014458163265306122,
+      "loss": 0.048,
+      "step": 11159
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00014456632653061223,
+      "loss": 0.1333,
+      "step": 11160
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014455102040816326,
+      "loss": 0.1507,
+      "step": 11161
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014453571428571426,
+      "loss": 0.2092,
+      "step": 11162
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.0001445204081632653,
+      "loss": 0.2221,
+      "step": 11163
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.0001445051020408163,
+      "loss": 0.1937,
+      "step": 11164
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014448979591836732,
+      "loss": 0.0699,
+      "step": 11165
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014447448979591835,
+      "loss": 0.0979,
+      "step": 11166
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014445918367346939,
+      "loss": 0.0553,
+      "step": 11167
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014444387755102042,
+      "loss": 0.1026,
+      "step": 11168
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014442857142857142,
+      "loss": 0.1661,
+      "step": 11169
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014441326530612245,
+      "loss": 0.198,
+      "step": 11170
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014439795918367345,
+      "loss": 0.066,
+      "step": 11171
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014438265306122448,
+      "loss": 0.0657,
+      "step": 11172
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.0001443673469387755,
+      "loss": 0.1389,
+      "step": 11173
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014435204081632652,
+      "loss": 0.0855,
+      "step": 11174
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014433673469387755,
+      "loss": 0.1008,
+      "step": 11175
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014432142857142858,
+      "loss": 0.1086,
+      "step": 11176
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014430612244897958,
+      "loss": 0.1302,
+      "step": 11177
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.0001442908163265306,
+      "loss": 0.0758,
+      "step": 11178
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014427551020408162,
+      "loss": 0.0788,
+      "step": 11179
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014426020408163265,
+      "loss": 0.1578,
+      "step": 11180
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00014424489795918365,
+      "loss": 0.1057,
+      "step": 11181
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00014422959183673468,
+      "loss": 0.2424,
+      "step": 11182
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.0001442142857142857,
+      "loss": 0.1062,
+      "step": 11183
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00014419897959183671,
+      "loss": 0.1049,
+      "step": 11184
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00014418367346938775,
+      "loss": 0.0929,
+      "step": 11185
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00014416836734693878,
+      "loss": 0.1043,
+      "step": 11186
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00014415306122448978,
+      "loss": 0.1316,
+      "step": 11187
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.0001441377551020408,
+      "loss": 0.1476,
+      "step": 11188
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.0001441224489795918,
+      "loss": 0.1397,
+      "step": 11189
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00014410714285714284,
+      "loss": 0.1061,
+      "step": 11190
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00014409183673469385,
+      "loss": 0.0519,
+      "step": 11191
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00014407653061224488,
+      "loss": 0.1121,
+      "step": 11192
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.0001440612244897959,
+      "loss": 0.0853,
+      "step": 11193
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00014404591836734694,
+      "loss": 0.1105,
+      "step": 11194
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00014403061224489794,
+      "loss": 0.1142,
+      "step": 11195
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00014401530612244897,
+      "loss": 0.0683,
+      "step": 11196
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00014399999999999998,
+      "loss": 0.1054,
+      "step": 11197
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.000143984693877551,
+      "loss": 0.0636,
+      "step": 11198
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.000143969387755102,
+      "loss": 0.0515,
+      "step": 11199
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00014395408163265304,
+      "loss": 0.0553,
+      "step": 11200
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00014393877551020407,
+      "loss": 0.1963,
+      "step": 11201
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.0001439234693877551,
+      "loss": 0.1578,
+      "step": 11202
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.0001439081632653061,
+      "loss": 0.056,
+      "step": 11203
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00014389285714285714,
+      "loss": 0.0455,
+      "step": 11204
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00014387755102040814,
+      "loss": 0.0884,
+      "step": 11205
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00014386224489795917,
+      "loss": 0.1696,
+      "step": 11206
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00014384693877551017,
+      "loss": 0.0841,
+      "step": 11207
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.0001438316326530612,
+      "loss": 0.0992,
+      "step": 11208
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00014381632653061223,
+      "loss": 0.0562,
+      "step": 11209
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00014380102040816327,
+      "loss": 0.0505,
+      "step": 11210
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00014378571428571427,
+      "loss": 0.0782,
+      "step": 11211
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.0001437704081632653,
+      "loss": 0.0733,
+      "step": 11212
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.0001437551020408163,
+      "loss": 0.1531,
+      "step": 11213
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00014373979591836733,
+      "loss": 0.1654,
+      "step": 11214
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00014372448979591836,
+      "loss": 0.1155,
+      "step": 11215
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00014370918367346937,
+      "loss": 0.1403,
+      "step": 11216
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.0001436938775510204,
+      "loss": 0.0657,
+      "step": 11217
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00014367857142857143,
+      "loss": 0.1486,
+      "step": 11218
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00014366326530612243,
+      "loss": 0.1059,
+      "step": 11219
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00014364795918367346,
+      "loss": 0.1205,
+      "step": 11220
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.0001436326530612245,
+      "loss": 0.0823,
+      "step": 11221
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0001436173469387755,
+      "loss": 0.1028,
+      "step": 11222
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00014360204081632653,
+      "loss": 0.144,
+      "step": 11223
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00014358673469387753,
+      "loss": 0.0571,
+      "step": 11224
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00014357142857142856,
+      "loss": 0.087,
+      "step": 11225
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00014355612244897956,
+      "loss": 0.1524,
+      "step": 11226
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0001435408163265306,
+      "loss": 0.1343,
+      "step": 11227
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00014352551020408163,
+      "loss": 0.1146,
+      "step": 11228
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00014351020408163266,
+      "loss": 0.1795,
+      "step": 11229
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00014349489795918366,
+      "loss": 0.2491,
+      "step": 11230
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0001434795918367347,
+      "loss": 0.1533,
+      "step": 11231
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0001434642857142857,
+      "loss": 0.1767,
+      "step": 11232
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00014344897959183672,
+      "loss": 0.1151,
+      "step": 11233
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00014343367346938773,
+      "loss": 0.1878,
+      "step": 11234
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00014341836734693876,
+      "loss": 0.0986,
+      "step": 11235
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.0001434030612244898,
+      "loss": 0.0993,
+      "step": 11236
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00014338775510204082,
+      "loss": 0.1517,
+      "step": 11237
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00014337244897959182,
+      "loss": 0.1069,
+      "step": 11238
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00014335714285714285,
+      "loss": 0.1667,
+      "step": 11239
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00014334183673469386,
+      "loss": 0.1343,
+      "step": 11240
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.0001433265306122449,
+      "loss": 0.1225,
+      "step": 11241
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.0001433112244897959,
+      "loss": 0.1008,
+      "step": 11242
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014329591836734692,
+      "loss": 0.0677,
+      "step": 11243
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014328061224489795,
+      "loss": 0.2375,
+      "step": 11244
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014326530612244898,
+      "loss": 0.0935,
+      "step": 11245
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014324999999999999,
+      "loss": 0.1871,
+      "step": 11246
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014323469387755102,
+      "loss": 0.1432,
+      "step": 11247
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014321938775510202,
+      "loss": 0.0482,
+      "step": 11248
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014320408163265305,
+      "loss": 0.1032,
+      "step": 11249
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014318877551020405,
+      "loss": 0.0641,
+      "step": 11250
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014317346938775508,
+      "loss": 0.1648,
+      "step": 11251
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00014315816326530611,
+      "loss": 0.0947,
+      "step": 11252
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014314285714285712,
+      "loss": 0.1137,
+      "step": 11253
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014312755102040815,
+      "loss": 0.1676,
+      "step": 11254
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014311224489795918,
+      "loss": 0.0563,
+      "step": 11255
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014309693877551018,
+      "loss": 0.0971,
+      "step": 11256
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.0001430816326530612,
+      "loss": 0.1985,
+      "step": 11257
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014306632653061222,
+      "loss": 0.066,
+      "step": 11258
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014305102040816325,
+      "loss": 0.0687,
+      "step": 11259
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014303571428571428,
+      "loss": 0.1302,
+      "step": 11260
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014302040816326528,
+      "loss": 0.0721,
+      "step": 11261
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.0001430051020408163,
+      "loss": 0.1156,
+      "step": 11262
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00014298979591836734,
+      "loss": 0.1866,
+      "step": 11263
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00014297448979591837,
+      "loss": 0.0868,
+      "step": 11264
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00014295918367346938,
+      "loss": 0.0983,
+      "step": 11265
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.0001429438775510204,
+      "loss": 0.0908,
+      "step": 11266
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.0001429285714285714,
+      "loss": 0.0625,
+      "step": 11267
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00014291326530612244,
+      "loss": 0.0997,
+      "step": 11268
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00014289795918367344,
+      "loss": 0.2121,
+      "step": 11269
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00014288265306122447,
+      "loss": 0.1585,
+      "step": 11270
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.0001428673469387755,
+      "loss": 0.0754,
+      "step": 11271
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00014285204081632654,
+      "loss": 0.066,
+      "step": 11272
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00014283673469387754,
+      "loss": 0.0982,
+      "step": 11273
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00014282142857142857,
+      "loss": 0.1638,
+      "step": 11274
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00014280612244897957,
+      "loss": 0.1787,
+      "step": 11275
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.0001427908163265306,
+      "loss": 0.2075,
+      "step": 11276
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.0001427755102040816,
+      "loss": 0.1147,
+      "step": 11277
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00014276020408163264,
+      "loss": 0.1044,
+      "step": 11278
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00014274489795918367,
+      "loss": 0.1835,
+      "step": 11279
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.0001427295918367347,
+      "loss": 0.1005,
+      "step": 11280
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.0001427142857142857,
+      "loss": 0.0555,
+      "step": 11281
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00014269897959183673,
+      "loss": 0.1603,
+      "step": 11282
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00014268367346938774,
+      "loss": 0.1485,
+      "step": 11283
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00014266836734693877,
+      "loss": 0.074,
+      "step": 11284
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00014265306122448977,
+      "loss": 0.1977,
+      "step": 11285
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.0001426377551020408,
+      "loss": 0.077,
+      "step": 11286
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00014262244897959183,
+      "loss": 0.0645,
+      "step": 11287
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00014260714285714283,
+      "loss": 0.2148,
+      "step": 11288
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00014259183673469387,
+      "loss": 0.0743,
+      "step": 11289
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.0001425765306122449,
+      "loss": 0.1115,
+      "step": 11290
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.0001425612244897959,
+      "loss": 0.0612,
+      "step": 11291
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00014254591836734693,
+      "loss": 0.1222,
+      "step": 11292
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00014253061224489793,
+      "loss": 0.0541,
+      "step": 11293
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00014251530612244896,
+      "loss": 0.2523,
+      "step": 11294
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.0001425,
+      "loss": 0.1883,
+      "step": 11295
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.000142484693877551,
+      "loss": 0.1585,
+      "step": 11296
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00014246938775510203,
+      "loss": 0.0659,
+      "step": 11297
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00014245408163265306,
+      "loss": 0.1773,
+      "step": 11298
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00014243877551020406,
+      "loss": 0.098,
+      "step": 11299
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.0001424234693877551,
+      "loss": 0.1608,
+      "step": 11300
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.0001424081632653061,
+      "loss": 0.0763,
+      "step": 11301
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00014239285714285713,
+      "loss": 0.0501,
+      "step": 11302
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00014237755102040813,
+      "loss": 0.1127,
+      "step": 11303
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00014236224489795916,
+      "loss": 0.0804,
+      "step": 11304
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.0001423469387755102,
+      "loss": 0.0634,
+      "step": 11305
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00014233163265306122,
+      "loss": 0.0506,
+      "step": 11306
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00014231632653061225,
+      "loss": 0.1413,
+      "step": 11307
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00014230102040816326,
+      "loss": 0.0769,
+      "step": 11308
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00014228571428571429,
+      "loss": 0.0463,
+      "step": 11309
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.0001422704081632653,
+      "loss": 0.1313,
+      "step": 11310
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00014225510204081632,
+      "loss": 0.0688,
+      "step": 11311
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00014223979591836732,
+      "loss": 0.1847,
+      "step": 11312
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00014222448979591835,
+      "loss": 0.1565,
+      "step": 11313
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00014220918367346938,
+      "loss": 0.1539,
+      "step": 11314
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014219387755102042,
+      "loss": 0.0781,
+      "step": 11315
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014217857142857142,
+      "loss": 0.1112,
+      "step": 11316
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014216326530612245,
+      "loss": 0.1264,
+      "step": 11317
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014214795918367345,
+      "loss": 0.0563,
+      "step": 11318
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014213265306122448,
+      "loss": 0.1111,
+      "step": 11319
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.0001421173469387755,
+      "loss": 0.1127,
+      "step": 11320
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014210204081632652,
+      "loss": 0.0844,
+      "step": 11321
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014208673469387755,
+      "loss": 0.0537,
+      "step": 11322
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014207142857142855,
+      "loss": 0.1247,
+      "step": 11323
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00014205612244897958,
+      "loss": 0.0694,
+      "step": 11324
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.0001420408163265306,
+      "loss": 0.0929,
+      "step": 11325
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014202551020408162,
+      "loss": 0.1113,
+      "step": 11326
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014201020408163265,
+      "loss": 0.0695,
+      "step": 11327
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014199489795918365,
+      "loss": 0.0463,
+      "step": 11328
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014197959183673468,
+      "loss": 0.1029,
+      "step": 11329
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014196428571428568,
+      "loss": 0.1761,
+      "step": 11330
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014194897959183671,
+      "loss": 0.1285,
+      "step": 11331
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014193367346938775,
+      "loss": 0.1083,
+      "step": 11332
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014191836734693878,
+      "loss": 0.0557,
+      "step": 11333
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00014190306122448978,
+      "loss": 0.0896,
+      "step": 11334
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.0001418877551020408,
+      "loss": 0.0487,
+      "step": 11335
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.0001418724489795918,
+      "loss": 0.0631,
+      "step": 11336
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00014185714285714284,
+      "loss": 0.1356,
+      "step": 11337
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00014184183673469385,
+      "loss": 0.0687,
+      "step": 11338
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00014182653061224488,
+      "loss": 0.0414,
+      "step": 11339
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.0001418112244897959,
+      "loss": 0.0955,
+      "step": 11340
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00014179591836734694,
+      "loss": 0.0801,
+      "step": 11341
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00014178061224489794,
+      "loss": 0.0551,
+      "step": 11342
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00014176530612244897,
+      "loss": 0.1804,
+      "step": 11343
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00014174999999999998,
+      "loss": 0.0943,
+      "step": 11344
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.000141734693877551,
+      "loss": 0.0953,
+      "step": 11345
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.000141719387755102,
+      "loss": 0.1314,
+      "step": 11346
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.00014170408163265304,
+      "loss": 0.0491,
+      "step": 11347
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.00014168877551020407,
+      "loss": 0.0742,
+      "step": 11348
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0001416734693877551,
+      "loss": 0.1114,
+      "step": 11349
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0001416581632653061,
+      "loss": 0.0566,
+      "step": 11350
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.00014164285714285714,
+      "loss": 0.0767,
+      "step": 11351
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.00014162755102040817,
+      "loss": 0.1079,
+      "step": 11352
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.00014161224489795917,
+      "loss": 0.1729,
+      "step": 11353
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0001415969387755102,
+      "loss": 0.1077,
+      "step": 11354
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0001415816326530612,
+      "loss": 0.1467,
+      "step": 11355
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00014156632653061223,
+      "loss": 0.1602,
+      "step": 11356
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00014155102040816326,
+      "loss": 0.0583,
+      "step": 11357
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00014153571428571427,
+      "loss": 0.0881,
+      "step": 11358
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.0001415204081632653,
+      "loss": 0.1305,
+      "step": 11359
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00014150510204081633,
+      "loss": 0.0482,
+      "step": 11360
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00014148979591836733,
+      "loss": 0.1033,
+      "step": 11361
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00014147448979591836,
+      "loss": 0.1028,
+      "step": 11362
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00014145918367346937,
+      "loss": 0.1068,
+      "step": 11363
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.0001414438775510204,
+      "loss": 0.0716,
+      "step": 11364
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.0001414285714285714,
+      "loss": 0.099,
+      "step": 11365
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00014141326530612243,
+      "loss": 0.1843,
+      "step": 11366
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00014139795918367346,
+      "loss": 0.1416,
+      "step": 11367
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.0001413826530612245,
+      "loss": 0.1289,
+      "step": 11368
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.0001413673469387755,
+      "loss": 0.1253,
+      "step": 11369
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00014135204081632653,
+      "loss": 0.1533,
+      "step": 11370
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00014133673469387753,
+      "loss": 0.1686,
+      "step": 11371
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00014132142857142856,
+      "loss": 0.09,
+      "step": 11372
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00014130612244897956,
+      "loss": 0.0525,
+      "step": 11373
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.0001412908163265306,
+      "loss": 0.0849,
+      "step": 11374
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00014127551020408162,
+      "loss": 0.1381,
+      "step": 11375
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00014126020408163266,
+      "loss": 0.0642,
+      "step": 11376
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00014124489795918366,
+      "loss": 0.054,
+      "step": 11377
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.0001412295918367347,
+      "loss": 0.0474,
+      "step": 11378
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.0001412142857142857,
+      "loss": 0.1299,
+      "step": 11379
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00014119897959183672,
+      "loss": 0.1221,
+      "step": 11380
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00014118367346938773,
+      "loss": 0.0747,
+      "step": 11381
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00014116836734693876,
+      "loss": 0.0824,
+      "step": 11382
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.0001411530612244898,
+      "loss": 0.0565,
+      "step": 11383
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00014113775510204082,
+      "loss": 0.0804,
+      "step": 11384
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00014112244897959182,
+      "loss": 0.1036,
+      "step": 11385
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00014110714285714285,
+      "loss": 0.1231,
+      "step": 11386
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014109183673469386,
+      "loss": 0.0754,
+      "step": 11387
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014107653061224489,
+      "loss": 0.1103,
+      "step": 11388
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.0001410612244897959,
+      "loss": 0.1033,
+      "step": 11389
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014104591836734692,
+      "loss": 0.1043,
+      "step": 11390
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014103061224489795,
+      "loss": 0.0936,
+      "step": 11391
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014101530612244895,
+      "loss": 0.1123,
+      "step": 11392
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 0.1356,
+      "step": 11393
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014098469387755102,
+      "loss": 0.1099,
+      "step": 11394
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014096938775510205,
+      "loss": 0.0579,
+      "step": 11395
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00014095408163265305,
+      "loss": 0.0809,
+      "step": 11396
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014093877551020408,
+      "loss": 0.0633,
+      "step": 11397
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014092346938775508,
+      "loss": 0.0624,
+      "step": 11398
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014090816326530611,
+      "loss": 0.2147,
+      "step": 11399
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014089285714285712,
+      "loss": 0.0953,
+      "step": 11400
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014087755102040815,
+      "loss": 0.0998,
+      "step": 11401
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014086224489795918,
+      "loss": 0.0438,
+      "step": 11402
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.0001408469387755102,
+      "loss": 0.0858,
+      "step": 11403
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.0001408316326530612,
+      "loss": 0.1227,
+      "step": 11404
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014081632653061224,
+      "loss": 0.1634,
+      "step": 11405
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014080102040816325,
+      "loss": 0.2348,
+      "step": 11406
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00014078571428571428,
+      "loss": 0.1367,
+      "step": 11407
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00014077040816326528,
+      "loss": 0.1106,
+      "step": 11408
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.0001407551020408163,
+      "loss": 0.1896,
+      "step": 11409
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00014073979591836734,
+      "loss": 0.0851,
+      "step": 11410
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00014072448979591837,
+      "loss": 0.0994,
+      "step": 11411
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00014070918367346938,
+      "loss": 0.1431,
+      "step": 11412
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.0001406938775510204,
+      "loss": 0.1015,
+      "step": 11413
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.0001406785714285714,
+      "loss": 0.0902,
+      "step": 11414
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00014066326530612244,
+      "loss": 0.1011,
+      "step": 11415
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00014064795918367344,
+      "loss": 0.044,
+      "step": 11416
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00014063265306122447,
+      "loss": 0.0461,
+      "step": 11417
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.0001406173469387755,
+      "loss": 0.2073,
+      "step": 11418
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00014060204081632654,
+      "loss": 0.1077,
+      "step": 11419
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00014058673469387754,
+      "loss": 0.0729,
+      "step": 11420
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00014057142857142857,
+      "loss": 0.1345,
+      "step": 11421
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00014055612244897957,
+      "loss": 0.0733,
+      "step": 11422
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.0001405408163265306,
+      "loss": 0.0545,
+      "step": 11423
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.0001405255102040816,
+      "loss": 0.2399,
+      "step": 11424
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00014051020408163264,
+      "loss": 0.095,
+      "step": 11425
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00014049489795918367,
+      "loss": 0.067,
+      "step": 11426
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00014047959183673467,
+      "loss": 0.125,
+      "step": 11427
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.0001404642857142857,
+      "loss": 0.2461,
+      "step": 11428
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00014044897959183673,
+      "loss": 0.1096,
+      "step": 11429
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00014043367346938774,
+      "loss": 0.2552,
+      "step": 11430
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00014041836734693877,
+      "loss": 0.1484,
+      "step": 11431
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00014040306122448977,
+      "loss": 0.0673,
+      "step": 11432
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.0001403877551020408,
+      "loss": 0.0639,
+      "step": 11433
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00014037244897959183,
+      "loss": 0.1352,
+      "step": 11434
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00014035714285714283,
+      "loss": 0.0413,
+      "step": 11435
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00014034183673469386,
+      "loss": 0.0655,
+      "step": 11436
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.0001403265306122449,
+      "loss": 0.1926,
+      "step": 11437
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.0001403112244897959,
+      "loss": 0.0985,
+      "step": 11438
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00014029591836734693,
+      "loss": 0.1751,
+      "step": 11439
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00014028061224489793,
+      "loss": 0.125,
+      "step": 11440
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00014026530612244896,
+      "loss": 0.0975,
+      "step": 11441
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00014025,
+      "loss": 0.0581,
+      "step": 11442
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.000140234693877551,
+      "loss": 0.1676,
+      "step": 11443
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00014021938775510203,
+      "loss": 0.1041,
+      "step": 11444
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00014020408163265306,
+      "loss": 0.1554,
+      "step": 11445
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.0001401887755102041,
+      "loss": 0.0502,
+      "step": 11446
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.0001401734693877551,
+      "loss": 0.0959,
+      "step": 11447
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00014015816326530612,
+      "loss": 0.1594,
+      "step": 11448
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00014014285714285713,
+      "loss": 0.1861,
+      "step": 11449
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00014012755102040816,
+      "loss": 0.0567,
+      "step": 11450
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00014011224489795916,
+      "loss": 0.0601,
+      "step": 11451
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.0001400969387755102,
+      "loss": 0.1342,
+      "step": 11452
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00014008163265306122,
+      "loss": 0.0502,
+      "step": 11453
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00014006632653061225,
+      "loss": 0.1678,
+      "step": 11454
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00014005102040816326,
+      "loss": 0.0954,
+      "step": 11455
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00014003571428571429,
+      "loss": 0.1429,
+      "step": 11456
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.0001400204081632653,
+      "loss": 0.1704,
+      "step": 11457
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00014000510204081632,
+      "loss": 0.1313,
+      "step": 11458
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013998979591836732,
+      "loss": 0.0514,
+      "step": 11459
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013997448979591835,
+      "loss": 0.091,
+      "step": 11460
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013995918367346938,
+      "loss": 0.0553,
+      "step": 11461
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.0001399438775510204,
+      "loss": 0.1897,
+      "step": 11462
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013992857142857142,
+      "loss": 0.058,
+      "step": 11463
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013991326530612245,
+      "loss": 0.0887,
+      "step": 11464
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013989795918367345,
+      "loss": 0.0531,
+      "step": 11465
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013988265306122448,
+      "loss": 0.0539,
+      "step": 11466
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013986734693877549,
+      "loss": 0.053,
+      "step": 11467
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013985204081632652,
+      "loss": 0.1248,
+      "step": 11468
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00013983673469387752,
+      "loss": 0.1152,
+      "step": 11469
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013982142857142855,
+      "loss": 0.0987,
+      "step": 11470
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013980612244897958,
+      "loss": 0.0433,
+      "step": 11471
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.0001397908163265306,
+      "loss": 0.0409,
+      "step": 11472
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013977551020408162,
+      "loss": 0.0425,
+      "step": 11473
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013976020408163265,
+      "loss": 0.1658,
+      "step": 11474
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013974489795918365,
+      "loss": 0.0613,
+      "step": 11475
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013972959183673468,
+      "loss": 0.0596,
+      "step": 11476
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013971428571428568,
+      "loss": 0.1125,
+      "step": 11477
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013969897959183671,
+      "loss": 0.0484,
+      "step": 11478
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013968367346938774,
+      "loss": 0.0582,
+      "step": 11479
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00013966836734693878,
+      "loss": 0.1288,
+      "step": 11480
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00013965306122448978,
+      "loss": 0.0785,
+      "step": 11481
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.0001396377551020408,
+      "loss": 0.0591,
+      "step": 11482
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.0001396224489795918,
+      "loss": 0.1118,
+      "step": 11483
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00013960714285714284,
+      "loss": 0.1217,
+      "step": 11484
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00013959183673469385,
+      "loss": 0.0345,
+      "step": 11485
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00013957653061224488,
+      "loss": 0.0799,
+      "step": 11486
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.0001395612244897959,
+      "loss": 0.156,
+      "step": 11487
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00013954591836734694,
+      "loss": 0.1007,
+      "step": 11488
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00013953061224489794,
+      "loss": 0.149,
+      "step": 11489
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00013951530612244897,
+      "loss": 0.0918,
+      "step": 11490
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.0001395,
+      "loss": 0.0493,
+      "step": 11491
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.000139484693877551,
+      "loss": 0.0529,
+      "step": 11492
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00013946938775510204,
+      "loss": 0.187,
+      "step": 11493
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00013945408163265304,
+      "loss": 0.1561,
+      "step": 11494
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00013943877551020407,
+      "loss": 0.0485,
+      "step": 11495
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.0001394234693877551,
+      "loss": 0.0631,
+      "step": 11496
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.0001394081632653061,
+      "loss": 0.1645,
+      "step": 11497
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00013939285714285714,
+      "loss": 0.0624,
+      "step": 11498
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00013937755102040817,
+      "loss": 0.157,
+      "step": 11499
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013936224489795917,
+      "loss": 0.1956,
+      "step": 11500
+    },
+    {
+      "epoch": 11.17,
+      "eval_cer": 0.24402014651217643,
+      "eval_loss": 0.954861044883728,
+      "eval_runtime": 238.1578,
+      "eval_samples_per_second": 37.009,
+      "eval_steps_per_second": 4.627,
+      "eval_wer": 0.6340451943886014,
+      "step": 11500
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.0001393469387755102,
+      "loss": 0.0937,
+      "step": 11501
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.0001393316326530612,
+      "loss": 0.0756,
+      "step": 11502
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013931632653061223,
+      "loss": 0.0993,
+      "step": 11503
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013930102040816324,
+      "loss": 0.0437,
+      "step": 11504
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013928571428571427,
+      "loss": 0.1267,
+      "step": 11505
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.0001392704081632653,
+      "loss": 0.1473,
+      "step": 11506
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013925510204081633,
+      "loss": 0.1441,
+      "step": 11507
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013923979591836733,
+      "loss": 0.0606,
+      "step": 11508
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013922448979591836,
+      "loss": 0.114,
+      "step": 11509
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013920918367346937,
+      "loss": 0.1066,
+      "step": 11510
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.0001391938775510204,
+      "loss": 0.08,
+      "step": 11511
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.0001391785714285714,
+      "loss": 0.1879,
+      "step": 11512
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00013916326530612243,
+      "loss": 0.116,
+      "step": 11513
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00013914795918367346,
+      "loss": 0.1038,
+      "step": 11514
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.0001391326530612245,
+      "loss": 0.0388,
+      "step": 11515
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.0001391173469387755,
+      "loss": 0.069,
+      "step": 11516
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00013910204081632653,
+      "loss": 0.0916,
+      "step": 11517
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00013908673469387753,
+      "loss": 0.1222,
+      "step": 11518
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00013907142857142856,
+      "loss": 0.1589,
+      "step": 11519
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00013905612244897956,
+      "loss": 0.0506,
+      "step": 11520
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.0001390408163265306,
+      "loss": 0.0496,
+      "step": 11521
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00013902551020408162,
+      "loss": 0.0389,
+      "step": 11522
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00013901020408163265,
+      "loss": 0.0527,
+      "step": 11523
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00013899489795918366,
+      "loss": 0.1732,
+      "step": 11524
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.0001389795918367347,
+      "loss": 0.0672,
+      "step": 11525
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.0001389642857142857,
+      "loss": 0.1152,
+      "step": 11526
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00013894897959183672,
+      "loss": 0.0532,
+      "step": 11527
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00013893367346938773,
+      "loss": 0.0538,
+      "step": 11528
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00013891836734693876,
+      "loss": 0.0555,
+      "step": 11529
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.0001389030612244898,
+      "loss": 0.1452,
+      "step": 11530
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.0001388877551020408,
+      "loss": 0.1921,
+      "step": 11531
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013887244897959182,
+      "loss": 0.0699,
+      "step": 11532
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013885714285714285,
+      "loss": 0.0541,
+      "step": 11533
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013884183673469388,
+      "loss": 0.0997,
+      "step": 11534
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013882653061224489,
+      "loss": 0.0797,
+      "step": 11535
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013881122448979592,
+      "loss": 0.0805,
+      "step": 11536
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013879591836734692,
+      "loss": 0.1931,
+      "step": 11537
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013878061224489795,
+      "loss": 0.0723,
+      "step": 11538
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013876530612244895,
+      "loss": 0.0902,
+      "step": 11539
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013874999999999998,
+      "loss": 0.0892,
+      "step": 11540
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00013873469387755102,
+      "loss": 0.0519,
+      "step": 11541
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00013871938775510205,
+      "loss": 0.1662,
+      "step": 11542
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00013870408163265305,
+      "loss": 0.0958,
+      "step": 11543
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00013868877551020408,
+      "loss": 0.1723,
+      "step": 11544
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00013867346938775508,
+      "loss": 0.0653,
+      "step": 11545
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.0001386581632653061,
+      "loss": 0.0577,
+      "step": 11546
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00013864285714285712,
+      "loss": 0.1767,
+      "step": 11547
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00013862755102040815,
+      "loss": 0.2491,
+      "step": 11548
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00013861224489795918,
+      "loss": 0.1126,
+      "step": 11549
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.0001385969387755102,
+      "loss": 0.0666,
+      "step": 11550
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.0001385816326530612,
+      "loss": 0.1416,
+      "step": 11551
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00013856632653061224,
+      "loss": 0.1034,
+      "step": 11552
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00013855102040816325,
+      "loss": 0.0509,
+      "step": 11553
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00013853571428571428,
+      "loss": 0.0374,
+      "step": 11554
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00013852040816326528,
+      "loss": 0.1712,
+      "step": 11555
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0001385051020408163,
+      "loss": 0.1372,
+      "step": 11556
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00013848979591836734,
+      "loss": 0.0779,
+      "step": 11557
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00013847448979591837,
+      "loss": 0.0736,
+      "step": 11558
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00013845918367346938,
+      "loss": 0.0549,
+      "step": 11559
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0001384438775510204,
+      "loss": 0.0604,
+      "step": 11560
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0001384285714285714,
+      "loss": 0.1626,
+      "step": 11561
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00013841326530612244,
+      "loss": 0.2098,
+      "step": 11562
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00013839795918367344,
+      "loss": 0.1139,
+      "step": 11563
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00013838265306122447,
+      "loss": 0.0857,
+      "step": 11564
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.0001383673469387755,
+      "loss": 0.0652,
+      "step": 11565
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.0001383520408163265,
+      "loss": 0.0527,
+      "step": 11566
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00013833673469387754,
+      "loss": 0.2227,
+      "step": 11567
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00013832142857142857,
+      "loss": 0.0966,
+      "step": 11568
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00013830612244897957,
+      "loss": 0.109,
+      "step": 11569
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.0001382908163265306,
+      "loss": 0.0734,
+      "step": 11570
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.0001382755102040816,
+      "loss": 0.1137,
+      "step": 11571
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00013826020408163264,
+      "loss": 0.1478,
+      "step": 11572
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00013824489795918364,
+      "loss": 0.0896,
+      "step": 11573
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00013822959183673467,
+      "loss": 0.2094,
+      "step": 11574
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.0001382142857142857,
+      "loss": 0.1808,
+      "step": 11575
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00013819897959183673,
+      "loss": 0.0672,
+      "step": 11576
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00013818367346938776,
+      "loss": 0.0974,
+      "step": 11577
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00013816836734693877,
+      "loss": 0.155,
+      "step": 11578
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.0001381530612244898,
+      "loss": 0.102,
+      "step": 11579
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.0001381377551020408,
+      "loss": 0.1103,
+      "step": 11580
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00013812244897959183,
+      "loss": 0.1566,
+      "step": 11581
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00013810714285714283,
+      "loss": 0.0586,
+      "step": 11582
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00013809183673469386,
+      "loss": 0.0672,
+      "step": 11583
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.0001380765306122449,
+      "loss": 0.0624,
+      "step": 11584
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00013806122448979593,
+      "loss": 0.0739,
+      "step": 11585
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00013804591836734693,
+      "loss": 0.0841,
+      "step": 11586
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00013803061224489796,
+      "loss": 0.0753,
+      "step": 11587
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00013801530612244896,
+      "loss": 0.0966,
+      "step": 11588
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.000138,
+      "loss": 0.1425,
+      "step": 11589
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.000137984693877551,
+      "loss": 0.0555,
+      "step": 11590
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00013796938775510203,
+      "loss": 0.0718,
+      "step": 11591
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00013795408163265306,
+      "loss": 0.0434,
+      "step": 11592
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013793877551020406,
+      "loss": 0.0596,
+      "step": 11593
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.0001379234693877551,
+      "loss": 0.1103,
+      "step": 11594
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013790816326530612,
+      "loss": 0.1349,
+      "step": 11595
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013789285714285713,
+      "loss": 0.1003,
+      "step": 11596
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013787755102040816,
+      "loss": 0.1176,
+      "step": 11597
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013786224489795916,
+      "loss": 0.0541,
+      "step": 11598
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.0001378469387755102,
+      "loss": 0.0731,
+      "step": 11599
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013783163265306122,
+      "loss": 0.149,
+      "step": 11600
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013781632653061222,
+      "loss": 0.0485,
+      "step": 11601
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013780102040816325,
+      "loss": 0.0659,
+      "step": 11602
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013778571428571429,
+      "loss": 0.0732,
+      "step": 11603
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.0001377704081632653,
+      "loss": 0.1251,
+      "step": 11604
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013775510204081632,
+      "loss": 0.1585,
+      "step": 11605
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013773979591836732,
+      "loss": 0.1214,
+      "step": 11606
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013772448979591835,
+      "loss": 0.2031,
+      "step": 11607
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013770918367346936,
+      "loss": 0.1184,
+      "step": 11608
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.0001376938775510204,
+      "loss": 0.0955,
+      "step": 11609
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013767857142857142,
+      "loss": 0.0961,
+      "step": 11610
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013766326530612245,
+      "loss": 0.0999,
+      "step": 11611
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013764795918367345,
+      "loss": 0.0912,
+      "step": 11612
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00013763265306122448,
+      "loss": 0.1159,
+      "step": 11613
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013761734693877549,
+      "loss": 0.0585,
+      "step": 11614
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013760204081632652,
+      "loss": 0.086,
+      "step": 11615
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013758673469387752,
+      "loss": 0.0608,
+      "step": 11616
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013757142857142855,
+      "loss": 0.1157,
+      "step": 11617
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013755612244897958,
+      "loss": 0.1736,
+      "step": 11618
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.0001375408163265306,
+      "loss": 0.0663,
+      "step": 11619
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013752551020408162,
+      "loss": 0.1321,
+      "step": 11620
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013751020408163265,
+      "loss": 0.0539,
+      "step": 11621
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013749489795918368,
+      "loss": 0.1203,
+      "step": 11622
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00013747959183673468,
+      "loss": 0.1713,
+      "step": 11623
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.0001374642857142857,
+      "loss": 0.1322,
+      "step": 11624
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.0001374489795918367,
+      "loss": 0.2028,
+      "step": 11625
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00013743367346938774,
+      "loss": 0.1125,
+      "step": 11626
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00013741836734693877,
+      "loss": 0.094,
+      "step": 11627
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00013740306122448978,
+      "loss": 0.0934,
+      "step": 11628
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.0001373877551020408,
+      "loss": 0.0417,
+      "step": 11629
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00013737244897959184,
+      "loss": 0.2524,
+      "step": 11630
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00013735714285714284,
+      "loss": 0.0776,
+      "step": 11631
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00013734183673469387,
+      "loss": 0.0859,
+      "step": 11632
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00013732653061224488,
+      "loss": 0.1076,
+      "step": 11633
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.0001373112244897959,
+      "loss": 0.1274,
+      "step": 11634
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013729591836734694,
+      "loss": 0.0488,
+      "step": 11635
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013728061224489794,
+      "loss": 0.1329,
+      "step": 11636
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013726530612244897,
+      "loss": 0.1599,
+      "step": 11637
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013725,
+      "loss": 0.0665,
+      "step": 11638
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.000137234693877551,
+      "loss": 0.1244,
+      "step": 11639
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013721938775510204,
+      "loss": 0.088,
+      "step": 11640
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013720408163265304,
+      "loss": 0.1461,
+      "step": 11641
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013718877551020407,
+      "loss": 0.1043,
+      "step": 11642
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013717346938775507,
+      "loss": 0.1327,
+      "step": 11643
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.0001371581632653061,
+      "loss": 0.0735,
+      "step": 11644
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00013714285714285713,
+      "loss": 0.093,
+      "step": 11645
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00013712755102040817,
+      "loss": 0.0605,
+      "step": 11646
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00013711224489795917,
+      "loss": 0.174,
+      "step": 11647
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0001370969387755102,
+      "loss": 0.1029,
+      "step": 11648
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0001370816326530612,
+      "loss": 0.1351,
+      "step": 11649
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00013706632653061223,
+      "loss": 0.1189,
+      "step": 11650
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00013705102040816324,
+      "loss": 0.1056,
+      "step": 11651
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00013703571428571427,
+      "loss": 0.0711,
+      "step": 11652
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0001370204081632653,
+      "loss": 0.0508,
+      "step": 11653
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00013700510204081633,
+      "loss": 0.0659,
+      "step": 11654
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00013698979591836733,
+      "loss": 0.1306,
+      "step": 11655
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00013697448979591836,
+      "loss": 0.1516,
+      "step": 11656
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00013695918367346937,
+      "loss": 0.1079,
+      "step": 11657
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.0001369438775510204,
+      "loss": 0.0598,
+      "step": 11658
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.0001369285714285714,
+      "loss": 0.1118,
+      "step": 11659
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00013691326530612243,
+      "loss": 0.0579,
+      "step": 11660
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00013689795918367346,
+      "loss": 0.0853,
+      "step": 11661
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.0001368826530612245,
+      "loss": 0.0776,
+      "step": 11662
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.0001368673469387755,
+      "loss": 0.1206,
+      "step": 11663
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00013685204081632653,
+      "loss": 0.0451,
+      "step": 11664
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013683673469387753,
+      "loss": 0.0879,
+      "step": 11665
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013682142857142856,
+      "loss": 0.0783,
+      "step": 11666
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013680612244897956,
+      "loss": 0.072,
+      "step": 11667
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.0001367908163265306,
+      "loss": 0.2365,
+      "step": 11668
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013677551020408162,
+      "loss": 0.2178,
+      "step": 11669
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013676020408163263,
+      "loss": 0.134,
+      "step": 11670
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013674489795918366,
+      "loss": 0.0486,
+      "step": 11671
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.0001367295918367347,
+      "loss": 0.1085,
+      "step": 11672
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013671428571428572,
+      "loss": 0.0551,
+      "step": 11673
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013669897959183672,
+      "loss": 0.2675,
+      "step": 11674
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013668367346938775,
+      "loss": 0.1112,
+      "step": 11675
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013666836734693876,
+      "loss": 0.139,
+      "step": 11676
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.0001366530612244898,
+      "loss": 0.0557,
+      "step": 11677
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.0001366377551020408,
+      "loss": 0.1056,
+      "step": 11678
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013662244897959182,
+      "loss": 0.0512,
+      "step": 11679
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013660714285714285,
+      "loss": 0.1053,
+      "step": 11680
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013659183673469388,
+      "loss": 0.1627,
+      "step": 11681
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013657653061224489,
+      "loss": 0.1065,
+      "step": 11682
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013656122448979592,
+      "loss": 0.0496,
+      "step": 11683
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013654591836734692,
+      "loss": 0.0495,
+      "step": 11684
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013653061224489795,
+      "loss": 0.1118,
+      "step": 11685
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013651530612244895,
+      "loss": 0.144,
+      "step": 11686
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013649999999999998,
+      "loss": 0.1382,
+      "step": 11687
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013648469387755101,
+      "loss": 0.1174,
+      "step": 11688
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013646938775510205,
+      "loss": 0.0768,
+      "step": 11689
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013645408163265305,
+      "loss": 0.1031,
+      "step": 11690
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013643877551020408,
+      "loss": 0.0851,
+      "step": 11691
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013642346938775508,
+      "loss": 0.0963,
+      "step": 11692
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.0001364081632653061,
+      "loss": 0.1302,
+      "step": 11693
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013639285714285712,
+      "loss": 0.0726,
+      "step": 11694
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00013637755102040815,
+      "loss": 0.0984,
+      "step": 11695
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00013636224489795918,
+      "loss": 0.1011,
+      "step": 11696
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.0001363469387755102,
+      "loss": 0.0482,
+      "step": 11697
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.0001363316326530612,
+      "loss": 0.1254,
+      "step": 11698
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00013631632653061224,
+      "loss": 0.1939,
+      "step": 11699
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00013630102040816325,
+      "loss": 0.1356,
+      "step": 11700
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00013628571428571428,
+      "loss": 0.1742,
+      "step": 11701
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00013627040816326528,
+      "loss": 0.1092,
+      "step": 11702
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.0001362551020408163,
+      "loss": 0.0924,
+      "step": 11703
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00013623979591836734,
+      "loss": 0.0747,
+      "step": 11704
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00013622448979591834,
+      "loss": 0.1485,
+      "step": 11705
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00013620918367346937,
+      "loss": 0.1849,
+      "step": 11706
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.0001361938775510204,
+      "loss": 0.1113,
+      "step": 11707
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.0001361785714285714,
+      "loss": 0.0853,
+      "step": 11708
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00013616326530612244,
+      "loss": 0.0503,
+      "step": 11709
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00013614795918367344,
+      "loss": 0.0505,
+      "step": 11710
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00013613265306122447,
+      "loss": 0.0875,
+      "step": 11711
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00013611734693877548,
+      "loss": 0.2585,
+      "step": 11712
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.0001361020408163265,
+      "loss": 0.1497,
+      "step": 11713
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00013608673469387754,
+      "loss": 0.1385,
+      "step": 11714
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00013607142857142857,
+      "loss": 0.0972,
+      "step": 11715
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.0001360561224489796,
+      "loss": 0.0922,
+      "step": 11716
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.0001360408163265306,
+      "loss": 0.1051,
+      "step": 11717
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00013602551020408163,
+      "loss": 0.2669,
+      "step": 11718
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00013601020408163264,
+      "loss": 0.2067,
+      "step": 11719
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00013599489795918367,
+      "loss": 0.1288,
+      "step": 11720
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00013597959183673467,
+      "loss": 0.0888,
+      "step": 11721
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.0001359642857142857,
+      "loss": 0.111,
+      "step": 11722
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00013594897959183673,
+      "loss": 0.0827,
+      "step": 11723
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00013593367346938776,
+      "loss": 0.0814,
+      "step": 11724
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00013591836734693877,
+      "loss": 0.1631,
+      "step": 11725
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.0001359030612244898,
+      "loss": 0.052,
+      "step": 11726
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.0001358877551020408,
+      "loss": 0.0956,
+      "step": 11727
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00013587244897959183,
+      "loss": 0.0992,
+      "step": 11728
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00013585714285714283,
+      "loss": 0.0768,
+      "step": 11729
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00013584183673469386,
+      "loss": 0.0809,
+      "step": 11730
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.0001358265306122449,
+      "loss": 0.1052,
+      "step": 11731
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.0001358112244897959,
+      "loss": 0.0624,
+      "step": 11732
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00013579591836734693,
+      "loss": 0.0622,
+      "step": 11733
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00013578061224489796,
+      "loss": 0.05,
+      "step": 11734
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00013576530612244896,
+      "loss": 0.0499,
+      "step": 11735
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00013575,
+      "loss": 0.1718,
+      "step": 11736
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.000135734693877551,
+      "loss": 0.3187,
+      "step": 11737
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00013571938775510203,
+      "loss": 0.1123,
+      "step": 11738
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00013570408163265306,
+      "loss": 0.0637,
+      "step": 11739
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00013568877551020406,
+      "loss": 0.092,
+      "step": 11740
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.0001356734693877551,
+      "loss": 0.0548,
+      "step": 11741
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00013565816326530612,
+      "loss": 0.1708,
+      "step": 11742
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00013564285714285713,
+      "loss": 0.279,
+      "step": 11743
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00013562755102040816,
+      "loss": 0.0597,
+      "step": 11744
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00013561224489795916,
+      "loss": 0.0593,
+      "step": 11745
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.0001355969387755102,
+      "loss": 0.1292,
+      "step": 11746
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.0001355816326530612,
+      "loss": 0.0547,
+      "step": 11747
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00013556632653061222,
+      "loss": 0.0874,
+      "step": 11748
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00013555102040816325,
+      "loss": 0.3388,
+      "step": 11749
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00013553571428571428,
+      "loss": 0.1079,
+      "step": 11750
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.0001355204081632653,
+      "loss": 0.045,
+      "step": 11751
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00013550510204081632,
+      "loss": 0.1061,
+      "step": 11752
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00013548979591836732,
+      "loss": 0.15,
+      "step": 11753
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00013547448979591835,
+      "loss": 0.0539,
+      "step": 11754
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00013545918367346936,
+      "loss": 0.0881,
+      "step": 11755
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.0001354438775510204,
+      "loss": 0.1143,
+      "step": 11756
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00013542857142857142,
+      "loss": 0.0669,
+      "step": 11757
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013541326530612245,
+      "loss": 0.0527,
+      "step": 11758
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013539795918367348,
+      "loss": 0.0481,
+      "step": 11759
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013538265306122448,
+      "loss": 0.1208,
+      "step": 11760
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.0001353673469387755,
+      "loss": 0.136,
+      "step": 11761
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013535204081632652,
+      "loss": 0.1483,
+      "step": 11762
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013533673469387755,
+      "loss": 0.184,
+      "step": 11763
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013532142857142855,
+      "loss": 0.057,
+      "step": 11764
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013530612244897958,
+      "loss": 0.052,
+      "step": 11765
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.0001352908163265306,
+      "loss": 0.1842,
+      "step": 11766
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013527551020408161,
+      "loss": 0.0818,
+      "step": 11767
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013526020408163265,
+      "loss": 0.0994,
+      "step": 11768
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013524489795918368,
+      "loss": 0.105,
+      "step": 11769
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013522959183673468,
+      "loss": 0.1175,
+      "step": 11770
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.0001352142857142857,
+      "loss": 0.1093,
+      "step": 11771
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.0001351989795918367,
+      "loss": 0.0989,
+      "step": 11772
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013518367346938774,
+      "loss": 0.164,
+      "step": 11773
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013516836734693877,
+      "loss": 0.0593,
+      "step": 11774
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013515306122448978,
+      "loss": 0.1534,
+      "step": 11775
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.0001351377551020408,
+      "loss": 0.1405,
+      "step": 11776
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013512244897959184,
+      "loss": 0.1409,
+      "step": 11777
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013510714285714284,
+      "loss": 0.05,
+      "step": 11778
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00013509183673469387,
+      "loss": 0.094,
+      "step": 11779
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00013507653061224488,
+      "loss": 0.1642,
+      "step": 11780
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.0001350612244897959,
+      "loss": 0.1622,
+      "step": 11781
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.0001350459183673469,
+      "loss": 0.1056,
+      "step": 11782
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00013503061224489794,
+      "loss": 0.0633,
+      "step": 11783
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00013501530612244897,
+      "loss": 0.0826,
+      "step": 11784
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.000135,
+      "loss": 0.1266,
+      "step": 11785
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.000134984693877551,
+      "loss": 0.0931,
+      "step": 11786
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00013496938775510204,
+      "loss": 0.1413,
+      "step": 11787
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00013495408163265304,
+      "loss": 0.058,
+      "step": 11788
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00013493877551020407,
+      "loss": 0.0751,
+      "step": 11789
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00013492346938775507,
+      "loss": 0.1341,
+      "step": 11790
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.0001349081632653061,
+      "loss": 0.0444,
+      "step": 11791
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00013489285714285713,
+      "loss": 0.0617,
+      "step": 11792
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00013487755102040816,
+      "loss": 0.2557,
+      "step": 11793
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00013486224489795917,
+      "loss": 0.1362,
+      "step": 11794
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.0001348469387755102,
+      "loss": 0.1708,
+      "step": 11795
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.0001348316326530612,
+      "loss": 0.0961,
+      "step": 11796
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00013481632653061223,
+      "loss": 0.0776,
+      "step": 11797
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00013480102040816324,
+      "loss": 0.0477,
+      "step": 11798
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00013478571428571427,
+      "loss": 0.1752,
+      "step": 11799
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.0001347704081632653,
+      "loss": 0.0785,
+      "step": 11800
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00013475510204081633,
+      "loss": 0.0479,
+      "step": 11801
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00013473979591836733,
+      "loss": 0.1151,
+      "step": 11802
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00013472448979591836,
+      "loss": 0.0874,
+      "step": 11803
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.0001347091836734694,
+      "loss": 0.0397,
+      "step": 11804
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.0001346938775510204,
+      "loss": 0.1519,
+      "step": 11805
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00013467857142857143,
+      "loss": 0.2292,
+      "step": 11806
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00013466326530612243,
+      "loss": 0.1055,
+      "step": 11807
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00013464795918367346,
+      "loss": 0.0697,
+      "step": 11808
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00013463265306122446,
+      "loss": 0.1289,
+      "step": 11809
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.0001346173469387755,
+      "loss": 0.0445,
+      "step": 11810
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00013460204081632652,
+      "loss": 0.0535,
+      "step": 11811
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00013458673469387756,
+      "loss": 0.0977,
+      "step": 11812
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00013457142857142856,
+      "loss": 0.0681,
+      "step": 11813
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.0001345561224489796,
+      "loss": 0.0958,
+      "step": 11814
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.0001345408163265306,
+      "loss": 0.0621,
+      "step": 11815
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00013452551020408162,
+      "loss": 0.1357,
+      "step": 11816
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00013451020408163263,
+      "loss": 0.1339,
+      "step": 11817
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00013449489795918366,
+      "loss": 0.2273,
+      "step": 11818
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.0001344795918367347,
+      "loss": 0.1097,
+      "step": 11819
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013446428571428572,
+      "loss": 0.1065,
+      "step": 11820
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013444897959183672,
+      "loss": 0.0607,
+      "step": 11821
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013443367346938775,
+      "loss": 0.0915,
+      "step": 11822
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013441836734693876,
+      "loss": 0.0893,
+      "step": 11823
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.0001344030612244898,
+      "loss": 0.1341,
+      "step": 11824
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.0001343877551020408,
+      "loss": 0.1041,
+      "step": 11825
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013437244897959182,
+      "loss": 0.0599,
+      "step": 11826
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013435714285714285,
+      "loss": 0.0818,
+      "step": 11827
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013434183673469388,
+      "loss": 0.1064,
+      "step": 11828
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013432653061224488,
+      "loss": 0.0431,
+      "step": 11829
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013431122448979592,
+      "loss": 0.1313,
+      "step": 11830
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013429591836734692,
+      "loss": 0.0747,
+      "step": 11831
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013428061224489795,
+      "loss": 0.0822,
+      "step": 11832
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013426530612244895,
+      "loss": 0.0479,
+      "step": 11833
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013424999999999998,
+      "loss": 0.0528,
+      "step": 11834
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013423469387755101,
+      "loss": 0.0824,
+      "step": 11835
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013421938775510204,
+      "loss": 0.0523,
+      "step": 11836
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013420408163265305,
+      "loss": 0.1378,
+      "step": 11837
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013418877551020408,
+      "loss": 0.1454,
+      "step": 11838
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013417346938775508,
+      "loss": 0.2209,
+      "step": 11839
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.0001341581632653061,
+      "loss": 0.1409,
+      "step": 11840
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00013414285714285712,
+      "loss": 0.1338,
+      "step": 11841
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00013412755102040815,
+      "loss": 0.0913,
+      "step": 11842
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00013411224489795918,
+      "loss": 0.1548,
+      "step": 11843
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00013409693877551018,
+      "loss": 0.102,
+      "step": 11844
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.0001340816326530612,
+      "loss": 0.1256,
+      "step": 11845
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00013406632653061224,
+      "loss": 0.0909,
+      "step": 11846
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00013405102040816325,
+      "loss": 0.0521,
+      "step": 11847
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00013403571428571428,
+      "loss": 0.0898,
+      "step": 11848
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.0001340204081632653,
+      "loss": 0.1193,
+      "step": 11849
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.0001340051020408163,
+      "loss": 0.137,
+      "step": 11850
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00013398979591836734,
+      "loss": 0.0982,
+      "step": 11851
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00013397448979591834,
+      "loss": 0.0906,
+      "step": 11852
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00013395918367346937,
+      "loss": 0.137,
+      "step": 11853
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.0001339438775510204,
+      "loss": 0.0467,
+      "step": 11854
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00013392857142857144,
+      "loss": 0.1321,
+      "step": 11855
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00013391326530612244,
+      "loss": 0.1845,
+      "step": 11856
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00013389795918367347,
+      "loss": 0.1362,
+      "step": 11857
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00013388265306122447,
+      "loss": 0.1052,
+      "step": 11858
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.0001338673469387755,
+      "loss": 0.0906,
+      "step": 11859
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.0001338520408163265,
+      "loss": 0.0809,
+      "step": 11860
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00013383673469387754,
+      "loss": 0.0919,
+      "step": 11861
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00013382142857142857,
+      "loss": 0.0901,
+      "step": 11862
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0001338061224489796,
+      "loss": 0.1083,
+      "step": 11863
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0001337908163265306,
+      "loss": 0.0614,
+      "step": 11864
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00013377551020408163,
+      "loss": 0.1014,
+      "step": 11865
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00013376020408163264,
+      "loss": 0.0418,
+      "step": 11866
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00013374489795918367,
+      "loss": 0.0784,
+      "step": 11867
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00013372959183673467,
+      "loss": 0.23,
+      "step": 11868
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0001337142857142857,
+      "loss": 0.1695,
+      "step": 11869
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00013369897959183673,
+      "loss": 0.0469,
+      "step": 11870
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00013368367346938773,
+      "loss": 0.293,
+      "step": 11871
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00013366836734693876,
+      "loss": 0.1367,
+      "step": 11872
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.0001336530612244898,
+      "loss": 0.0504,
+      "step": 11873
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.0001336377551020408,
+      "loss": 0.0844,
+      "step": 11874
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00013362244897959183,
+      "loss": 0.0805,
+      "step": 11875
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00013360714285714283,
+      "loss": 0.0936,
+      "step": 11876
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00013359183673469386,
+      "loss": 0.1047,
+      "step": 11877
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.0001335765306122449,
+      "loss": 0.1942,
+      "step": 11878
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.0001335612244897959,
+      "loss": 0.0639,
+      "step": 11879
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00013354591836734693,
+      "loss": 0.1657,
+      "step": 11880
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00013353061224489796,
+      "loss": 0.1111,
+      "step": 11881
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00013351530612244896,
+      "loss": 0.145,
+      "step": 11882
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.0001335,
+      "loss": 0.0683,
+      "step": 11883
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.000133484693877551,
+      "loss": 0.0789,
+      "step": 11884
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00013346938775510203,
+      "loss": 0.1136,
+      "step": 11885
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00013345408163265303,
+      "loss": 0.1004,
+      "step": 11886
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00013343877551020406,
+      "loss": 0.2408,
+      "step": 11887
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.0001334234693877551,
+      "loss": 0.1821,
+      "step": 11888
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00013340816326530612,
+      "loss": 0.1137,
+      "step": 11889
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00013339285714285712,
+      "loss": 0.2037,
+      "step": 11890
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00013337755102040816,
+      "loss": 0.1244,
+      "step": 11891
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013336224489795916,
+      "loss": 0.0849,
+      "step": 11892
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.0001333469387755102,
+      "loss": 0.1549,
+      "step": 11893
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013333163265306122,
+      "loss": 0.1409,
+      "step": 11894
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013331632653061222,
+      "loss": 0.0917,
+      "step": 11895
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013330102040816325,
+      "loss": 0.1017,
+      "step": 11896
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013328571428571428,
+      "loss": 0.0975,
+      "step": 11897
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013327040816326532,
+      "loss": 0.0976,
+      "step": 11898
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013325510204081632,
+      "loss": 0.253,
+      "step": 11899
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013323979591836735,
+      "loss": 0.0605,
+      "step": 11900
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00013322448979591835,
+      "loss": 0.1039,
+      "step": 11901
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00013320918367346938,
+      "loss": 0.1471,
+      "step": 11902
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.0001331938775510204,
+      "loss": 0.0788,
+      "step": 11903
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00013317857142857142,
+      "loss": 0.0906,
+      "step": 11904
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00013316326530612245,
+      "loss": 0.1129,
+      "step": 11905
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00013314795918367345,
+      "loss": 0.1115,
+      "step": 11906
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00013313265306122448,
+      "loss": 0.1483,
+      "step": 11907
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.0001331173469387755,
+      "loss": 0.0996,
+      "step": 11908
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00013310204081632652,
+      "loss": 0.1221,
+      "step": 11909
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00013308673469387755,
+      "loss": 0.0911,
+      "step": 11910
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00013307142857142855,
+      "loss": 0.0866,
+      "step": 11911
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00013305612244897958,
+      "loss": 0.0656,
+      "step": 11912
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.0001330408163265306,
+      "loss": 0.1052,
+      "step": 11913
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00013302551020408161,
+      "loss": 0.1378,
+      "step": 11914
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00013301020408163264,
+      "loss": 0.0832,
+      "step": 11915
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00013299489795918368,
+      "loss": 0.0791,
+      "step": 11916
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00013297959183673468,
+      "loss": 0.1237,
+      "step": 11917
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.0001329642857142857,
+      "loss": 0.2698,
+      "step": 11918
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.0001329489795918367,
+      "loss": 0.1404,
+      "step": 11919
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00013293367346938774,
+      "loss": 0.1406,
+      "step": 11920
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00013291836734693875,
+      "loss": 0.0463,
+      "step": 11921
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00013290306122448978,
+      "loss": 0.0936,
+      "step": 11922
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.0001328877551020408,
+      "loss": 0.0818,
+      "step": 11923
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00013287244897959184,
+      "loss": 0.1334,
+      "step": 11924
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00013285714285714284,
+      "loss": 0.1551,
+      "step": 11925
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00013284183673469387,
+      "loss": 0.1638,
+      "step": 11926
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00013282653061224488,
+      "loss": 0.1387,
+      "step": 11927
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.0001328112244897959,
+      "loss": 0.0428,
+      "step": 11928
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.0001327959183673469,
+      "loss": 0.1304,
+      "step": 11929
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00013278061224489794,
+      "loss": 0.0789,
+      "step": 11930
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00013276530612244897,
+      "loss": 0.1126,
+      "step": 11931
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00013275,
+      "loss": 0.0796,
+      "step": 11932
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.000132734693877551,
+      "loss": 0.0617,
+      "step": 11933
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00013271938775510204,
+      "loss": 0.1632,
+      "step": 11934
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00013270408163265304,
+      "loss": 0.0998,
+      "step": 11935
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00013268877551020407,
+      "loss": 0.0556,
+      "step": 11936
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00013267346938775507,
+      "loss": 0.131,
+      "step": 11937
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.0001326581632653061,
+      "loss": 0.0666,
+      "step": 11938
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00013264285714285713,
+      "loss": 0.1331,
+      "step": 11939
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00013262755102040816,
+      "loss": 0.136,
+      "step": 11940
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00013261224489795917,
+      "loss": 0.0701,
+      "step": 11941
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.0001325969387755102,
+      "loss": 0.0365,
+      "step": 11942
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.00013258163265306123,
+      "loss": 0.168,
+      "step": 11943
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.00013256632653061223,
+      "loss": 0.1458,
+      "step": 11944
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.00013255102040816326,
+      "loss": 0.1432,
+      "step": 11945
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.00013253571428571427,
+      "loss": 0.0516,
+      "step": 11946
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.0001325204081632653,
+      "loss": 0.122,
+      "step": 11947
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.0001325051020408163,
+      "loss": 0.129,
+      "step": 11948
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.00013248979591836733,
+      "loss": 0.2008,
+      "step": 11949
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.00013247448979591836,
+      "loss": 0.071,
+      "step": 11950
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.0001324591836734694,
+      "loss": 0.0941,
+      "step": 11951
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.0001324438775510204,
+      "loss": 0.0631,
+      "step": 11952
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.00013242857142857143,
+      "loss": 0.0833,
+      "step": 11953
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00013241326530612243,
+      "loss": 0.1282,
+      "step": 11954
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00013239795918367346,
+      "loss": 0.0981,
+      "step": 11955
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00013238265306122446,
+      "loss": 0.2937,
+      "step": 11956
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0001323673469387755,
+      "loss": 0.0962,
+      "step": 11957
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00013235204081632652,
+      "loss": 0.1422,
+      "step": 11958
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00013233673469387755,
+      "loss": 0.1346,
+      "step": 11959
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00013232142857142856,
+      "loss": 0.0546,
+      "step": 11960
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0001323061224489796,
+      "loss": 0.2594,
+      "step": 11961
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0001322908163265306,
+      "loss": 0.1816,
+      "step": 11962
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00013227551020408162,
+      "loss": 0.1369,
+      "step": 11963
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00013226020408163263,
+      "loss": 0.1036,
+      "step": 11964
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00013224489795918366,
+      "loss": 0.0605,
+      "step": 11965
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.0001322295918367347,
+      "loss": 0.0792,
+      "step": 11966
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00013221428571428572,
+      "loss": 0.0532,
+      "step": 11967
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00013219897959183672,
+      "loss": 0.1885,
+      "step": 11968
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00013218367346938775,
+      "loss": 0.1609,
+      "step": 11969
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00013216836734693876,
+      "loss": 0.0584,
+      "step": 11970
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00013215306122448979,
+      "loss": 0.103,
+      "step": 11971
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.0001321377551020408,
+      "loss": 0.0858,
+      "step": 11972
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00013212244897959182,
+      "loss": 0.0513,
+      "step": 11973
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013210714285714285,
+      "loss": 0.1729,
+      "step": 11974
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013209183673469388,
+      "loss": 0.2398,
+      "step": 11975
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013207653061224488,
+      "loss": 0.1076,
+      "step": 11976
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013206122448979592,
+      "loss": 0.0591,
+      "step": 11977
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013204591836734692,
+      "loss": 0.1028,
+      "step": 11978
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013203061224489795,
+      "loss": 0.0868,
+      "step": 11979
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013201530612244895,
+      "loss": 0.1018,
+      "step": 11980
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013199999999999998,
+      "loss": 0.1218,
+      "step": 11981
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013198469387755101,
+      "loss": 0.1216,
+      "step": 11982
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013196938775510202,
+      "loss": 0.088,
+      "step": 11983
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00013195408163265305,
+      "loss": 0.1012,
+      "step": 11984
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.00013193877551020408,
+      "loss": 0.0617,
+      "step": 11985
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.0001319234693877551,
+      "loss": 0.1783,
+      "step": 11986
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.0001319081632653061,
+      "loss": 0.1845,
+      "step": 11987
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.00013189285714285714,
+      "loss": 0.1048,
+      "step": 11988
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.00013187755102040815,
+      "loss": 0.0503,
+      "step": 11989
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.00013186224489795918,
+      "loss": 0.0776,
+      "step": 11990
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.00013184693877551018,
+      "loss": 0.0554,
+      "step": 11991
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.0001318316326530612,
+      "loss": 0.0925,
+      "step": 11992
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.00013181632653061224,
+      "loss": 0.1562,
+      "step": 11993
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.00013180102040816327,
+      "loss": 0.1646,
+      "step": 11994
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00013178571428571428,
+      "loss": 0.1014,
+      "step": 11995
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.0001317704081632653,
+      "loss": 0.1455,
+      "step": 11996
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.0001317551020408163,
+      "loss": 0.0627,
+      "step": 11997
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00013173979591836734,
+      "loss": 0.0844,
+      "step": 11998
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00013172448979591834,
+      "loss": 0.1372,
+      "step": 11999
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00013170918367346937,
+      "loss": 0.0828,
+      "step": 12000
+    },
+    {
+      "epoch": 11.65,
+      "eval_cer": 0.2460375828014862,
+      "eval_loss": 0.9597804546356201,
+      "eval_runtime": 236.5269,
+      "eval_samples_per_second": 37.264,
+      "eval_steps_per_second": 4.659,
+      "eval_wer": 0.6402904568873655,
+      "step": 12000
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.0001316938775510204,
+      "loss": 0.1071,
+      "step": 12001
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00013167857142857143,
+      "loss": 0.1786,
+      "step": 12002
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00013166326530612244,
+      "loss": 0.1626,
+      "step": 12003
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00013164795918367347,
+      "loss": 0.0571,
+      "step": 12004
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00013163265306122447,
+      "loss": 0.169,
+      "step": 12005
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.0001316173469387755,
+      "loss": 0.12,
+      "step": 12006
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.0001316020408163265,
+      "loss": 0.0612,
+      "step": 12007
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00013158673469387754,
+      "loss": 0.0523,
+      "step": 12008
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00013157142857142857,
+      "loss": 0.0542,
+      "step": 12009
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00013155612244897957,
+      "loss": 0.0476,
+      "step": 12010
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.0001315408163265306,
+      "loss": 0.0925,
+      "step": 12011
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00013152551020408163,
+      "loss": 0.106,
+      "step": 12012
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00013151020408163264,
+      "loss": 0.0708,
+      "step": 12013
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00013149489795918367,
+      "loss": 0.0586,
+      "step": 12014
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00013147959183673467,
+      "loss": 0.0958,
+      "step": 12015
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.0001314642857142857,
+      "loss": 0.0556,
+      "step": 12016
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00013144897959183673,
+      "loss": 0.0447,
+      "step": 12017
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00013143367346938773,
+      "loss": 0.1908,
+      "step": 12018
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00013141836734693876,
+      "loss": 0.0525,
+      "step": 12019
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.0001314030612244898,
+      "loss": 0.0845,
+      "step": 12020
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.0001313877551020408,
+      "loss": 0.044,
+      "step": 12021
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00013137244897959183,
+      "loss": 0.0931,
+      "step": 12022
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00013135714285714283,
+      "loss": 0.1232,
+      "step": 12023
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00013134183673469386,
+      "loss": 0.2136,
+      "step": 12024
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00013132653061224487,
+      "loss": 0.1587,
+      "step": 12025
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.0001313112244897959,
+      "loss": 0.0557,
+      "step": 12026
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00013129591836734693,
+      "loss": 0.0978,
+      "step": 12027
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00013128061224489796,
+      "loss": 0.2095,
+      "step": 12028
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00013126530612244896,
+      "loss": 0.0642,
+      "step": 12029
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00013125,
+      "loss": 0.1482,
+      "step": 12030
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00013123469387755102,
+      "loss": 0.2374,
+      "step": 12031
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00013121938775510203,
+      "loss": 0.1507,
+      "step": 12032
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00013120408163265306,
+      "loss": 0.0787,
+      "step": 12033
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00013118877551020406,
+      "loss": 0.0628,
+      "step": 12034
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.0001311734693877551,
+      "loss": 0.0866,
+      "step": 12035
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00013115816326530612,
+      "loss": 0.0699,
+      "step": 12036
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00013114285714285715,
+      "loss": 0.1395,
+      "step": 12037
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00013112755102040815,
+      "loss": 0.1467,
+      "step": 12038
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00013111224489795919,
+      "loss": 0.0997,
+      "step": 12039
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.0001310969387755102,
+      "loss": 0.0441,
+      "step": 12040
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00013108163265306122,
+      "loss": 0.0874,
+      "step": 12041
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00013106632653061222,
+      "loss": 0.1801,
+      "step": 12042
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00013105102040816325,
+      "loss": 0.1584,
+      "step": 12043
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00013103571428571428,
+      "loss": 0.148,
+      "step": 12044
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.0001310204081632653,
+      "loss": 0.1007,
+      "step": 12045
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013100510204081632,
+      "loss": 0.0948,
+      "step": 12046
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013098979591836735,
+      "loss": 0.1275,
+      "step": 12047
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013097448979591835,
+      "loss": 0.098,
+      "step": 12048
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013095918367346938,
+      "loss": 0.1737,
+      "step": 12049
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013094387755102039,
+      "loss": 0.0707,
+      "step": 12050
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013092857142857142,
+      "loss": 0.118,
+      "step": 12051
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013091326530612242,
+      "loss": 0.086,
+      "step": 12052
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013089795918367345,
+      "loss": 0.058,
+      "step": 12053
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013088265306122448,
+      "loss": 0.1052,
+      "step": 12054
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.0001308673469387755,
+      "loss": 0.1165,
+      "step": 12055
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00013085204081632652,
+      "loss": 0.1114,
+      "step": 12056
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00013083673469387755,
+      "loss": 0.1761,
+      "step": 12057
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00013082142857142855,
+      "loss": 0.1376,
+      "step": 12058
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00013080612244897958,
+      "loss": 0.0783,
+      "step": 12059
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00013079081632653058,
+      "loss": 0.0965,
+      "step": 12060
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00013077551020408161,
+      "loss": 0.1327,
+      "step": 12061
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00013076020408163264,
+      "loss": 0.1846,
+      "step": 12062
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00013074489795918367,
+      "loss": 0.0603,
+      "step": 12063
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00013072959183673468,
+      "loss": 0.1053,
+      "step": 12064
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.0001307142857142857,
+      "loss": 0.0835,
+      "step": 12065
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.0001306989795918367,
+      "loss": 0.0604,
+      "step": 12066
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00013068367346938774,
+      "loss": 0.0931,
+      "step": 12067
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00013066836734693875,
+      "loss": 0.0949,
+      "step": 12068
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00013065306122448978,
+      "loss": 0.0909,
+      "step": 12069
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.0001306377551020408,
+      "loss": 0.0941,
+      "step": 12070
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00013062244897959184,
+      "loss": 0.1021,
+      "step": 12071
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00013060714285714284,
+      "loss": 0.0855,
+      "step": 12072
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00013059183673469387,
+      "loss": 0.0762,
+      "step": 12073
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00013057653061224488,
+      "loss": 0.1068,
+      "step": 12074
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.0001305612244897959,
+      "loss": 0.1424,
+      "step": 12075
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00013054591836734694,
+      "loss": 0.1022,
+      "step": 12076
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00013053061224489794,
+      "loss": 0.0923,
+      "step": 12077
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00013051530612244897,
+      "loss": 0.0654,
+      "step": 12078
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.0001305,
+      "loss": 0.0557,
+      "step": 12079
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.000130484693877551,
+      "loss": 0.1271,
+      "step": 12080
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00013046938775510203,
+      "loss": 0.2786,
+      "step": 12081
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00013045408163265307,
+      "loss": 0.1038,
+      "step": 12082
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00013043877551020407,
+      "loss": 0.1141,
+      "step": 12083
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.0001304234693877551,
+      "loss": 0.1001,
+      "step": 12084
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.0001304081632653061,
+      "loss": 0.1241,
+      "step": 12085
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00013039285714285713,
+      "loss": 0.0718,
+      "step": 12086
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00013037755102040814,
+      "loss": 0.2367,
+      "step": 12087
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00013036224489795917,
+      "loss": 0.1085,
+      "step": 12088
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.0001303469387755102,
+      "loss": 0.0579,
+      "step": 12089
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00013033163265306123,
+      "loss": 0.1392,
+      "step": 12090
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00013031632653061223,
+      "loss": 0.0473,
+      "step": 12091
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00013030102040816326,
+      "loss": 0.0447,
+      "step": 12092
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00013028571428571427,
+      "loss": 0.2576,
+      "step": 12093
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.0001302704081632653,
+      "loss": 0.1335,
+      "step": 12094
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.0001302551020408163,
+      "loss": 0.1221,
+      "step": 12095
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00013023979591836733,
+      "loss": 0.1008,
+      "step": 12096
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00013022448979591836,
+      "loss": 0.0699,
+      "step": 12097
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.0001302091836734694,
+      "loss": 0.0399,
+      "step": 12098
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.0001301938775510204,
+      "loss": 0.136,
+      "step": 12099
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00013017857142857143,
+      "loss": 0.0743,
+      "step": 12100
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00013016326530612243,
+      "loss": 0.1925,
+      "step": 12101
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00013014795918367346,
+      "loss": 0.0616,
+      "step": 12102
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00013013265306122446,
+      "loss": 0.1484,
+      "step": 12103
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.0001301173469387755,
+      "loss": 0.0415,
+      "step": 12104
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00013010204081632652,
+      "loss": 0.1392,
+      "step": 12105
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00013008673469387755,
+      "loss": 0.1549,
+      "step": 12106
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00013007142857142856,
+      "loss": 0.0481,
+      "step": 12107
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.0001300561224489796,
+      "loss": 0.065,
+      "step": 12108
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.0001300408163265306,
+      "loss": 0.1136,
+      "step": 12109
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00013002551020408162,
+      "loss": 0.0877,
+      "step": 12110
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00013001020408163263,
+      "loss": 0.1494,
+      "step": 12111
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00012999489795918366,
+      "loss": 0.1658,
+      "step": 12112
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.0001299795918367347,
+      "loss": 0.1063,
+      "step": 12113
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00012996428571428572,
+      "loss": 0.0727,
+      "step": 12114
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00012994897959183672,
+      "loss": 0.034,
+      "step": 12115
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00012993367346938775,
+      "loss": 0.1495,
+      "step": 12116
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00012991836734693875,
+      "loss": 0.2301,
+      "step": 12117
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012990306122448979,
+      "loss": 0.1645,
+      "step": 12118
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.0001298877551020408,
+      "loss": 0.0651,
+      "step": 12119
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012987244897959182,
+      "loss": 0.1777,
+      "step": 12120
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012985714285714285,
+      "loss": 0.087,
+      "step": 12121
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012984183673469385,
+      "loss": 0.0491,
+      "step": 12122
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012982653061224488,
+      "loss": 0.0861,
+      "step": 12123
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012981122448979591,
+      "loss": 0.1816,
+      "step": 12124
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012979591836734695,
+      "loss": 0.0943,
+      "step": 12125
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012978061224489795,
+      "loss": 0.18,
+      "step": 12126
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012976530612244898,
+      "loss": 0.0939,
+      "step": 12127
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012974999999999998,
+      "loss": 0.1252,
+      "step": 12128
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.000129734693877551,
+      "loss": 0.0498,
+      "step": 12129
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00012971938775510202,
+      "loss": 0.0579,
+      "step": 12130
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00012970408163265305,
+      "loss": 0.1494,
+      "step": 12131
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00012968877551020408,
+      "loss": 0.1269,
+      "step": 12132
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.0001296734693877551,
+      "loss": 0.0442,
+      "step": 12133
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.0001296581632653061,
+      "loss": 0.0471,
+      "step": 12134
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00012964285714285714,
+      "loss": 0.0829,
+      "step": 12135
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00012962755102040815,
+      "loss": 0.069,
+      "step": 12136
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00012961224489795918,
+      "loss": 0.3224,
+      "step": 12137
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00012959693877551018,
+      "loss": 0.1423,
+      "step": 12138
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.0001295816326530612,
+      "loss": 0.0646,
+      "step": 12139
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.00012956632653061224,
+      "loss": 0.1888,
+      "step": 12140
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.00012955102040816327,
+      "loss": 0.0965,
+      "step": 12141
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.00012953571428571427,
+      "loss": 0.1305,
+      "step": 12142
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.0001295204081632653,
+      "loss": 0.1988,
+      "step": 12143
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.0001295051020408163,
+      "loss": 0.1042,
+      "step": 12144
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.00012948979591836734,
+      "loss": 0.0625,
+      "step": 12145
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.00012947448979591834,
+      "loss": 0.1102,
+      "step": 12146
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.00012945918367346937,
+      "loss": 0.127,
+      "step": 12147
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.0001294438775510204,
+      "loss": 0.0494,
+      "step": 12148
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.0001294285714285714,
+      "loss": 0.1,
+      "step": 12149
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00012941326530612244,
+      "loss": 0.0938,
+      "step": 12150
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00012939795918367347,
+      "loss": 0.054,
+      "step": 12151
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00012938265306122447,
+      "loss": 0.1272,
+      "step": 12152
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.0001293673469387755,
+      "loss": 0.0426,
+      "step": 12153
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.0001293520408163265,
+      "loss": 0.1244,
+      "step": 12154
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00012933673469387754,
+      "loss": 0.1838,
+      "step": 12155
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00012932142857142857,
+      "loss": 0.1476,
+      "step": 12156
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00012930612244897957,
+      "loss": 0.2041,
+      "step": 12157
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.0001292908163265306,
+      "loss": 0.1016,
+      "step": 12158
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00012927551020408163,
+      "loss": 0.1371,
+      "step": 12159
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00012926020408163263,
+      "loss": 0.047,
+      "step": 12160
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00012924489795918367,
+      "loss": 0.1583,
+      "step": 12161
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00012922959183673467,
+      "loss": 0.1077,
+      "step": 12162
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.0001292142857142857,
+      "loss": 0.1013,
+      "step": 12163
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.0001291989795918367,
+      "loss": 0.053,
+      "step": 12164
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00012918367346938773,
+      "loss": 0.0927,
+      "step": 12165
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00012916836734693876,
+      "loss": 0.0949,
+      "step": 12166
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.0001291530612244898,
+      "loss": 0.0766,
+      "step": 12167
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00012913775510204082,
+      "loss": 0.1049,
+      "step": 12168
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00012912244897959183,
+      "loss": 0.0524,
+      "step": 12169
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00012910714285714286,
+      "loss": 0.0924,
+      "step": 12170
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00012909183673469386,
+      "loss": 0.0805,
+      "step": 12171
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.0001290765306122449,
+      "loss": 0.1299,
+      "step": 12172
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.0001290612244897959,
+      "loss": 0.0816,
+      "step": 12173
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00012904591836734693,
+      "loss": 0.2119,
+      "step": 12174
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00012903061224489796,
+      "loss": 0.1838,
+      "step": 12175
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.000129015306122449,
+      "loss": 0.0542,
+      "step": 12176
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.000129,
+      "loss": 0.0997,
+      "step": 12177
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00012898469387755102,
+      "loss": 0.1433,
+      "step": 12178
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00012896938775510203,
+      "loss": 0.0943,
+      "step": 12179
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012895408163265306,
+      "loss": 0.1369,
+      "step": 12180
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012893877551020406,
+      "loss": 0.2142,
+      "step": 12181
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.0001289234693877551,
+      "loss": 0.1018,
+      "step": 12182
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012890816326530612,
+      "loss": 0.0662,
+      "step": 12183
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012889285714285712,
+      "loss": 0.0421,
+      "step": 12184
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012887755102040815,
+      "loss": 0.0856,
+      "step": 12185
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012886224489795919,
+      "loss": 0.0806,
+      "step": 12186
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.0001288469387755102,
+      "loss": 0.063,
+      "step": 12187
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012883163265306122,
+      "loss": 0.0705,
+      "step": 12188
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012881632653061222,
+      "loss": 0.143,
+      "step": 12189
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00012880102040816325,
+      "loss": 0.1585,
+      "step": 12190
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012878571428571426,
+      "loss": 0.0639,
+      "step": 12191
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.0001287704081632653,
+      "loss": 0.078,
+      "step": 12192
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012875510204081632,
+      "loss": 0.1193,
+      "step": 12193
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012873979591836735,
+      "loss": 0.103,
+      "step": 12194
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012872448979591835,
+      "loss": 0.0941,
+      "step": 12195
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012870918367346938,
+      "loss": 0.1257,
+      "step": 12196
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012869387755102039,
+      "loss": 0.0444,
+      "step": 12197
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012867857142857142,
+      "loss": 0.0503,
+      "step": 12198
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012866326530612242,
+      "loss": 0.1386,
+      "step": 12199
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012866326530612242,
+      "loss": 0.1567,
+      "step": 12200
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00012864795918367345,
+      "loss": 0.0867,
+      "step": 12201
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00012863265306122448,
+      "loss": 0.1,
+      "step": 12202
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.0001286173469387755,
+      "loss": 0.0717,
+      "step": 12203
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00012860204081632651,
+      "loss": 0.0403,
+      "step": 12204
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00012858673469387755,
+      "loss": 0.147,
+      "step": 12205
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00012857142857142855,
+      "loss": 0.1247,
+      "step": 12206
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00012855612244897958,
+      "loss": 0.2578,
+      "step": 12207
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00012854081632653058,
+      "loss": 0.1152,
+      "step": 12208
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.0001285255102040816,
+      "loss": 0.0885,
+      "step": 12209
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00012851020408163264,
+      "loss": 0.0843,
+      "step": 12210
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012849489795918367,
+      "loss": 0.1028,
+      "step": 12211
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012847959183673468,
+      "loss": 0.1899,
+      "step": 12212
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.0001284642857142857,
+      "loss": 0.0492,
+      "step": 12213
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012844897959183674,
+      "loss": 0.1792,
+      "step": 12214
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012843367346938774,
+      "loss": 0.0604,
+      "step": 12215
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012841836734693877,
+      "loss": 0.0892,
+      "step": 12216
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012840306122448978,
+      "loss": 0.0797,
+      "step": 12217
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.0001283877551020408,
+      "loss": 0.1135,
+      "step": 12218
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012837244897959184,
+      "loss": 0.0844,
+      "step": 12219
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012835714285714284,
+      "loss": 0.0607,
+      "step": 12220
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012834183673469387,
+      "loss": 0.1264,
+      "step": 12221
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.0001283265306122449,
+      "loss": 0.0821,
+      "step": 12222
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.0001283112244897959,
+      "loss": 0.0639,
+      "step": 12223
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012829591836734694,
+      "loss": 0.106,
+      "step": 12224
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012828061224489794,
+      "loss": 0.0755,
+      "step": 12225
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012826530612244897,
+      "loss": 0.0936,
+      "step": 12226
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012824999999999997,
+      "loss": 0.1357,
+      "step": 12227
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.000128234693877551,
+      "loss": 0.0518,
+      "step": 12228
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012821938775510203,
+      "loss": 0.0723,
+      "step": 12229
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012820408163265306,
+      "loss": 0.2995,
+      "step": 12230
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012818877551020407,
+      "loss": 0.1142,
+      "step": 12231
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.0001281734693877551,
+      "loss": 0.1693,
+      "step": 12232
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.0001281581632653061,
+      "loss": 0.0565,
+      "step": 12233
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00012814285714285713,
+      "loss": 0.0589,
+      "step": 12234
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00012812755102040814,
+      "loss": 0.1663,
+      "step": 12235
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00012811224489795917,
+      "loss": 0.1174,
+      "step": 12236
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.0001280969387755102,
+      "loss": 0.0681,
+      "step": 12237
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00012808163265306123,
+      "loss": 0.063,
+      "step": 12238
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00012806632653061223,
+      "loss": 0.0478,
+      "step": 12239
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00012805102040816326,
+      "loss": 0.0782,
+      "step": 12240
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00012803571428571427,
+      "loss": 0.0568,
+      "step": 12241
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0001280204081632653,
+      "loss": 0.0926,
+      "step": 12242
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0001280051020408163,
+      "loss": 0.1502,
+      "step": 12243
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.00012798979591836733,
+      "loss": 0.0701,
+      "step": 12244
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.00012797448979591836,
+      "loss": 0.1029,
+      "step": 12245
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0001279591836734694,
+      "loss": 0.1367,
+      "step": 12246
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0001279438775510204,
+      "loss": 0.0508,
+      "step": 12247
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.00012792857142857142,
+      "loss": 0.0483,
+      "step": 12248
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.00012791326530612243,
+      "loss": 0.3224,
+      "step": 12249
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.00012789795918367346,
+      "loss": 0.1159,
+      "step": 12250
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.00012788265306122446,
+      "loss": 0.0919,
+      "step": 12251
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0001278673469387755,
+      "loss": 0.1546,
+      "step": 12252
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.00012785204081632652,
+      "loss": 0.069,
+      "step": 12253
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.00012783673469387755,
+      "loss": 0.0651,
+      "step": 12254
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.00012782142857142856,
+      "loss": 0.1001,
+      "step": 12255
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0001278061224489796,
+      "loss": 0.1406,
+      "step": 12256
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0001277908163265306,
+      "loss": 0.0582,
+      "step": 12257
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.00012777551020408162,
+      "loss": 0.0732,
+      "step": 12258
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.00012776020408163265,
+      "loss": 0.0951,
+      "step": 12259
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.00012774489795918366,
+      "loss": 0.0837,
+      "step": 12260
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0001277295918367347,
+      "loss": 0.0896,
+      "step": 12261
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0001277142857142857,
+      "loss": 0.1156,
+      "step": 12262
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012769897959183672,
+      "loss": 0.154,
+      "step": 12263
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012768367346938775,
+      "loss": 0.108,
+      "step": 12264
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012766836734693878,
+      "loss": 0.1202,
+      "step": 12265
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012765306122448979,
+      "loss": 0.0712,
+      "step": 12266
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012763775510204082,
+      "loss": 0.0742,
+      "step": 12267
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012762244897959182,
+      "loss": 0.0589,
+      "step": 12268
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012760714285714285,
+      "loss": 0.0735,
+      "step": 12269
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012759183673469385,
+      "loss": 0.1872,
+      "step": 12270
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012757653061224488,
+      "loss": 0.1265,
+      "step": 12271
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012756122448979591,
+      "loss": 0.0661,
+      "step": 12272
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00012754591836734694,
+      "loss": 0.1365,
+      "step": 12273
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00012753061224489795,
+      "loss": 0.1236,
+      "step": 12274
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00012751530612244898,
+      "loss": 0.1017,
+      "step": 12275
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00012749999999999998,
+      "loss": 0.0692,
+      "step": 12276
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.000127484693877551,
+      "loss": 0.1367,
+      "step": 12277
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00012746938775510202,
+      "loss": 0.0945,
+      "step": 12278
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00012745408163265305,
+      "loss": 0.0797,
+      "step": 12279
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00012743877551020408,
+      "loss": 0.1491,
+      "step": 12280
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.0001274234693877551,
+      "loss": 0.0724,
+      "step": 12281
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.0001274081632653061,
+      "loss": 0.1875,
+      "step": 12282
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00012739285714285714,
+      "loss": 0.0967,
+      "step": 12283
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00012737755102040815,
+      "loss": 0.092,
+      "step": 12284
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00012736224489795918,
+      "loss": 0.0491,
+      "step": 12285
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00012734693877551018,
+      "loss": 0.1467,
+      "step": 12286
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.0001273316326530612,
+      "loss": 0.1751,
+      "step": 12287
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00012731632653061224,
+      "loss": 0.1593,
+      "step": 12288
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00012730102040816324,
+      "loss": 0.0816,
+      "step": 12289
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00012728571428571427,
+      "loss": 0.1219,
+      "step": 12290
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.0001272704081632653,
+      "loss": 0.0508,
+      "step": 12291
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.0001272551020408163,
+      "loss": 0.0555,
+      "step": 12292
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00012723979591836734,
+      "loss": 0.0842,
+      "step": 12293
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00012722448979591834,
+      "loss": 0.0959,
+      "step": 12294
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00012720918367346937,
+      "loss": 0.1105,
+      "step": 12295
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0001271938775510204,
+      "loss": 0.0658,
+      "step": 12296
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0001271785714285714,
+      "loss": 0.1058,
+      "step": 12297
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00012716326530612244,
+      "loss": 0.086,
+      "step": 12298
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00012714795918367347,
+      "loss": 0.1134,
+      "step": 12299
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00012713265306122447,
+      "loss": 0.1059,
+      "step": 12300
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0001271173469387755,
+      "loss": 0.1086,
+      "step": 12301
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0001271020408163265,
+      "loss": 0.0528,
+      "step": 12302
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00012708673469387754,
+      "loss": 0.0817,
+      "step": 12303
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00012707142857142857,
+      "loss": 0.1012,
+      "step": 12304
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00012705612244897957,
+      "loss": 0.1713,
+      "step": 12305
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.0001270408163265306,
+      "loss": 0.1519,
+      "step": 12306
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00012702551020408163,
+      "loss": 0.1877,
+      "step": 12307
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00012701020408163266,
+      "loss": 0.1073,
+      "step": 12308
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00012699489795918366,
+      "loss": 0.1837,
+      "step": 12309
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.0001269795918367347,
+      "loss": 0.0766,
+      "step": 12310
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.0001269642857142857,
+      "loss": 0.0879,
+      "step": 12311
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00012694897959183673,
+      "loss": 0.3299,
+      "step": 12312
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00012693367346938773,
+      "loss": 0.106,
+      "step": 12313
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.00012691836734693876,
+      "loss": 0.0695,
+      "step": 12314
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.0001269030612244898,
+      "loss": 0.104,
+      "step": 12315
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.00012688775510204082,
+      "loss": 0.1399,
+      "step": 12316
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.00012687244897959183,
+      "loss": 0.1547,
+      "step": 12317
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.00012685714285714286,
+      "loss": 0.1635,
+      "step": 12318
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.00012684183673469386,
+      "loss": 0.1094,
+      "step": 12319
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.0001268265306122449,
+      "loss": 0.1095,
+      "step": 12320
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.0001268112244897959,
+      "loss": 0.0603,
+      "step": 12321
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.00012679591836734693,
+      "loss": 0.0662,
+      "step": 12322
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.00012678061224489796,
+      "loss": 0.0518,
+      "step": 12323
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012676530612244896,
+      "loss": 0.1229,
+      "step": 12324
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012675,
+      "loss": 0.0667,
+      "step": 12325
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012673469387755102,
+      "loss": 0.0636,
+      "step": 12326
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012671938775510202,
+      "loss": 0.0575,
+      "step": 12327
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012670408163265306,
+      "loss": 0.1022,
+      "step": 12328
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012668877551020406,
+      "loss": 0.0382,
+      "step": 12329
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.0001266734693877551,
+      "loss": 0.0482,
+      "step": 12330
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.0001266581632653061,
+      "loss": 0.1854,
+      "step": 12331
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012664285714285712,
+      "loss": 0.1373,
+      "step": 12332
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012662755102040815,
+      "loss": 0.0662,
+      "step": 12333
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00012661224489795918,
+      "loss": 0.0679,
+      "step": 12334
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.0001265969387755102,
+      "loss": 0.1518,
+      "step": 12335
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012658163265306122,
+      "loss": 0.0606,
+      "step": 12336
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012656632653061222,
+      "loss": 0.2125,
+      "step": 12337
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012655102040816325,
+      "loss": 0.2109,
+      "step": 12338
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012653571428571426,
+      "loss": 0.0693,
+      "step": 12339
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.0001265204081632653,
+      "loss": 0.1375,
+      "step": 12340
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012650510204081632,
+      "loss": 0.0692,
+      "step": 12341
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012648979591836735,
+      "loss": 0.1292,
+      "step": 12342
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012647448979591835,
+      "loss": 0.3264,
+      "step": 12343
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012645918367346938,
+      "loss": 0.1166,
+      "step": 12344
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012644387755102039,
+      "loss": 0.133,
+      "step": 12345
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012642857142857142,
+      "loss": 0.0983,
+      "step": 12346
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012641326530612242,
+      "loss": 0.1145,
+      "step": 12347
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012639795918367345,
+      "loss": 0.092,
+      "step": 12348
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012638265306122448,
+      "loss": 0.123,
+      "step": 12349
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.0001263673469387755,
+      "loss": 0.1957,
+      "step": 12350
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012635204081632651,
+      "loss": 0.107,
+      "step": 12351
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012633673469387754,
+      "loss": 0.1422,
+      "step": 12352
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012632142857142858,
+      "loss": 0.141,
+      "step": 12353
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012630612244897958,
+      "loss": 0.042,
+      "step": 12354
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.0001262908163265306,
+      "loss": 0.2159,
+      "step": 12355
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.0001262755102040816,
+      "loss": 0.1233,
+      "step": 12356
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00012626020408163264,
+      "loss": 0.0748,
+      "step": 12357
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00012624489795918367,
+      "loss": 0.094,
+      "step": 12358
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00012622959183673468,
+      "loss": 0.116,
+      "step": 12359
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.0001262142857142857,
+      "loss": 0.0635,
+      "step": 12360
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00012619897959183674,
+      "loss": 0.1436,
+      "step": 12361
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00012618367346938774,
+      "loss": 0.0775,
+      "step": 12362
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00012616836734693877,
+      "loss": 0.0442,
+      "step": 12363
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00012615306122448978,
+      "loss": 0.0384,
+      "step": 12364
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.0001261377551020408,
+      "loss": 0.0456,
+      "step": 12365
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.0001261224489795918,
+      "loss": 0.0813,
+      "step": 12366
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00012610714285714284,
+      "loss": 0.0513,
+      "step": 12367
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00012609183673469387,
+      "loss": 0.0769,
+      "step": 12368
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.0001260765306122449,
+      "loss": 0.0881,
+      "step": 12369
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.0001260612244897959,
+      "loss": 0.0625,
+      "step": 12370
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00012604591836734694,
+      "loss": 0.0594,
+      "step": 12371
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00012603061224489794,
+      "loss": 0.1411,
+      "step": 12372
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00012601530612244897,
+      "loss": 0.1784,
+      "step": 12373
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00012599999999999997,
+      "loss": 0.0506,
+      "step": 12374
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.000125984693877551,
+      "loss": 0.0568,
+      "step": 12375
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00012596938775510203,
+      "loss": 0.0843,
+      "step": 12376
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00012595408163265306,
+      "loss": 0.0627,
+      "step": 12377
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00012593877551020407,
+      "loss": 0.1296,
+      "step": 12378
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.0001259234693877551,
+      "loss": 0.2091,
+      "step": 12379
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.0001259081632653061,
+      "loss": 0.1704,
+      "step": 12380
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00012589285714285713,
+      "loss": 0.1182,
+      "step": 12381
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00012587755102040814,
+      "loss": 0.0563,
+      "step": 12382
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00012586224489795917,
+      "loss": 0.0464,
+      "step": 12383
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.0001258469387755102,
+      "loss": 0.1128,
+      "step": 12384
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00012583163265306123,
+      "loss": 0.2247,
+      "step": 12385
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00012581632653061223,
+      "loss": 0.1878,
+      "step": 12386
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00012580102040816326,
+      "loss": 0.0882,
+      "step": 12387
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00012578571428571426,
+      "loss": 0.0866,
+      "step": 12388
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.0001257704081632653,
+      "loss": 0.1604,
+      "step": 12389
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.0001257551020408163,
+      "loss": 0.1526,
+      "step": 12390
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00012573979591836733,
+      "loss": 0.0896,
+      "step": 12391
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00012572448979591836,
+      "loss": 0.1111,
+      "step": 12392
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.0001257091836734694,
+      "loss": 0.1268,
+      "step": 12393
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.0001256938775510204,
+      "loss": 0.1226,
+      "step": 12394
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00012567857142857142,
+      "loss": 0.0903,
+      "step": 12395
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00012566326530612245,
+      "loss": 0.087,
+      "step": 12396
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00012564795918367346,
+      "loss": 0.05,
+      "step": 12397
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.0001256326530612245,
+      "loss": 0.063,
+      "step": 12398
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.0001256173469387755,
+      "loss": 0.1277,
+      "step": 12399
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00012560204081632652,
+      "loss": 0.1851,
+      "step": 12400
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00012558673469387753,
+      "loss": 0.1247,
+      "step": 12401
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00012557142857142856,
+      "loss": 0.0456,
+      "step": 12402
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.0001255561224489796,
+      "loss": 0.0516,
+      "step": 12403
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00012554081632653062,
+      "loss": 0.0891,
+      "step": 12404
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00012552551020408162,
+      "loss": 0.182,
+      "step": 12405
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00012551020408163265,
+      "loss": 0.0934,
+      "step": 12406
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012549489795918366,
+      "loss": 0.0405,
+      "step": 12407
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012547959183673469,
+      "loss": 0.1272,
+      "step": 12408
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.0001254642857142857,
+      "loss": 0.0408,
+      "step": 12409
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012544897959183672,
+      "loss": 0.0871,
+      "step": 12410
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012543367346938775,
+      "loss": 0.1521,
+      "step": 12411
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012541836734693878,
+      "loss": 0.0952,
+      "step": 12412
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012540306122448978,
+      "loss": 0.0898,
+      "step": 12413
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012538775510204082,
+      "loss": 0.0535,
+      "step": 12414
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012537244897959182,
+      "loss": 0.0508,
+      "step": 12415
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00012535714285714285,
+      "loss": 0.08,
+      "step": 12416
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012534183673469385,
+      "loss": 0.1119,
+      "step": 12417
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012532653061224488,
+      "loss": 0.0975,
+      "step": 12418
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012531122448979591,
+      "loss": 0.098,
+      "step": 12419
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012529591836734694,
+      "loss": 0.0829,
+      "step": 12420
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012528061224489795,
+      "loss": 0.0482,
+      "step": 12421
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012526530612244898,
+      "loss": 0.0787,
+      "step": 12422
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012524999999999998,
+      "loss": 0.0938,
+      "step": 12423
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.000125234693877551,
+      "loss": 0.0699,
+      "step": 12424
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012521938775510202,
+      "loss": 0.1354,
+      "step": 12425
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00012520408163265305,
+      "loss": 0.1001,
+      "step": 12426
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012518877551020408,
+      "loss": 0.0572,
+      "step": 12427
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012517346938775508,
+      "loss": 0.1086,
+      "step": 12428
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.0001251581632653061,
+      "loss": 0.2371,
+      "step": 12429
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012514285714285714,
+      "loss": 0.0573,
+      "step": 12430
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012512755102040814,
+      "loss": 0.0597,
+      "step": 12431
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012511224489795918,
+      "loss": 0.1332,
+      "step": 12432
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012509693877551018,
+      "loss": 0.1445,
+      "step": 12433
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.0001250816326530612,
+      "loss": 0.0532,
+      "step": 12434
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012506632653061224,
+      "loss": 0.0873,
+      "step": 12435
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012505102040816324,
+      "loss": 0.1083,
+      "step": 12436
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012503571428571427,
+      "loss": 0.074,
+      "step": 12437
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.0001250204081632653,
+      "loss": 0.0532,
+      "step": 12438
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.0001250051020408163,
+      "loss": 0.046,
+      "step": 12439
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00012498979591836734,
+      "loss": 0.0912,
+      "step": 12440
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00012497448979591837,
+      "loss": 0.1169,
+      "step": 12441
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00012495918367346937,
+      "loss": 0.1258,
+      "step": 12442
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.0001249438775510204,
+      "loss": 0.0907,
+      "step": 12443
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.0001249285714285714,
+      "loss": 0.1266,
+      "step": 12444
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00012491326530612244,
+      "loss": 0.043,
+      "step": 12445
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00012489795918367347,
+      "loss": 0.0542,
+      "step": 12446
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.0001248826530612245,
+      "loss": 0.0776,
+      "step": 12447
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.0001248673469387755,
+      "loss": 0.1855,
+      "step": 12448
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.00012485204081632653,
+      "loss": 0.1055,
+      "step": 12449
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.00012483673469387754,
+      "loss": 0.0908,
+      "step": 12450
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.00012482142857142857,
+      "loss": 0.05,
+      "step": 12451
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.00012480612244897957,
+      "loss": 0.0624,
+      "step": 12452
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.0001247908163265306,
+      "loss": 0.0513,
+      "step": 12453
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.00012477551020408163,
+      "loss": 0.1192,
+      "step": 12454
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.00012476020408163266,
+      "loss": 0.1431,
+      "step": 12455
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.00012474489795918366,
+      "loss": 0.0606,
+      "step": 12456
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.0001247295918367347,
+      "loss": 0.2103,
+      "step": 12457
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0001247142857142857,
+      "loss": 0.09,
+      "step": 12458
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00012469897959183673,
+      "loss": 0.0445,
+      "step": 12459
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00012468367346938773,
+      "loss": 0.0822,
+      "step": 12460
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00012466836734693876,
+      "loss": 0.2394,
+      "step": 12461
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0001246530612244898,
+      "loss": 0.1098,
+      "step": 12462
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0001246377551020408,
+      "loss": 0.0834,
+      "step": 12463
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00012462244897959183,
+      "loss": 0.1052,
+      "step": 12464
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00012460714285714286,
+      "loss": 0.0918,
+      "step": 12465
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00012459183673469386,
+      "loss": 0.1298,
+      "step": 12466
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0001245765306122449,
+      "loss": 0.1513,
+      "step": 12467
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0001245612244897959,
+      "loss": 0.0982,
+      "step": 12468
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00012454591836734693,
+      "loss": 0.081,
+      "step": 12469
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00012453061224489793,
+      "loss": 0.0901,
+      "step": 12470
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00012451530612244896,
+      "loss": 0.0431,
+      "step": 12471
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0001245,
+      "loss": 0.0811,
+      "step": 12472
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00012448469387755102,
+      "loss": 0.0973,
+      "step": 12473
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00012446938775510202,
+      "loss": 0.0656,
+      "step": 12474
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00012445408163265305,
+      "loss": 0.0864,
+      "step": 12475
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00012443877551020406,
+      "loss": 0.1256,
+      "step": 12476
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0001244234693877551,
+      "loss": 0.1424,
+      "step": 12477
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0001244081632653061,
+      "loss": 0.0508,
+      "step": 12478
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012439285714285712,
+      "loss": 0.1033,
+      "step": 12479
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012437755102040815,
+      "loss": 0.1234,
+      "step": 12480
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012436224489795918,
+      "loss": 0.0566,
+      "step": 12481
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.0001243469387755102,
+      "loss": 0.1232,
+      "step": 12482
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012433163265306122,
+      "loss": 0.0896,
+      "step": 12483
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012431632653061222,
+      "loss": 0.1981,
+      "step": 12484
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012430102040816325,
+      "loss": 0.0819,
+      "step": 12485
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012428571428571428,
+      "loss": 0.1917,
+      "step": 12486
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012427040816326529,
+      "loss": 0.0667,
+      "step": 12487
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012425510204081632,
+      "loss": 0.1081,
+      "step": 12488
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012423979591836735,
+      "loss": 0.0675,
+      "step": 12489
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012422448979591835,
+      "loss": 0.0696,
+      "step": 12490
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012420918367346938,
+      "loss": 0.1247,
+      "step": 12491
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.0001241938775510204,
+      "loss": 0.0832,
+      "step": 12492
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012417857142857142,
+      "loss": 0.108,
+      "step": 12493
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012416326530612245,
+      "loss": 0.0527,
+      "step": 12494
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012414795918367345,
+      "loss": 0.1953,
+      "step": 12495
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012413265306122448,
+      "loss": 0.2114,
+      "step": 12496
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.0001241173469387755,
+      "loss": 0.0597,
+      "step": 12497
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012410204081632651,
+      "loss": 0.2189,
+      "step": 12498
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012408673469387754,
+      "loss": 0.1381,
+      "step": 12499
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00012407142857142857,
+      "loss": 0.059,
+      "step": 12500
+    },
+    {
+      "epoch": 12.14,
+      "eval_cer": 0.2530611854135353,
+      "eval_loss": 0.9971762299537659,
+      "eval_runtime": 238.669,
+      "eval_samples_per_second": 36.93,
+      "eval_steps_per_second": 4.617,
+      "eval_wer": 0.6573839106218897,
+      "step": 12500
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00012405612244897958,
+      "loss": 0.0702,
+      "step": 12501
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.0001240408163265306,
+      "loss": 0.1318,
+      "step": 12502
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.0001240255102040816,
+      "loss": 0.0802,
+      "step": 12503
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00012401020408163264,
+      "loss": 0.1835,
+      "step": 12504
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00012399489795918365,
+      "loss": 0.1879,
+      "step": 12505
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00012397959183673468,
+      "loss": 0.111,
+      "step": 12506
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.0001239642857142857,
+      "loss": 0.0852,
+      "step": 12507
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00012394897959183674,
+      "loss": 0.1042,
+      "step": 12508
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00012393367346938774,
+      "loss": 0.1321,
+      "step": 12509
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00012391836734693877,
+      "loss": 0.2104,
+      "step": 12510
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00012390306122448978,
+      "loss": 0.2885,
+      "step": 12511
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.0001238877551020408,
+      "loss": 0.149,
+      "step": 12512
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.0001238724489795918,
+      "loss": 0.1275,
+      "step": 12513
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00012385714285714284,
+      "loss": 0.0921,
+      "step": 12514
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00012384183673469387,
+      "loss": 0.0492,
+      "step": 12515
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.0001238265306122449,
+      "loss": 0.0993,
+      "step": 12516
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.0001238112244897959,
+      "loss": 0.2229,
+      "step": 12517
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00012379591836734693,
+      "loss": 0.1414,
+      "step": 12518
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00012378061224489794,
+      "loss": 0.0976,
+      "step": 12519
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00012376530612244897,
+      "loss": 0.0889,
+      "step": 12520
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00012374999999999997,
+      "loss": 0.0945,
+      "step": 12521
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.000123734693877551,
+      "loss": 0.0453,
+      "step": 12522
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00012371938775510203,
+      "loss": 0.1337,
+      "step": 12523
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00012370408163265306,
+      "loss": 0.0778,
+      "step": 12524
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00012368877551020407,
+      "loss": 0.0485,
+      "step": 12525
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.0001236734693877551,
+      "loss": 0.1368,
+      "step": 12526
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.0001236581632653061,
+      "loss": 0.0993,
+      "step": 12527
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00012364285714285713,
+      "loss": 0.0825,
+      "step": 12528
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00012362755102040814,
+      "loss": 0.0495,
+      "step": 12529
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00012361224489795917,
+      "loss": 0.145,
+      "step": 12530
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0001235969387755102,
+      "loss": 0.1264,
+      "step": 12531
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0001235816326530612,
+      "loss": 0.0737,
+      "step": 12532
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00012356632653061223,
+      "loss": 0.0489,
+      "step": 12533
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00012355102040816326,
+      "loss": 0.0542,
+      "step": 12534
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0001235357142857143,
+      "loss": 0.0593,
+      "step": 12535
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0001235204081632653,
+      "loss": 0.1346,
+      "step": 12536
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00012350510204081633,
+      "loss": 0.174,
+      "step": 12537
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00012348979591836733,
+      "loss": 0.0976,
+      "step": 12538
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00012347448979591836,
+      "loss": 0.073,
+      "step": 12539
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00012345918367346936,
+      "loss": 0.0758,
+      "step": 12540
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.0001234438775510204,
+      "loss": 0.1727,
+      "step": 12541
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.00012342857142857142,
+      "loss": 0.2719,
+      "step": 12542
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.00012341326530612245,
+      "loss": 0.0574,
+      "step": 12543
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.00012339795918367346,
+      "loss": 0.0554,
+      "step": 12544
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.0001233826530612245,
+      "loss": 0.0717,
+      "step": 12545
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.0001233673469387755,
+      "loss": 0.1084,
+      "step": 12546
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.00012335204081632652,
+      "loss": 0.0474,
+      "step": 12547
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.00012333673469387753,
+      "loss": 0.064,
+      "step": 12548
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.00012332142857142856,
+      "loss": 0.18,
+      "step": 12549
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.0001233061224489796,
+      "loss": 0.1033,
+      "step": 12550
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012329081632653062,
+      "loss": 0.1734,
+      "step": 12551
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012327551020408162,
+      "loss": 0.0962,
+      "step": 12552
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012326020408163265,
+      "loss": 0.0911,
+      "step": 12553
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012324489795918365,
+      "loss": 0.2616,
+      "step": 12554
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012322959183673469,
+      "loss": 0.0646,
+      "step": 12555
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.0001232142857142857,
+      "loss": 0.1334,
+      "step": 12556
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012319897959183672,
+      "loss": 0.0658,
+      "step": 12557
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012318367346938775,
+      "loss": 0.0498,
+      "step": 12558
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012316836734693878,
+      "loss": 0.0866,
+      "step": 12559
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012315306122448978,
+      "loss": 0.1536,
+      "step": 12560
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012313775510204081,
+      "loss": 0.2271,
+      "step": 12561
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012312244897959182,
+      "loss": 0.0715,
+      "step": 12562
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012310714285714285,
+      "loss": 0.0531,
+      "step": 12563
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012309183673469385,
+      "loss": 0.0463,
+      "step": 12564
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012307653061224488,
+      "loss": 0.1277,
+      "step": 12565
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.0001230612244897959,
+      "loss": 0.0512,
+      "step": 12566
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012304591836734692,
+      "loss": 0.244,
+      "step": 12567
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012303061224489795,
+      "loss": 0.1711,
+      "step": 12568
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012301530612244898,
+      "loss": 0.0527,
+      "step": 12569
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00012299999999999998,
+      "loss": 0.0467,
+      "step": 12570
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.000122984693877551,
+      "loss": 0.0992,
+      "step": 12571
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00012296938775510202,
+      "loss": 0.0787,
+      "step": 12572
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00012295408163265305,
+      "loss": 0.1871,
+      "step": 12573
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00012293877551020408,
+      "loss": 0.0634,
+      "step": 12574
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00012292346938775508,
+      "loss": 0.0736,
+      "step": 12575
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.0001229081632653061,
+      "loss": 0.0842,
+      "step": 12576
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00012289285714285714,
+      "loss": 0.0879,
+      "step": 12577
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00012287755102040817,
+      "loss": 0.0582,
+      "step": 12578
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00012286224489795917,
+      "loss": 0.1293,
+      "step": 12579
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.0001228469387755102,
+      "loss": 0.1039,
+      "step": 12580
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.0001228316326530612,
+      "loss": 0.0562,
+      "step": 12581
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00012281632653061224,
+      "loss": 0.0571,
+      "step": 12582
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00012280102040816324,
+      "loss": 0.0832,
+      "step": 12583
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00012278571428571427,
+      "loss": 0.0801,
+      "step": 12584
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.0001227704081632653,
+      "loss": 0.0651,
+      "step": 12585
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00012275510204081633,
+      "loss": 0.1154,
+      "step": 12586
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00012273979591836734,
+      "loss": 0.0471,
+      "step": 12587
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00012272448979591837,
+      "loss": 0.0356,
+      "step": 12588
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00012270918367346937,
+      "loss": 0.1335,
+      "step": 12589
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.0001226938775510204,
+      "loss": 0.057,
+      "step": 12590
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.0001226785714285714,
+      "loss": 0.0934,
+      "step": 12591
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00012266326530612244,
+      "loss": 0.1053,
+      "step": 12592
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00012264795918367347,
+      "loss": 0.057,
+      "step": 12593
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.0001226326530612245,
+      "loss": 0.1727,
+      "step": 12594
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.0001226173469387755,
+      "loss": 0.1343,
+      "step": 12595
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00012260204081632653,
+      "loss": 0.0401,
+      "step": 12596
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00012258673469387753,
+      "loss": 0.1108,
+      "step": 12597
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00012257142857142857,
+      "loss": 0.1786,
+      "step": 12598
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00012255612244897957,
+      "loss": 0.1033,
+      "step": 12599
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.0001225408163265306,
+      "loss": 0.1772,
+      "step": 12600
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00012252551020408163,
+      "loss": 0.0881,
+      "step": 12601
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00012251020408163263,
+      "loss": 0.0509,
+      "step": 12602
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00012249489795918366,
+      "loss": 0.0476,
+      "step": 12603
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.0001224795918367347,
+      "loss": 0.1813,
+      "step": 12604
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.0001224642857142857,
+      "loss": 0.0633,
+      "step": 12605
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00012244897959183673,
+      "loss": 0.088,
+      "step": 12606
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00012243367346938773,
+      "loss": 0.0691,
+      "step": 12607
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00012241836734693876,
+      "loss": 0.0811,
+      "step": 12608
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00012240306122448977,
+      "loss": 0.0974,
+      "step": 12609
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.0001223877551020408,
+      "loss": 0.225,
+      "step": 12610
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00012237244897959183,
+      "loss": 0.0674,
+      "step": 12611
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00012235714285714286,
+      "loss": 0.0902,
+      "step": 12612
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00012234183673469386,
+      "loss": 0.1278,
+      "step": 12613
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.0001223265306122449,
+      "loss": 0.0428,
+      "step": 12614
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.0001223112244897959,
+      "loss": 0.0425,
+      "step": 12615
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00012229591836734693,
+      "loss": 0.1937,
+      "step": 12616
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00012228061224489793,
+      "loss": 0.1814,
+      "step": 12617
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00012226530612244896,
+      "loss": 0.0564,
+      "step": 12618
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00012225,
+      "loss": 0.0842,
+      "step": 12619
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00012223469387755102,
+      "loss": 0.1055,
+      "step": 12620
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00012221938775510205,
+      "loss": 0.1169,
+      "step": 12621
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00012220408163265305,
+      "loss": 0.0827,
+      "step": 12622
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00012218877551020409,
+      "loss": 0.2196,
+      "step": 12623
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.0001221734693877551,
+      "loss": 0.239,
+      "step": 12624
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00012215816326530612,
+      "loss": 0.1517,
+      "step": 12625
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00012214285714285712,
+      "loss": 0.0907,
+      "step": 12626
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00012212755102040815,
+      "loss": 0.071,
+      "step": 12627
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00012211224489795918,
+      "loss": 0.1157,
+      "step": 12628
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.0001220969387755102,
+      "loss": 0.144,
+      "step": 12629
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00012208163265306122,
+      "loss": 0.1305,
+      "step": 12630
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00012206632653061223,
+      "loss": 0.1018,
+      "step": 12631
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00012205102040816325,
+      "loss": 0.0448,
+      "step": 12632
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012203571428571428,
+      "loss": 0.0389,
+      "step": 12633
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.0001220204081632653,
+      "loss": 0.0616,
+      "step": 12634
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012200510204081632,
+      "loss": 0.176,
+      "step": 12635
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012198979591836733,
+      "loss": 0.0594,
+      "step": 12636
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012197448979591836,
+      "loss": 0.0934,
+      "step": 12637
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012195918367346938,
+      "loss": 0.055,
+      "step": 12638
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.0001219438775510204,
+      "loss": 0.042,
+      "step": 12639
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012192857142857141,
+      "loss": 0.0464,
+      "step": 12640
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012191326530612245,
+      "loss": 0.1391,
+      "step": 12641
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012189795918367345,
+      "loss": 0.1385,
+      "step": 12642
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00012188265306122448,
+      "loss": 0.0583,
+      "step": 12643
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.0001218673469387755,
+      "loss": 0.053,
+      "step": 12644
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012185204081632653,
+      "loss": 0.0991,
+      "step": 12645
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012183673469387753,
+      "loss": 0.0568,
+      "step": 12646
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012182142857142856,
+      "loss": 0.0974,
+      "step": 12647
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012180612244897958,
+      "loss": 0.2617,
+      "step": 12648
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012179081632653061,
+      "loss": 0.1244,
+      "step": 12649
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012177551020408161,
+      "loss": 0.083,
+      "step": 12650
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012176020408163264,
+      "loss": 0.0671,
+      "step": 12651
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012174489795918366,
+      "loss": 0.091,
+      "step": 12652
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00012172959183673469,
+      "loss": 0.0984,
+      "step": 12653
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012171428571428569,
+      "loss": 0.1781,
+      "step": 12654
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012169897959183672,
+      "loss": 0.0724,
+      "step": 12655
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012168367346938774,
+      "loss": 0.115,
+      "step": 12656
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012166836734693877,
+      "loss": 0.1454,
+      "step": 12657
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012165306122448977,
+      "loss": 0.1634,
+      "step": 12658
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.0001216377551020408,
+      "loss": 0.0602,
+      "step": 12659
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012162244897959182,
+      "loss": 0.0911,
+      "step": 12660
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012160714285714285,
+      "loss": 0.1993,
+      "step": 12661
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012159183673469386,
+      "loss": 0.0581,
+      "step": 12662
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00012157653061224489,
+      "loss": 0.1298,
+      "step": 12663
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.0001215612244897959,
+      "loss": 0.0979,
+      "step": 12664
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012154591836734693,
+      "loss": 0.0685,
+      "step": 12665
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012153061224489794,
+      "loss": 0.153,
+      "step": 12666
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012151530612244897,
+      "loss": 0.1115,
+      "step": 12667
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.0001215,
+      "loss": 0.0948,
+      "step": 12668
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012148469387755102,
+      "loss": 0.084,
+      "step": 12669
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012146938775510203,
+      "loss": 0.1327,
+      "step": 12670
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012145408163265305,
+      "loss": 0.0367,
+      "step": 12671
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012143877551020408,
+      "loss": 0.1436,
+      "step": 12672
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012142346938775508,
+      "loss": 0.1471,
+      "step": 12673
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00012140816326530611,
+      "loss": 0.2132,
+      "step": 12674
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00012139285714285713,
+      "loss": 0.137,
+      "step": 12675
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00012137755102040816,
+      "loss": 0.0545,
+      "step": 12676
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00012136224489795917,
+      "loss": 0.0734,
+      "step": 12677
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.0001213469387755102,
+      "loss": 0.0473,
+      "step": 12678
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00012133163265306121,
+      "loss": 0.0968,
+      "step": 12679
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00012131632653061224,
+      "loss": 0.0983,
+      "step": 12680
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00012130102040816325,
+      "loss": 0.0641,
+      "step": 12681
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00012128571428571428,
+      "loss": 0.0888,
+      "step": 12682
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.0001212704081632653,
+      "loss": 0.0995,
+      "step": 12683
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00012125510204081632,
+      "loss": 0.0761,
+      "step": 12684
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012123979591836733,
+      "loss": 0.0648,
+      "step": 12685
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012122448979591836,
+      "loss": 0.2199,
+      "step": 12686
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012120918367346938,
+      "loss": 0.1395,
+      "step": 12687
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.0001211938775510204,
+      "loss": 0.0519,
+      "step": 12688
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012117857142857141,
+      "loss": 0.1158,
+      "step": 12689
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012116326530612244,
+      "loss": 0.0636,
+      "step": 12690
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012114795918367346,
+      "loss": 0.1018,
+      "step": 12691
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012113265306122449,
+      "loss": 0.0854,
+      "step": 12692
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012111734693877549,
+      "loss": 0.161,
+      "step": 12693
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00012110204081632652,
+      "loss": 0.0895,
+      "step": 12694
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012108673469387754,
+      "loss": 0.1294,
+      "step": 12695
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012107142857142857,
+      "loss": 0.0856,
+      "step": 12696
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012105612244897957,
+      "loss": 0.051,
+      "step": 12697
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.0001210408163265306,
+      "loss": 0.1296,
+      "step": 12698
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012102551020408162,
+      "loss": 0.1853,
+      "step": 12699
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012101020408163265,
+      "loss": 0.0527,
+      "step": 12700
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012099489795918365,
+      "loss": 0.0811,
+      "step": 12701
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012097959183673469,
+      "loss": 0.1151,
+      "step": 12702
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.0001209642857142857,
+      "loss": 0.0828,
+      "step": 12703
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012094897959183672,
+      "loss": 0.0838,
+      "step": 12704
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00012093367346938774,
+      "loss": 0.1098,
+      "step": 12705
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012091836734693877,
+      "loss": 0.0959,
+      "step": 12706
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012090306122448978,
+      "loss": 0.1315,
+      "step": 12707
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.0001208877551020408,
+      "loss": 0.0846,
+      "step": 12708
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012087244897959182,
+      "loss": 0.0439,
+      "step": 12709
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012085714285714285,
+      "loss": 0.0871,
+      "step": 12710
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012084183673469387,
+      "loss": 0.1935,
+      "step": 12711
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012082653061224488,
+      "loss": 0.0575,
+      "step": 12712
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012081122448979591,
+      "loss": 0.0604,
+      "step": 12713
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012079591836734693,
+      "loss": 0.0337,
+      "step": 12714
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00012078061224489796,
+      "loss": 0.0873,
+      "step": 12715
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012076530612244896,
+      "loss": 0.0385,
+      "step": 12716
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012075,
+      "loss": 0.22,
+      "step": 12717
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012073469387755101,
+      "loss": 0.1137,
+      "step": 12718
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012071938775510204,
+      "loss": 0.0426,
+      "step": 12719
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012070408163265305,
+      "loss": 0.1789,
+      "step": 12720
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012068877551020408,
+      "loss": 0.1736,
+      "step": 12721
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012067346938775509,
+      "loss": 0.0789,
+      "step": 12722
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012065816326530612,
+      "loss": 0.0905,
+      "step": 12723
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012064285714285713,
+      "loss": 0.0512,
+      "step": 12724
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00012062755102040816,
+      "loss": 0.0403,
+      "step": 12725
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012061224489795917,
+      "loss": 0.0474,
+      "step": 12726
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.0001205969387755102,
+      "loss": 0.0968,
+      "step": 12727
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012058163265306121,
+      "loss": 0.0881,
+      "step": 12728
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012056632653061224,
+      "loss": 0.1735,
+      "step": 12729
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012055102040816326,
+      "loss": 0.0974,
+      "step": 12730
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012053571428571429,
+      "loss": 0.0962,
+      "step": 12731
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012052040816326529,
+      "loss": 0.049,
+      "step": 12732
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012050510204081632,
+      "loss": 0.0406,
+      "step": 12733
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012048979591836734,
+      "loss": 0.085,
+      "step": 12734
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00012047448979591835,
+      "loss": 0.1547,
+      "step": 12735
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012045918367346937,
+      "loss": 0.1528,
+      "step": 12736
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.0001204438775510204,
+      "loss": 0.1671,
+      "step": 12737
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012042857142857142,
+      "loss": 0.1455,
+      "step": 12738
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012041326530612244,
+      "loss": 0.0448,
+      "step": 12739
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012039795918367345,
+      "loss": 0.1499,
+      "step": 12740
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012038265306122448,
+      "loss": 0.1099,
+      "step": 12741
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.0001203673469387755,
+      "loss": 0.2098,
+      "step": 12742
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012035204081632652,
+      "loss": 0.0546,
+      "step": 12743
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012033673469387753,
+      "loss": 0.0892,
+      "step": 12744
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012032142857142856,
+      "loss": 0.1252,
+      "step": 12745
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00012030612244897957,
+      "loss": 0.0339,
+      "step": 12746
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.0001202908163265306,
+      "loss": 0.0419,
+      "step": 12747
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00012027551020408162,
+      "loss": 0.1927,
+      "step": 12748
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00012026020408163265,
+      "loss": 0.0493,
+      "step": 12749
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00012024489795918365,
+      "loss": 0.0528,
+      "step": 12750
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00012022959183673468,
+      "loss": 0.0826,
+      "step": 12751
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.0001202142857142857,
+      "loss": 0.1197,
+      "step": 12752
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00012019897959183673,
+      "loss": 0.1202,
+      "step": 12753
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00012018367346938773,
+      "loss": 0.163,
+      "step": 12754
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00012016836734693876,
+      "loss": 0.0597,
+      "step": 12755
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00012015306122448978,
+      "loss": 0.0933,
+      "step": 12756
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012013775510204081,
+      "loss": 0.0384,
+      "step": 12757
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012012244897959184,
+      "loss": 0.0819,
+      "step": 12758
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012010714285714284,
+      "loss": 0.0501,
+      "step": 12759
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012009183673469387,
+      "loss": 0.2518,
+      "step": 12760
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012007653061224489,
+      "loss": 0.0554,
+      "step": 12761
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012006122448979592,
+      "loss": 0.0462,
+      "step": 12762
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012004591836734692,
+      "loss": 0.0505,
+      "step": 12763
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012003061224489796,
+      "loss": 0.1413,
+      "step": 12764
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00012001530612244897,
+      "loss": 0.113,
+      "step": 12765
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 0.0544,
+      "step": 12766
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.000119984693877551,
+      "loss": 0.1225,
+      "step": 12767
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011996938775510204,
+      "loss": 0.0935,
+      "step": 12768
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011995408163265305,
+      "loss": 0.0837,
+      "step": 12769
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011993877551020407,
+      "loss": 0.0543,
+      "step": 12770
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011992346938775509,
+      "loss": 0.0485,
+      "step": 12771
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011990816326530612,
+      "loss": 0.0982,
+      "step": 12772
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011989285714285714,
+      "loss": 0.1293,
+      "step": 12773
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011987755102040815,
+      "loss": 0.1458,
+      "step": 12774
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011986224489795917,
+      "loss": 0.0969,
+      "step": 12775
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.0001198469387755102,
+      "loss": 0.1372,
+      "step": 12776
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011983163265306122,
+      "loss": 0.1656,
+      "step": 12777
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011981632653061223,
+      "loss": 0.0644,
+      "step": 12778
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011980102040816325,
+      "loss": 0.1884,
+      "step": 12779
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011978571428571428,
+      "loss": 0.0676,
+      "step": 12780
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011977040816326529,
+      "loss": 0.1426,
+      "step": 12781
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011975510204081632,
+      "loss": 0.2198,
+      "step": 12782
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011973979591836733,
+      "loss": 0.1151,
+      "step": 12783
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011972448979591836,
+      "loss": 0.0456,
+      "step": 12784
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011970918367346937,
+      "loss": 0.0922,
+      "step": 12785
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.0001196938775510204,
+      "loss": 0.1513,
+      "step": 12786
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011967857142857141,
+      "loss": 0.0568,
+      "step": 12787
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011966326530612244,
+      "loss": 0.0447,
+      "step": 12788
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011964795918367345,
+      "loss": 0.1239,
+      "step": 12789
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011963265306122448,
+      "loss": 0.0806,
+      "step": 12790
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.0001196173469387755,
+      "loss": 0.0592,
+      "step": 12791
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011960204081632653,
+      "loss": 0.1175,
+      "step": 12792
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011958673469387753,
+      "loss": 0.1285,
+      "step": 12793
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011957142857142856,
+      "loss": 0.1045,
+      "step": 12794
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011955612244897958,
+      "loss": 0.1379,
+      "step": 12795
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011954081632653061,
+      "loss": 0.0574,
+      "step": 12796
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011952551020408161,
+      "loss": 0.1173,
+      "step": 12797
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011951020408163264,
+      "loss": 0.2235,
+      "step": 12798
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011949489795918366,
+      "loss": 0.1394,
+      "step": 12799
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011947959183673469,
+      "loss": 0.1034,
+      "step": 12800
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011946428571428569,
+      "loss": 0.1233,
+      "step": 12801
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011944897959183672,
+      "loss": 0.1653,
+      "step": 12802
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011943367346938775,
+      "loss": 0.0764,
+      "step": 12803
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011941836734693877,
+      "loss": 0.1936,
+      "step": 12804
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011940306122448979,
+      "loss": 0.0991,
+      "step": 12805
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.0001193877551020408,
+      "loss": 0.0603,
+      "step": 12806
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011937244897959184,
+      "loss": 0.0837,
+      "step": 12807
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00011935714285714285,
+      "loss": 0.0909,
+      "step": 12808
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011934183673469387,
+      "loss": 0.1798,
+      "step": 12809
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011932653061224489,
+      "loss": 0.2257,
+      "step": 12810
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011931122448979592,
+      "loss": 0.1194,
+      "step": 12811
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011929591836734692,
+      "loss": 0.1294,
+      "step": 12812
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011928061224489795,
+      "loss": 0.1023,
+      "step": 12813
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011926530612244897,
+      "loss": 0.0893,
+      "step": 12814
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011925,
+      "loss": 0.0876,
+      "step": 12815
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.000119234693877551,
+      "loss": 0.0659,
+      "step": 12816
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011921938775510203,
+      "loss": 0.2772,
+      "step": 12817
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011920408163265305,
+      "loss": 0.147,
+      "step": 12818
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011918877551020408,
+      "loss": 0.1013,
+      "step": 12819
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011917346938775508,
+      "loss": 0.0962,
+      "step": 12820
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011915816326530611,
+      "loss": 0.0582,
+      "step": 12821
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011914285714285713,
+      "loss": 0.0734,
+      "step": 12822
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011912755102040816,
+      "loss": 0.2253,
+      "step": 12823
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011911224489795916,
+      "loss": 0.0934,
+      "step": 12824
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.0001190969387755102,
+      "loss": 0.1235,
+      "step": 12825
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011908163265306121,
+      "loss": 0.1205,
+      "step": 12826
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011906632653061224,
+      "loss": 0.0618,
+      "step": 12827
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.00011905102040816325,
+      "loss": 0.0495,
+      "step": 12828
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011903571428571428,
+      "loss": 0.1728,
+      "step": 12829
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.0001190204081632653,
+      "loss": 0.16,
+      "step": 12830
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011900510204081632,
+      "loss": 0.0348,
+      "step": 12831
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011898979591836733,
+      "loss": 0.0639,
+      "step": 12832
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011897448979591836,
+      "loss": 0.1243,
+      "step": 12833
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011895918367346938,
+      "loss": 0.0435,
+      "step": 12834
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.0001189438775510204,
+      "loss": 0.1632,
+      "step": 12835
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011892857142857141,
+      "loss": 0.1517,
+      "step": 12836
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011891326530612244,
+      "loss": 0.0565,
+      "step": 12837
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011889795918367346,
+      "loss": 0.053,
+      "step": 12838
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011888265306122449,
+      "loss": 0.0837,
+      "step": 12839
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011886734693877549,
+      "loss": 0.1259,
+      "step": 12840
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011885204081632652,
+      "loss": 0.1656,
+      "step": 12841
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011883673469387754,
+      "loss": 0.195,
+      "step": 12842
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011882142857142856,
+      "loss": 0.1303,
+      "step": 12843
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011880612244897957,
+      "loss": 0.0785,
+      "step": 12844
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.0001187908163265306,
+      "loss": 0.0535,
+      "step": 12845
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011877551020408162,
+      "loss": 0.069,
+      "step": 12846
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011876020408163264,
+      "loss": 0.0923,
+      "step": 12847
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011874489795918367,
+      "loss": 0.1391,
+      "step": 12848
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011872959183673468,
+      "loss": 0.0578,
+      "step": 12849
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00011871428571428572,
+      "loss": 0.2157,
+      "step": 12850
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00011869897959183672,
+      "loss": 0.072,
+      "step": 12851
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00011868367346938775,
+      "loss": 0.1064,
+      "step": 12852
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00011866836734693877,
+      "loss": 0.0812,
+      "step": 12853
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.0001186530612244898,
+      "loss": 0.0733,
+      "step": 12854
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.0001186377551020408,
+      "loss": 0.1766,
+      "step": 12855
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00011862244897959183,
+      "loss": 0.0925,
+      "step": 12856
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00011860714285714285,
+      "loss": 0.2078,
+      "step": 12857
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00011859183673469388,
+      "loss": 0.0432,
+      "step": 12858
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00011857653061224488,
+      "loss": 0.0485,
+      "step": 12859
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011856122448979591,
+      "loss": 0.1817,
+      "step": 12860
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011854591836734693,
+      "loss": 0.0709,
+      "step": 12861
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011853061224489796,
+      "loss": 0.0623,
+      "step": 12862
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011851530612244896,
+      "loss": 0.0757,
+      "step": 12863
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.0001185,
+      "loss": 0.1001,
+      "step": 12864
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011848469387755101,
+      "loss": 0.0931,
+      "step": 12865
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011846938775510204,
+      "loss": 0.1461,
+      "step": 12866
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011845408163265304,
+      "loss": 0.1445,
+      "step": 12867
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011843877551020408,
+      "loss": 0.0572,
+      "step": 12868
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011842346938775509,
+      "loss": 0.1041,
+      "step": 12869
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011840816326530612,
+      "loss": 0.1035,
+      "step": 12870
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011839285714285713,
+      "loss": 0.0757,
+      "step": 12871
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011837755102040816,
+      "loss": 0.0331,
+      "step": 12872
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011836224489795917,
+      "loss": 0.1076,
+      "step": 12873
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011834693877551019,
+      "loss": 0.1036,
+      "step": 12874
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011833163265306121,
+      "loss": 0.069,
+      "step": 12875
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011831632653061224,
+      "loss": 0.0605,
+      "step": 12876
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011830102040816326,
+      "loss": 0.0813,
+      "step": 12877
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011828571428571427,
+      "loss": 0.0472,
+      "step": 12878
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011827040816326529,
+      "loss": 0.1856,
+      "step": 12879
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011825510204081632,
+      "loss": 0.0955,
+      "step": 12880
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011823979591836734,
+      "loss": 0.1406,
+      "step": 12881
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011822448979591835,
+      "loss": 0.0992,
+      "step": 12882
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011820918367346937,
+      "loss": 0.1424,
+      "step": 12883
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.0001181938775510204,
+      "loss": 0.0544,
+      "step": 12884
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.0001181785714285714,
+      "loss": 0.1791,
+      "step": 12885
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011816326530612244,
+      "loss": 0.1516,
+      "step": 12886
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011814795918367345,
+      "loss": 0.0643,
+      "step": 12887
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011813265306122448,
+      "loss": 0.0556,
+      "step": 12888
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011811734693877549,
+      "loss": 0.0761,
+      "step": 12889
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011810204081632652,
+      "loss": 0.1157,
+      "step": 12890
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011808673469387753,
+      "loss": 0.1445,
+      "step": 12891
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011807142857142856,
+      "loss": 0.3056,
+      "step": 12892
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011805612244897957,
+      "loss": 0.0636,
+      "step": 12893
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.0001180408163265306,
+      "loss": 0.0531,
+      "step": 12894
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011802551020408163,
+      "loss": 0.0929,
+      "step": 12895
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011801020408163265,
+      "loss": 0.1007,
+      "step": 12896
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011799489795918368,
+      "loss": 0.1187,
+      "step": 12897
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011797959183673468,
+      "loss": 0.2217,
+      "step": 12898
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011796428571428571,
+      "loss": 0.0577,
+      "step": 12899
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00011794897959183673,
+      "loss": 0.0588,
+      "step": 12900
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011793367346938776,
+      "loss": 0.0638,
+      "step": 12901
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011791836734693876,
+      "loss": 0.0988,
+      "step": 12902
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011790306122448979,
+      "loss": 0.0703,
+      "step": 12903
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011788775510204081,
+      "loss": 0.1617,
+      "step": 12904
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011787244897959183,
+      "loss": 0.0925,
+      "step": 12905
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011785714285714284,
+      "loss": 0.104,
+      "step": 12906
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011784183673469387,
+      "loss": 0.0839,
+      "step": 12907
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011782653061224489,
+      "loss": 0.056,
+      "step": 12908
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011781122448979591,
+      "loss": 0.0856,
+      "step": 12909
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011779591836734692,
+      "loss": 0.0834,
+      "step": 12910
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011778061224489796,
+      "loss": 0.0734,
+      "step": 12911
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011776530612244897,
+      "loss": 0.0777,
+      "step": 12912
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011774999999999999,
+      "loss": 0.1105,
+      "step": 12913
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.000117734693877551,
+      "loss": 0.0422,
+      "step": 12914
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011771938775510204,
+      "loss": 0.0497,
+      "step": 12915
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011770408163265304,
+      "loss": 0.1744,
+      "step": 12916
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011768877551020407,
+      "loss": 0.246,
+      "step": 12917
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011767346938775509,
+      "loss": 0.1141,
+      "step": 12918
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011765816326530612,
+      "loss": 0.096,
+      "step": 12919
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011764285714285712,
+      "loss": 0.1274,
+      "step": 12920
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011762755102040815,
+      "loss": 0.1526,
+      "step": 12921
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011761224489795917,
+      "loss": 0.1643,
+      "step": 12922
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.0001175969387755102,
+      "loss": 0.2027,
+      "step": 12923
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.0001175816326530612,
+      "loss": 0.0958,
+      "step": 12924
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011756632653061223,
+      "loss": 0.0937,
+      "step": 12925
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011755102040816325,
+      "loss": 0.1825,
+      "step": 12926
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011753571428571428,
+      "loss": 0.0616,
+      "step": 12927
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011752040816326528,
+      "loss": 0.0939,
+      "step": 12928
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011750510204081632,
+      "loss": 0.195,
+      "step": 12929
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011748979591836733,
+      "loss": 0.1832,
+      "step": 12930
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011747448979591836,
+      "loss": 0.0413,
+      "step": 12931
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011745918367346937,
+      "loss": 0.0857,
+      "step": 12932
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.0001174438775510204,
+      "loss": 0.0601,
+      "step": 12933
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011742857142857141,
+      "loss": 0.0717,
+      "step": 12934
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011741326530612244,
+      "loss": 0.1473,
+      "step": 12935
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011739795918367345,
+      "loss": 0.0937,
+      "step": 12936
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011738265306122448,
+      "loss": 0.1611,
+      "step": 12937
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.0001173673469387755,
+      "loss": 0.0879,
+      "step": 12938
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011735204081632653,
+      "loss": 0.0914,
+      "step": 12939
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011733673469387754,
+      "loss": 0.0845,
+      "step": 12940
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011732142857142856,
+      "loss": 0.0802,
+      "step": 12941
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011730612244897959,
+      "loss": 0.1392,
+      "step": 12942
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011729081632653061,
+      "loss": 0.0706,
+      "step": 12943
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011727551020408162,
+      "loss": 0.0477,
+      "step": 12944
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011726020408163264,
+      "loss": 0.1141,
+      "step": 12945
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011724489795918367,
+      "loss": 0.054,
+      "step": 12946
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011722959183673469,
+      "loss": 0.066,
+      "step": 12947
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.0001172142857142857,
+      "loss": 0.0902,
+      "step": 12948
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011719897959183672,
+      "loss": 0.1788,
+      "step": 12949
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011718367346938775,
+      "loss": 0.0895,
+      "step": 12950
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011716836734693876,
+      "loss": 0.0557,
+      "step": 12951
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00011715306122448979,
+      "loss": 0.0782,
+      "step": 12952
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.0001171377551020408,
+      "loss": 0.1307,
+      "step": 12953
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00011712244897959183,
+      "loss": 0.1299,
+      "step": 12954
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00011710714285714284,
+      "loss": 0.0651,
+      "step": 12955
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00011709183673469387,
+      "loss": 0.1363,
+      "step": 12956
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00011707653061224489,
+      "loss": 0.0836,
+      "step": 12957
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00011706122448979592,
+      "loss": 0.0796,
+      "step": 12958
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00011704591836734692,
+      "loss": 0.0502,
+      "step": 12959
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00011703061224489795,
+      "loss": 0.2601,
+      "step": 12960
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00011701530612244897,
+      "loss": 0.101,
+      "step": 12961
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.000117,
+      "loss": 0.0458,
+      "step": 12962
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.000116984693877551,
+      "loss": 0.0892,
+      "step": 12963
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011696938775510203,
+      "loss": 0.0446,
+      "step": 12964
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011695408163265305,
+      "loss": 0.084,
+      "step": 12965
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011693877551020408,
+      "loss": 0.0516,
+      "step": 12966
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011692346938775508,
+      "loss": 0.1419,
+      "step": 12967
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011690816326530611,
+      "loss": 0.1085,
+      "step": 12968
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011689285714285713,
+      "loss": 0.0385,
+      "step": 12969
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011687755102040816,
+      "loss": 0.0474,
+      "step": 12970
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011686224489795916,
+      "loss": 0.0488,
+      "step": 12971
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.0001168469387755102,
+      "loss": 0.0783,
+      "step": 12972
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011683163265306121,
+      "loss": 0.1331,
+      "step": 12973
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011681632653061224,
+      "loss": 0.1371,
+      "step": 12974
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011680102040816325,
+      "loss": 0.0923,
+      "step": 12975
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011678571428571428,
+      "loss": 0.0527,
+      "step": 12976
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.0001167704081632653,
+      "loss": 0.0436,
+      "step": 12977
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011675510204081632,
+      "loss": 0.111,
+      "step": 12978
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011673979591836733,
+      "loss": 0.1411,
+      "step": 12979
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011672448979591836,
+      "loss": 0.0622,
+      "step": 12980
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011670918367346937,
+      "loss": 0.059,
+      "step": 12981
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011669387755102039,
+      "loss": 0.1059,
+      "step": 12982
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011667857142857141,
+      "loss": 0.083,
+      "step": 12983
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00011666326530612244,
+      "loss": 0.0364,
+      "step": 12984
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00011664795918367347,
+      "loss": 0.0457,
+      "step": 12985
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00011663265306122447,
+      "loss": 0.1437,
+      "step": 12986
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.0001166173469387755,
+      "loss": 0.1359,
+      "step": 12987
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00011660204081632652,
+      "loss": 0.1697,
+      "step": 12988
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00011658673469387755,
+      "loss": 0.0885,
+      "step": 12989
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00011657142857142856,
+      "loss": 0.0477,
+      "step": 12990
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00011655612244897959,
+      "loss": 0.1747,
+      "step": 12991
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.0001165408163265306,
+      "loss": 0.105,
+      "step": 12992
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00011652551020408163,
+      "loss": 0.0941,
+      "step": 12993
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00011651020408163264,
+      "loss": 0.177,
+      "step": 12994
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00011649489795918367,
+      "loss": 0.0797,
+      "step": 12995
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00011647959183673468,
+      "loss": 0.115,
+      "step": 12996
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00011646428571428571,
+      "loss": 0.1211,
+      "step": 12997
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00011644897959183672,
+      "loss": 0.0641,
+      "step": 12998
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00011643367346938775,
+      "loss": 0.0995,
+      "step": 12999
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00011641836734693877,
+      "loss": 0.0505,
+      "step": 13000
+    },
+    {
+      "epoch": 12.62,
+      "eval_cer": 0.25245369165601234,
+      "eval_loss": 0.9836429953575134,
+      "eval_runtime": 234.9526,
+      "eval_samples_per_second": 37.514,
+      "eval_steps_per_second": 4.69,
+      "eval_wer": 0.6534016632062309,
+      "step": 13000
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.0001164030612244898,
+      "loss": 0.0989,
+      "step": 13001
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.0001163877551020408,
+      "loss": 0.0633,
+      "step": 13002
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00011637244897959183,
+      "loss": 0.0594,
+      "step": 13003
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011635714285714285,
+      "loss": 0.1655,
+      "step": 13004
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011634183673469388,
+      "loss": 0.1106,
+      "step": 13005
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011632653061224488,
+      "loss": 0.1672,
+      "step": 13006
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011631122448979591,
+      "loss": 0.1324,
+      "step": 13007
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011629591836734693,
+      "loss": 0.0945,
+      "step": 13008
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011628061224489796,
+      "loss": 0.038,
+      "step": 13009
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011626530612244896,
+      "loss": 0.0959,
+      "step": 13010
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011624999999999999,
+      "loss": 0.1007,
+      "step": 13011
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011623469387755101,
+      "loss": 0.1333,
+      "step": 13012
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011621938775510203,
+      "loss": 0.0575,
+      "step": 13013
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011620408163265304,
+      "loss": 0.1346,
+      "step": 13014
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011618877551020407,
+      "loss": 0.059,
+      "step": 13015
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011617346938775509,
+      "loss": 0.1013,
+      "step": 13016
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011615816326530611,
+      "loss": 0.1585,
+      "step": 13017
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011614285714285713,
+      "loss": 0.1086,
+      "step": 13018
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011612755102040816,
+      "loss": 0.0948,
+      "step": 13019
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011611224489795917,
+      "loss": 0.075,
+      "step": 13020
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011609693877551019,
+      "loss": 0.0422,
+      "step": 13021
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011608163265306121,
+      "loss": 0.0497,
+      "step": 13022
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011606632653061224,
+      "loss": 0.0654,
+      "step": 13023
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011605102040816324,
+      "loss": 0.1398,
+      "step": 13024
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011603571428571427,
+      "loss": 0.0963,
+      "step": 13025
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011602040816326529,
+      "loss": 0.0521,
+      "step": 13026
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011600510204081632,
+      "loss": 0.0843,
+      "step": 13027
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011598979591836732,
+      "loss": 0.1464,
+      "step": 13028
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011597448979591835,
+      "loss": 0.1504,
+      "step": 13029
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011595918367346938,
+      "loss": 0.1035,
+      "step": 13030
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.0001159438775510204,
+      "loss": 0.0548,
+      "step": 13031
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011592857142857143,
+      "loss": 0.1322,
+      "step": 13032
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011591326530612243,
+      "loss": 0.1032,
+      "step": 13033
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011589795918367347,
+      "loss": 0.0768,
+      "step": 13034
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00011588265306122448,
+      "loss": 0.1173,
+      "step": 13035
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00011586734693877551,
+      "loss": 0.1241,
+      "step": 13036
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00011585204081632652,
+      "loss": 0.1822,
+      "step": 13037
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00011583673469387755,
+      "loss": 0.0886,
+      "step": 13038
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00011582142857142856,
+      "loss": 0.0491,
+      "step": 13039
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.0001158061224489796,
+      "loss": 0.0341,
+      "step": 13040
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.0001157908163265306,
+      "loss": 0.0984,
+      "step": 13041
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00011577551020408163,
+      "loss": 0.1057,
+      "step": 13042
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00011576020408163265,
+      "loss": 0.1039,
+      "step": 13043
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00011574489795918366,
+      "loss": 0.053,
+      "step": 13044
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011572959183673468,
+      "loss": 0.0419,
+      "step": 13045
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011571428571428571,
+      "loss": 0.0488,
+      "step": 13046
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011569897959183673,
+      "loss": 0.1615,
+      "step": 13047
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011568367346938774,
+      "loss": 0.1812,
+      "step": 13048
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011566836734693876,
+      "loss": 0.2175,
+      "step": 13049
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011565306122448979,
+      "loss": 0.0991,
+      "step": 13050
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011563775510204081,
+      "loss": 0.1221,
+      "step": 13051
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011562244897959183,
+      "loss": 0.0661,
+      "step": 13052
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011560714285714284,
+      "loss": 0.0643,
+      "step": 13053
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011559183673469387,
+      "loss": 0.2257,
+      "step": 13054
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011557653061224488,
+      "loss": 0.0581,
+      "step": 13055
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011556122448979591,
+      "loss": 0.1018,
+      "step": 13056
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011554591836734692,
+      "loss": 0.0972,
+      "step": 13057
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011553061224489795,
+      "loss": 0.1595,
+      "step": 13058
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011551530612244896,
+      "loss": 0.047,
+      "step": 13059
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011549999999999999,
+      "loss": 0.1003,
+      "step": 13060
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.000115484693877551,
+      "loss": 0.0617,
+      "step": 13061
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011546938775510204,
+      "loss": 0.0956,
+      "step": 13062
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011545408163265304,
+      "loss": 0.0421,
+      "step": 13063
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011543877551020407,
+      "loss": 0.1568,
+      "step": 13064
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011542346938775509,
+      "loss": 0.0967,
+      "step": 13065
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00011540816326530612,
+      "loss": 0.083,
+      "step": 13066
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00011539285714285712,
+      "loss": 0.1787,
+      "step": 13067
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00011537755102040815,
+      "loss": 0.1854,
+      "step": 13068
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00011536224489795917,
+      "loss": 0.1944,
+      "step": 13069
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.0001153469387755102,
+      "loss": 0.1374,
+      "step": 13070
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.0001153316326530612,
+      "loss": 0.1009,
+      "step": 13071
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00011531632653061223,
+      "loss": 0.1213,
+      "step": 13072
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00011530102040816325,
+      "loss": 0.1864,
+      "step": 13073
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00011528571428571428,
+      "loss": 0.0674,
+      "step": 13074
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.0001152704081632653,
+      "loss": 0.0831,
+      "step": 13075
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011525510204081631,
+      "loss": 0.0916,
+      "step": 13076
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011523979591836735,
+      "loss": 0.0949,
+      "step": 13077
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011522448979591836,
+      "loss": 0.0431,
+      "step": 13078
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011520918367346938,
+      "loss": 0.1503,
+      "step": 13079
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.0001151938775510204,
+      "loss": 0.1484,
+      "step": 13080
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011517857142857143,
+      "loss": 0.0467,
+      "step": 13081
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011516326530612244,
+      "loss": 0.0838,
+      "step": 13082
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011514795918367346,
+      "loss": 0.0801,
+      "step": 13083
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011513265306122448,
+      "loss": 0.0935,
+      "step": 13084
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011511734693877551,
+      "loss": 0.0901,
+      "step": 13085
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011510204081632651,
+      "loss": 0.063,
+      "step": 13086
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011508673469387754,
+      "loss": 0.0981,
+      "step": 13087
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011507142857142856,
+      "loss": 0.0903,
+      "step": 13088
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011505612244897959,
+      "loss": 0.0905,
+      "step": 13089
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011504081632653059,
+      "loss": 0.0794,
+      "step": 13090
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011502551020408162,
+      "loss": 0.0832,
+      "step": 13091
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011501020408163264,
+      "loss": 0.1834,
+      "step": 13092
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011499489795918367,
+      "loss": 0.1332,
+      "step": 13093
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011497959183673467,
+      "loss": 0.1819,
+      "step": 13094
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.0001149642857142857,
+      "loss": 0.091,
+      "step": 13095
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00011494897959183672,
+      "loss": 0.0413,
+      "step": 13096
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011493367346938775,
+      "loss": 0.101,
+      "step": 13097
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011491836734693876,
+      "loss": 0.1754,
+      "step": 13098
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011490306122448979,
+      "loss": 0.057,
+      "step": 13099
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.0001148877551020408,
+      "loss": 0.0959,
+      "step": 13100
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011487244897959183,
+      "loss": 0.0526,
+      "step": 13101
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011485714285714284,
+      "loss": 0.1316,
+      "step": 13102
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011484183673469387,
+      "loss": 0.1415,
+      "step": 13103
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011482653061224489,
+      "loss": 0.1449,
+      "step": 13104
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011481122448979592,
+      "loss": 0.1447,
+      "step": 13105
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011479591836734692,
+      "loss": 0.091,
+      "step": 13106
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011478061224489795,
+      "loss": 0.0706,
+      "step": 13107
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011476530612244897,
+      "loss": 0.1106,
+      "step": 13108
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011475,
+      "loss": 0.0836,
+      "step": 13109
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.000114734693877551,
+      "loss": 0.2325,
+      "step": 13110
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011471938775510203,
+      "loss": 0.0712,
+      "step": 13111
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011470408163265305,
+      "loss": 0.0612,
+      "step": 13112
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011468877551020408,
+      "loss": 0.1719,
+      "step": 13113
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011467346938775508,
+      "loss": 0.0698,
+      "step": 13114
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011465816326530611,
+      "loss": 0.1055,
+      "step": 13115
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011464285714285713,
+      "loss": 0.1148,
+      "step": 13116
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011462755102040816,
+      "loss": 0.1976,
+      "step": 13117
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00011461224489795916,
+      "loss": 0.1313,
+      "step": 13118
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.0001145969387755102,
+      "loss": 0.1095,
+      "step": 13119
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00011458163265306122,
+      "loss": 0.1318,
+      "step": 13120
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00011456632653061223,
+      "loss": 0.0396,
+      "step": 13121
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00011455102040816326,
+      "loss": 0.0669,
+      "step": 13122
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00011453571428571428,
+      "loss": 0.3547,
+      "step": 13123
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.0001145204081632653,
+      "loss": 0.1105,
+      "step": 13124
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00011450510204081631,
+      "loss": 0.0966,
+      "step": 13125
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00011448979591836734,
+      "loss": 0.1513,
+      "step": 13126
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00011447448979591836,
+      "loss": 0.0936,
+      "step": 13127
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011445918367346939,
+      "loss": 0.0505,
+      "step": 13128
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011444387755102039,
+      "loss": 0.1198,
+      "step": 13129
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011442857142857142,
+      "loss": 0.1394,
+      "step": 13130
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011441326530612244,
+      "loss": 0.1021,
+      "step": 13131
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011439795918367347,
+      "loss": 0.1014,
+      "step": 13132
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011438265306122447,
+      "loss": 0.067,
+      "step": 13133
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.0001143673469387755,
+      "loss": 0.0859,
+      "step": 13134
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011435204081632652,
+      "loss": 0.0837,
+      "step": 13135
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011433673469387755,
+      "loss": 0.0686,
+      "step": 13136
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011432142857142855,
+      "loss": 0.0591,
+      "step": 13137
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011430612244897959,
+      "loss": 0.0859,
+      "step": 13138
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.0001142908163265306,
+      "loss": 0.1227,
+      "step": 13139
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011427551020408163,
+      "loss": 0.1016,
+      "step": 13140
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011426020408163264,
+      "loss": 0.1017,
+      "step": 13141
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011424489795918367,
+      "loss": 0.1255,
+      "step": 13142
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011422959183673468,
+      "loss": 0.1854,
+      "step": 13143
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011421428571428571,
+      "loss": 0.0529,
+      "step": 13144
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011419897959183672,
+      "loss": 0.0878,
+      "step": 13145
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011418367346938775,
+      "loss": 0.0731,
+      "step": 13146
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00011416836734693877,
+      "loss": 0.0497,
+      "step": 13147
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.0001141530612244898,
+      "loss": 0.1637,
+      "step": 13148
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.0001141377551020408,
+      "loss": 0.0538,
+      "step": 13149
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011412244897959183,
+      "loss": 0.0855,
+      "step": 13150
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011410714285714285,
+      "loss": 0.054,
+      "step": 13151
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011409183673469386,
+      "loss": 0.0391,
+      "step": 13152
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011407653061224488,
+      "loss": 0.192,
+      "step": 13153
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011406122448979591,
+      "loss": 0.345,
+      "step": 13154
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011404591836734693,
+      "loss": 0.0923,
+      "step": 13155
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011403061224489795,
+      "loss": 0.0942,
+      "step": 13156
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011401530612244896,
+      "loss": 0.0427,
+      "step": 13157
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 0.0883,
+      "step": 13158
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011398469387755101,
+      "loss": 0.0938,
+      "step": 13159
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011396938775510203,
+      "loss": 0.1384,
+      "step": 13160
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011395408163265304,
+      "loss": 0.0563,
+      "step": 13161
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011393877551020407,
+      "loss": 0.0914,
+      "step": 13162
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011392346938775508,
+      "loss": 0.0501,
+      "step": 13163
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011390816326530611,
+      "loss": 0.0678,
+      "step": 13164
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011389285714285713,
+      "loss": 0.0543,
+      "step": 13165
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011387755102040816,
+      "loss": 0.0709,
+      "step": 13166
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011386224489795919,
+      "loss": 0.1311,
+      "step": 13167
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011384693877551019,
+      "loss": 0.0547,
+      "step": 13168
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011383163265306122,
+      "loss": 0.047,
+      "step": 13169
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011381632653061224,
+      "loss": 0.0883,
+      "step": 13170
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011380102040816327,
+      "loss": 0.1011,
+      "step": 13171
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011378571428571427,
+      "loss": 0.0541,
+      "step": 13172
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.0001137704081632653,
+      "loss": 0.0642,
+      "step": 13173
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011375510204081632,
+      "loss": 0.1103,
+      "step": 13174
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011373979591836735,
+      "loss": 0.0519,
+      "step": 13175
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011372448979591835,
+      "loss": 0.0626,
+      "step": 13176
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011370918367346938,
+      "loss": 0.1077,
+      "step": 13177
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.0001136938775510204,
+      "loss": 0.0504,
+      "step": 13178
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011367857142857143,
+      "loss": 0.0956,
+      "step": 13179
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011366326530612243,
+      "loss": 0.0576,
+      "step": 13180
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011364795918367346,
+      "loss": 0.1312,
+      "step": 13181
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011363265306122448,
+      "loss": 0.1365,
+      "step": 13182
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.0001136173469387755,
+      "loss": 0.0558,
+      "step": 13183
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011360204081632652,
+      "loss": 0.045,
+      "step": 13184
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011358673469387755,
+      "loss": 0.2278,
+      "step": 13185
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011357142857142856,
+      "loss": 0.1518,
+      "step": 13186
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011355612244897958,
+      "loss": 0.0931,
+      "step": 13187
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.0001135408163265306,
+      "loss": 0.0509,
+      "step": 13188
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00011352551020408163,
+      "loss": 0.0901,
+      "step": 13189
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011351020408163264,
+      "loss": 0.0869,
+      "step": 13190
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011349489795918366,
+      "loss": 0.0959,
+      "step": 13191
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011347959183673468,
+      "loss": 0.1107,
+      "step": 13192
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011346428571428571,
+      "loss": 0.1715,
+      "step": 13193
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011344897959183671,
+      "loss": 0.0797,
+      "step": 13194
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011343367346938774,
+      "loss": 0.1234,
+      "step": 13195
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011341836734693876,
+      "loss": 0.1192,
+      "step": 13196
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011340306122448979,
+      "loss": 0.0647,
+      "step": 13197
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.0001133877551020408,
+      "loss": 0.1478,
+      "step": 13198
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011337244897959182,
+      "loss": 0.1122,
+      "step": 13199
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00011335714285714284,
+      "loss": 0.0914,
+      "step": 13200
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00011334183673469387,
+      "loss": 0.1023,
+      "step": 13201
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00011332653061224488,
+      "loss": 0.0826,
+      "step": 13202
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.0001133112244897959,
+      "loss": 0.0911,
+      "step": 13203
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00011329591836734692,
+      "loss": 0.2011,
+      "step": 13204
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00011328061224489795,
+      "loss": 0.1073,
+      "step": 13205
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00011326530612244896,
+      "loss": 0.0951,
+      "step": 13206
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00011324999999999999,
+      "loss": 0.1279,
+      "step": 13207
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.000113234693877551,
+      "loss": 0.0502,
+      "step": 13208
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00011321938775510204,
+      "loss": 0.0563,
+      "step": 13209
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011320408163265304,
+      "loss": 0.1934,
+      "step": 13210
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011318877551020407,
+      "loss": 0.1061,
+      "step": 13211
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.0001131734693877551,
+      "loss": 0.0958,
+      "step": 13212
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011315816326530612,
+      "loss": 0.0617,
+      "step": 13213
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011314285714285713,
+      "loss": 0.0774,
+      "step": 13214
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011312755102040815,
+      "loss": 0.0805,
+      "step": 13215
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011311224489795918,
+      "loss": 0.1077,
+      "step": 13216
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.0001130969387755102,
+      "loss": 0.2564,
+      "step": 13217
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011308163265306122,
+      "loss": 0.0898,
+      "step": 13218
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011306632653061223,
+      "loss": 0.0406,
+      "step": 13219
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011305102040816326,
+      "loss": 0.099,
+      "step": 13220
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011303571428571428,
+      "loss": 0.0851,
+      "step": 13221
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.0001130204081632653,
+      "loss": 0.0698,
+      "step": 13222
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011300510204081631,
+      "loss": 0.0843,
+      "step": 13223
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011298979591836734,
+      "loss": 0.0938,
+      "step": 13224
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011297448979591835,
+      "loss": 0.0481,
+      "step": 13225
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011295918367346938,
+      "loss": 0.1298,
+      "step": 13226
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.0001129438775510204,
+      "loss": 0.0977,
+      "step": 13227
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011292857142857143,
+      "loss": 0.0985,
+      "step": 13228
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011291326530612243,
+      "loss": 0.145,
+      "step": 13229
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011289795918367346,
+      "loss": 0.0931,
+      "step": 13230
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011288265306122448,
+      "loss": 0.0509,
+      "step": 13231
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011286734693877551,
+      "loss": 0.1436,
+      "step": 13232
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011285204081632651,
+      "loss": 0.0601,
+      "step": 13233
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011283673469387754,
+      "loss": 0.0849,
+      "step": 13234
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011282142857142856,
+      "loss": 0.0445,
+      "step": 13235
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011280612244897959,
+      "loss": 0.2345,
+      "step": 13236
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011279081632653059,
+      "loss": 0.059,
+      "step": 13237
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011277551020408162,
+      "loss": 0.1368,
+      "step": 13238
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011276020408163264,
+      "loss": 0.0916,
+      "step": 13239
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00011274489795918367,
+      "loss": 0.1069,
+      "step": 13240
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00011272959183673467,
+      "loss": 0.1162,
+      "step": 13241
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.0001127142857142857,
+      "loss": 0.2503,
+      "step": 13242
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00011269897959183672,
+      "loss": 0.062,
+      "step": 13243
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00011268367346938775,
+      "loss": 0.1276,
+      "step": 13244
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00011266836734693876,
+      "loss": 0.0431,
+      "step": 13245
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00011265306122448979,
+      "loss": 0.0489,
+      "step": 13246
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.0001126377551020408,
+      "loss": 0.0798,
+      "step": 13247
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00011262244897959183,
+      "loss": 0.1237,
+      "step": 13248
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00011260714285714284,
+      "loss": 0.1213,
+      "step": 13249
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00011259183673469387,
+      "loss": 0.2014,
+      "step": 13250
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011257653061224488,
+      "loss": 0.0526,
+      "step": 13251
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011256122448979592,
+      "loss": 0.1037,
+      "step": 13252
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011254591836734692,
+      "loss": 0.1143,
+      "step": 13253
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011253061224489795,
+      "loss": 0.1559,
+      "step": 13254
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011251530612244897,
+      "loss": 0.14,
+      "step": 13255
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.0001125,
+      "loss": 0.158,
+      "step": 13256
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011248469387755101,
+      "loss": 0.0427,
+      "step": 13257
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011246938775510203,
+      "loss": 0.1133,
+      "step": 13258
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011245408163265306,
+      "loss": 0.116,
+      "step": 13259
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00011243877551020406,
+      "loss": 0.1338,
+      "step": 13260
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.0001124234693877551,
+      "loss": 0.1291,
+      "step": 13261
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00011240816326530611,
+      "loss": 0.057,
+      "step": 13262
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00011239285714285714,
+      "loss": 0.1488,
+      "step": 13263
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00011237755102040815,
+      "loss": 0.089,
+      "step": 13264
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00011236224489795918,
+      "loss": 0.0592,
+      "step": 13265
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.0001123469387755102,
+      "loss": 0.1142,
+      "step": 13266
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00011233163265306122,
+      "loss": 0.0854,
+      "step": 13267
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00011231632653061223,
+      "loss": 0.1048,
+      "step": 13268
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00011230102040816326,
+      "loss": 0.0918,
+      "step": 13269
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00011228571428571428,
+      "loss": 0.0461,
+      "step": 13270
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.0001122704081632653,
+      "loss": 0.0397,
+      "step": 13271
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011225510204081631,
+      "loss": 0.0719,
+      "step": 13272
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011223979591836734,
+      "loss": 0.1895,
+      "step": 13273
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011222448979591836,
+      "loss": 0.0538,
+      "step": 13274
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011220918367346939,
+      "loss": 0.1344,
+      "step": 13275
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011219387755102039,
+      "loss": 0.0816,
+      "step": 13276
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011217857142857142,
+      "loss": 0.0904,
+      "step": 13277
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011216326530612244,
+      "loss": 0.0853,
+      "step": 13278
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011214795918367347,
+      "loss": 0.1348,
+      "step": 13279
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00011213265306122447,
+      "loss": 0.0648,
+      "step": 13280
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.0001121173469387755,
+      "loss": 0.0578,
+      "step": 13281
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011210204081632652,
+      "loss": 0.0618,
+      "step": 13282
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011208673469387755,
+      "loss": 0.0396,
+      "step": 13283
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011207142857142855,
+      "loss": 0.0838,
+      "step": 13284
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011205612244897958,
+      "loss": 0.1032,
+      "step": 13285
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.0001120408163265306,
+      "loss": 0.1012,
+      "step": 13286
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011202551020408163,
+      "loss": 0.1052,
+      "step": 13287
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011201020408163264,
+      "loss": 0.1369,
+      "step": 13288
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011199489795918367,
+      "loss": 0.0655,
+      "step": 13289
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011197959183673468,
+      "loss": 0.0415,
+      "step": 13290
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.0001119642857142857,
+      "loss": 0.109,
+      "step": 13291
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00011194897959183672,
+      "loss": 0.2169,
+      "step": 13292
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011193367346938775,
+      "loss": 0.0609,
+      "step": 13293
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011191836734693876,
+      "loss": 0.1818,
+      "step": 13294
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011190306122448978,
+      "loss": 0.1196,
+      "step": 13295
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.0001118877551020408,
+      "loss": 0.0463,
+      "step": 13296
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011187244897959183,
+      "loss": 0.111,
+      "step": 13297
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011185714285714285,
+      "loss": 0.0513,
+      "step": 13298
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011184183673469386,
+      "loss": 0.0745,
+      "step": 13299
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011182653061224488,
+      "loss": 0.0658,
+      "step": 13300
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011181122448979591,
+      "loss": 0.0559,
+      "step": 13301
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00011179591836734694,
+      "loss": 0.0824,
+      "step": 13302
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011178061224489794,
+      "loss": 0.0754,
+      "step": 13303
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011176530612244898,
+      "loss": 0.1222,
+      "step": 13304
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011174999999999999,
+      "loss": 0.0624,
+      "step": 13305
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011173469387755102,
+      "loss": 0.1486,
+      "step": 13306
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011171938775510203,
+      "loss": 0.0928,
+      "step": 13307
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011170408163265306,
+      "loss": 0.0435,
+      "step": 13308
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011168877551020407,
+      "loss": 0.0438,
+      "step": 13309
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.0001116734693877551,
+      "loss": 0.0745,
+      "step": 13310
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011165816326530611,
+      "loss": 0.1224,
+      "step": 13311
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00011164285714285714,
+      "loss": 0.1387,
+      "step": 13312
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011162755102040816,
+      "loss": 0.077,
+      "step": 13313
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011161224489795919,
+      "loss": 0.0933,
+      "step": 13314
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011159693877551019,
+      "loss": 0.082,
+      "step": 13315
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011158163265306122,
+      "loss": 0.0443,
+      "step": 13316
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011156632653061224,
+      "loss": 0.1093,
+      "step": 13317
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011155102040816327,
+      "loss": 0.0544,
+      "step": 13318
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011153571428571427,
+      "loss": 0.0857,
+      "step": 13319
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.0001115204081632653,
+      "loss": 0.0876,
+      "step": 13320
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011150510204081632,
+      "loss": 0.0432,
+      "step": 13321
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011148979591836734,
+      "loss": 0.0745,
+      "step": 13322
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00011147448979591835,
+      "loss": 0.1825,
+      "step": 13323
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00011145918367346938,
+      "loss": 0.1411,
+      "step": 13324
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.0001114438775510204,
+      "loss": 0.1519,
+      "step": 13325
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00011142857142857142,
+      "loss": 0.0905,
+      "step": 13326
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00011141326530612243,
+      "loss": 0.0577,
+      "step": 13327
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00011139795918367346,
+      "loss": 0.0509,
+      "step": 13328
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00011138265306122448,
+      "loss": 0.0674,
+      "step": 13329
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.0001113673469387755,
+      "loss": 0.1491,
+      "step": 13330
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00011135204081632652,
+      "loss": 0.0876,
+      "step": 13331
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00011133673469387755,
+      "loss": 0.0397,
+      "step": 13332
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00011132142857142855,
+      "loss": 0.0887,
+      "step": 13333
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011130612244897958,
+      "loss": 0.0365,
+      "step": 13334
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.0001112908163265306,
+      "loss": 0.0831,
+      "step": 13335
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011127551020408163,
+      "loss": 0.0967,
+      "step": 13336
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011126020408163263,
+      "loss": 0.1459,
+      "step": 13337
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011124489795918366,
+      "loss": 0.0988,
+      "step": 13338
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011122959183673468,
+      "loss": 0.0977,
+      "step": 13339
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011121428571428571,
+      "loss": 0.0799,
+      "step": 13340
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011119897959183671,
+      "loss": 0.0589,
+      "step": 13341
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011118367346938774,
+      "loss": 0.0964,
+      "step": 13342
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00011116836734693876,
+      "loss": 0.0775,
+      "step": 13343
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00011115306122448979,
+      "loss": 0.1008,
+      "step": 13344
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.0001111377551020408,
+      "loss": 0.041,
+      "step": 13345
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00011112244897959182,
+      "loss": 0.087,
+      "step": 13346
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00011110714285714286,
+      "loss": 0.0829,
+      "step": 13347
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00011109183673469387,
+      "loss": 0.0541,
+      "step": 13348
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.0001110765306122449,
+      "loss": 0.1493,
+      "step": 13349
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.0001110612244897959,
+      "loss": 0.1294,
+      "step": 13350
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00011104591836734694,
+      "loss": 0.1682,
+      "step": 13351
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00011103061224489795,
+      "loss": 0.0449,
+      "step": 13352
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00011101530612244897,
+      "loss": 0.1115,
+      "step": 13353
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011099999999999999,
+      "loss": 0.0505,
+      "step": 13354
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011098469387755102,
+      "loss": 0.218,
+      "step": 13355
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011096938775510204,
+      "loss": 0.2292,
+      "step": 13356
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011095408163265305,
+      "loss": 0.1292,
+      "step": 13357
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011093877551020407,
+      "loss": 0.0691,
+      "step": 13358
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.0001109234693877551,
+      "loss": 0.1281,
+      "step": 13359
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011090816326530612,
+      "loss": 0.1525,
+      "step": 13360
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011089285714285713,
+      "loss": 0.1512,
+      "step": 13361
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011087755102040815,
+      "loss": 0.1914,
+      "step": 13362
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011086224489795918,
+      "loss": 0.0641,
+      "step": 13363
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00011084693877551018,
+      "loss": 0.0851,
+      "step": 13364
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00011083163265306122,
+      "loss": 0.0889,
+      "step": 13365
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00011081632653061223,
+      "loss": 0.1214,
+      "step": 13366
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00011080102040816326,
+      "loss": 0.283,
+      "step": 13367
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00011078571428571427,
+      "loss": 0.0854,
+      "step": 13368
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.0001107704081632653,
+      "loss": 0.0487,
+      "step": 13369
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00011075510204081631,
+      "loss": 0.0853,
+      "step": 13370
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00011073979591836734,
+      "loss": 0.0896,
+      "step": 13371
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00011072448979591835,
+      "loss": 0.1286,
+      "step": 13372
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00011070918367346938,
+      "loss": 0.245,
+      "step": 13373
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.0001106938775510204,
+      "loss": 0.0615,
+      "step": 13374
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011067857142857143,
+      "loss": 0.0906,
+      "step": 13375
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011066326530612243,
+      "loss": 0.0605,
+      "step": 13376
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011064795918367346,
+      "loss": 0.0631,
+      "step": 13377
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011063265306122448,
+      "loss": 0.0997,
+      "step": 13378
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011061734693877551,
+      "loss": 0.1462,
+      "step": 13379
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011060204081632651,
+      "loss": 0.0569,
+      "step": 13380
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011058673469387754,
+      "loss": 0.0518,
+      "step": 13381
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011057142857142856,
+      "loss": 0.0802,
+      "step": 13382
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011055612244897959,
+      "loss": 0.0826,
+      "step": 13383
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00011054081632653059,
+      "loss": 0.1438,
+      "step": 13384
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011052551020408162,
+      "loss": 0.1154,
+      "step": 13385
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011051020408163264,
+      "loss": 0.1969,
+      "step": 13386
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011049489795918367,
+      "loss": 0.0573,
+      "step": 13387
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011047959183673467,
+      "loss": 0.0907,
+      "step": 13388
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.0001104642857142857,
+      "loss": 0.1231,
+      "step": 13389
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011044897959183672,
+      "loss": 0.0823,
+      "step": 13390
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011043367346938775,
+      "loss": 0.1077,
+      "step": 13391
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011041836734693876,
+      "loss": 0.1335,
+      "step": 13392
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011040306122448979,
+      "loss": 0.0835,
+      "step": 13393
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011038775510204082,
+      "loss": 0.0284,
+      "step": 13394
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00011037244897959182,
+      "loss": 0.0697,
+      "step": 13395
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00011035714285714285,
+      "loss": 0.0849,
+      "step": 13396
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00011034183673469387,
+      "loss": 0.0707,
+      "step": 13397
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.0001103265306122449,
+      "loss": 0.0957,
+      "step": 13398
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.0001103112244897959,
+      "loss": 0.0777,
+      "step": 13399
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00011029591836734693,
+      "loss": 0.0447,
+      "step": 13400
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00011028061224489795,
+      "loss": 0.0444,
+      "step": 13401
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00011026530612244898,
+      "loss": 0.1106,
+      "step": 13402
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00011024999999999998,
+      "loss": 0.2301,
+      "step": 13403
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00011023469387755101,
+      "loss": 0.1312,
+      "step": 13404
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00011021938775510203,
+      "loss": 0.0484,
+      "step": 13405
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00011020408163265306,
+      "loss": 0.0871,
+      "step": 13406
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00011018877551020406,
+      "loss": 0.0419,
+      "step": 13407
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.0001101734693877551,
+      "loss": 0.0448,
+      "step": 13408
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00011015816326530611,
+      "loss": 0.2126,
+      "step": 13409
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00011014285714285714,
+      "loss": 0.1624,
+      "step": 13410
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00011012755102040815,
+      "loss": 0.0829,
+      "step": 13411
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00011011224489795918,
+      "loss": 0.0474,
+      "step": 13412
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.0001100969387755102,
+      "loss": 0.1305,
+      "step": 13413
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00011008163265306122,
+      "loss": 0.0825,
+      "step": 13414
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00011006632653061223,
+      "loss": 0.103,
+      "step": 13415
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00011005102040816326,
+      "loss": 0.0572,
+      "step": 13416
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00011003571428571428,
+      "loss": 0.0975,
+      "step": 13417
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.0001100204081632653,
+      "loss": 0.0447,
+      "step": 13418
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00011000510204081631,
+      "loss": 0.047,
+      "step": 13419
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00010998979591836734,
+      "loss": 0.041,
+      "step": 13420
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00010997448979591836,
+      "loss": 0.1021,
+      "step": 13421
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00010995918367346939,
+      "loss": 0.1005,
+      "step": 13422
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00010994387755102039,
+      "loss": 0.1307,
+      "step": 13423
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00010992857142857142,
+      "loss": 0.0922,
+      "step": 13424
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00010991326530612244,
+      "loss": 0.0608,
+      "step": 13425
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00010989795918367347,
+      "loss": 0.0929,
+      "step": 13426
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010988265306122447,
+      "loss": 0.0686,
+      "step": 13427
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.0001098673469387755,
+      "loss": 0.0976,
+      "step": 13428
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010985204081632652,
+      "loss": 0.1289,
+      "step": 13429
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010983673469387754,
+      "loss": 0.0523,
+      "step": 13430
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010982142857142855,
+      "loss": 0.0931,
+      "step": 13431
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010980612244897958,
+      "loss": 0.3069,
+      "step": 13432
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.0001097908163265306,
+      "loss": 0.0337,
+      "step": 13433
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010977551020408162,
+      "loss": 0.0842,
+      "step": 13434
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010976020408163264,
+      "loss": 0.1683,
+      "step": 13435
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010974489795918367,
+      "loss": 0.0813,
+      "step": 13436
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010972959183673468,
+      "loss": 0.0455,
+      "step": 13437
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.0001097142857142857,
+      "loss": 0.1733,
+      "step": 13438
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010969897959183673,
+      "loss": 0.0622,
+      "step": 13439
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010968367346938775,
+      "loss": 0.1715,
+      "step": 13440
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010966836734693878,
+      "loss": 0.1413,
+      "step": 13441
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010965306122448978,
+      "loss": 0.0399,
+      "step": 13442
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010963775510204081,
+      "loss": 0.1569,
+      "step": 13443
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010962244897959183,
+      "loss": 0.0457,
+      "step": 13444
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010960714285714286,
+      "loss": 0.0793,
+      "step": 13445
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010959183673469386,
+      "loss": 0.0674,
+      "step": 13446
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.0001095765306122449,
+      "loss": 0.13,
+      "step": 13447
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010956122448979591,
+      "loss": 0.1392,
+      "step": 13448
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010954591836734694,
+      "loss": 0.0533,
+      "step": 13449
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010953061224489794,
+      "loss": 0.0314,
+      "step": 13450
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010951530612244897,
+      "loss": 0.0732,
+      "step": 13451
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010949999999999999,
+      "loss": 0.1346,
+      "step": 13452
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010948469387755102,
+      "loss": 0.1569,
+      "step": 13453
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010946938775510203,
+      "loss": 0.0499,
+      "step": 13454
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010945408163265306,
+      "loss": 0.0933,
+      "step": 13455
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00010943877551020407,
+      "loss": 0.0471,
+      "step": 13456
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.0001094234693877551,
+      "loss": 0.0704,
+      "step": 13457
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010940816326530611,
+      "loss": 0.0499,
+      "step": 13458
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010939285714285714,
+      "loss": 0.0908,
+      "step": 13459
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010937755102040815,
+      "loss": 0.0435,
+      "step": 13460
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010936224489795917,
+      "loss": 0.0488,
+      "step": 13461
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010934693877551019,
+      "loss": 0.1384,
+      "step": 13462
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010933163265306122,
+      "loss": 0.0495,
+      "step": 13463
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010931632653061224,
+      "loss": 0.0818,
+      "step": 13464
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010930102040816325,
+      "loss": 0.1337,
+      "step": 13465
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010928571428571427,
+      "loss": 0.0848,
+      "step": 13466
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.0001092704081632653,
+      "loss": 0.121,
+      "step": 13467
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010925510204081632,
+      "loss": 0.0863,
+      "step": 13468
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010923979591836733,
+      "loss": 0.0482,
+      "step": 13469
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010922448979591835,
+      "loss": 0.0642,
+      "step": 13470
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010920918367346938,
+      "loss": 0.0426,
+      "step": 13471
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010919387755102039,
+      "loss": 0.1249,
+      "step": 13472
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010917857142857142,
+      "loss": 0.0483,
+      "step": 13473
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010916326530612243,
+      "loss": 0.1146,
+      "step": 13474
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010914795918367346,
+      "loss": 0.1266,
+      "step": 13475
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00010913265306122447,
+      "loss": 0.0857,
+      "step": 13476
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.0001091173469387755,
+      "loss": 0.149,
+      "step": 13477
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010910204081632651,
+      "loss": 0.05,
+      "step": 13478
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010908673469387755,
+      "loss": 0.0573,
+      "step": 13479
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010907142857142855,
+      "loss": 0.0804,
+      "step": 13480
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010905612244897958,
+      "loss": 0.0878,
+      "step": 13481
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.0001090408163265306,
+      "loss": 0.0638,
+      "step": 13482
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010902551020408163,
+      "loss": 0.0735,
+      "step": 13483
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010901020408163266,
+      "loss": 0.0468,
+      "step": 13484
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010899489795918366,
+      "loss": 0.1925,
+      "step": 13485
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010897959183673469,
+      "loss": 0.0849,
+      "step": 13486
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010896428571428571,
+      "loss": 0.0614,
+      "step": 13487
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010894897959183674,
+      "loss": 0.0654,
+      "step": 13488
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010893367346938774,
+      "loss": 0.0488,
+      "step": 13489
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010891836734693877,
+      "loss": 0.1303,
+      "step": 13490
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010890306122448979,
+      "loss": 0.1367,
+      "step": 13491
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010888775510204081,
+      "loss": 0.0546,
+      "step": 13492
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010887244897959182,
+      "loss": 0.0447,
+      "step": 13493
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010885714285714285,
+      "loss": 0.089,
+      "step": 13494
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010884183673469387,
+      "loss": 0.0563,
+      "step": 13495
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010882653061224489,
+      "loss": 0.0689,
+      "step": 13496
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.0001088112244897959,
+      "loss": 0.2655,
+      "step": 13497
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010879591836734694,
+      "loss": 0.0997,
+      "step": 13498
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010878061224489795,
+      "loss": 0.0843,
+      "step": 13499
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010876530612244897,
+      "loss": 0.0336,
+      "step": 13500
+    },
+    {
+      "epoch": 13.11,
+      "eval_cer": 0.25190015866205157,
+      "eval_loss": 1.0619421005249023,
+      "eval_runtime": 227.4074,
+      "eval_samples_per_second": 38.759,
+      "eval_steps_per_second": 4.846,
+      "eval_wer": 0.6563842292016829,
+      "step": 13500
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010874999999999999,
+      "loss": 0.05,
+      "step": 13501
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010873469387755102,
+      "loss": 0.0599,
+      "step": 13502
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010871938775510202,
+      "loss": 0.2085,
+      "step": 13503
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010870408163265305,
+      "loss": 0.0828,
+      "step": 13504
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010868877551020407,
+      "loss": 0.0447,
+      "step": 13505
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.0001086734693877551,
+      "loss": 0.079,
+      "step": 13506
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.0001086581632653061,
+      "loss": 0.0406,
+      "step": 13507
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010864285714285713,
+      "loss": 0.0941,
+      "step": 13508
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010862755102040815,
+      "loss": 0.0905,
+      "step": 13509
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010861224489795918,
+      "loss": 0.1259,
+      "step": 13510
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010859693877551018,
+      "loss": 0.1236,
+      "step": 13511
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010858163265306121,
+      "loss": 0.0937,
+      "step": 13512
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010856632653061223,
+      "loss": 0.0444,
+      "step": 13513
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010855102040816326,
+      "loss": 0.1174,
+      "step": 13514
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010853571428571427,
+      "loss": 0.1167,
+      "step": 13515
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.0001085204081632653,
+      "loss": 0.1328,
+      "step": 13516
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010850510204081631,
+      "loss": 0.09,
+      "step": 13517
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010848979591836734,
+      "loss": 0.0774,
+      "step": 13518
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010847448979591835,
+      "loss": 0.0846,
+      "step": 13519
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010845918367346938,
+      "loss": 0.087,
+      "step": 13520
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.0001084438775510204,
+      "loss": 0.17,
+      "step": 13521
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010842857142857143,
+      "loss": 0.0533,
+      "step": 13522
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010841326530612243,
+      "loss": 0.089,
+      "step": 13523
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010839795918367346,
+      "loss": 0.0537,
+      "step": 13524
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010838265306122448,
+      "loss": 0.1589,
+      "step": 13525
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010836734693877551,
+      "loss": 0.0359,
+      "step": 13526
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010835204081632651,
+      "loss": 0.0479,
+      "step": 13527
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010833673469387754,
+      "loss": 0.1016,
+      "step": 13528
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00010832142857142857,
+      "loss": 0.0669,
+      "step": 13529
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010830612244897959,
+      "loss": 0.1191,
+      "step": 13530
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.0001082908163265306,
+      "loss": 0.0841,
+      "step": 13531
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010827551020408162,
+      "loss": 0.0992,
+      "step": 13532
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010826020408163265,
+      "loss": 0.0475,
+      "step": 13533
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010824489795918366,
+      "loss": 0.1124,
+      "step": 13534
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010822959183673469,
+      "loss": 0.1151,
+      "step": 13535
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.0001082142857142857,
+      "loss": 0.0933,
+      "step": 13536
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010819897959183673,
+      "loss": 0.1241,
+      "step": 13537
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010818367346938774,
+      "loss": 0.0568,
+      "step": 13538
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010816836734693877,
+      "loss": 0.1163,
+      "step": 13539
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010815306122448979,
+      "loss": 0.15,
+      "step": 13540
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010813775510204082,
+      "loss": 0.1453,
+      "step": 13541
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010812244897959182,
+      "loss": 0.1197,
+      "step": 13542
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010810714285714285,
+      "loss": 0.0869,
+      "step": 13543
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010809183673469387,
+      "loss": 0.0527,
+      "step": 13544
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.0001080765306122449,
+      "loss": 0.0307,
+      "step": 13545
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.0001080612244897959,
+      "loss": 0.0875,
+      "step": 13546
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010804591836734693,
+      "loss": 0.0871,
+      "step": 13547
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010803061224489795,
+      "loss": 0.1034,
+      "step": 13548
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010801530612244898,
+      "loss": 0.0364,
+      "step": 13549
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010799999999999998,
+      "loss": 0.0933,
+      "step": 13550
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010798469387755101,
+      "loss": 0.0778,
+      "step": 13551
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010796938775510203,
+      "loss": 0.1176,
+      "step": 13552
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010795408163265306,
+      "loss": 0.2101,
+      "step": 13553
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010793877551020406,
+      "loss": 0.0578,
+      "step": 13554
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.0001079234693877551,
+      "loss": 0.1712,
+      "step": 13555
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010790816326530611,
+      "loss": 0.0361,
+      "step": 13556
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010789285714285714,
+      "loss": 0.0852,
+      "step": 13557
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010787755102040815,
+      "loss": 0.0337,
+      "step": 13558
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010786224489795918,
+      "loss": 0.1978,
+      "step": 13559
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010784693877551019,
+      "loss": 0.0553,
+      "step": 13560
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010783163265306122,
+      "loss": 0.0527,
+      "step": 13561
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010781632653061223,
+      "loss": 0.0428,
+      "step": 13562
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010780102040816326,
+      "loss": 0.0885,
+      "step": 13563
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010778571428571427,
+      "loss": 0.0414,
+      "step": 13564
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010777040816326529,
+      "loss": 0.1215,
+      "step": 13565
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010775510204081631,
+      "loss": 0.055,
+      "step": 13566
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010773979591836734,
+      "loss": 0.0571,
+      "step": 13567
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010772448979591836,
+      "loss": 0.0546,
+      "step": 13568
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010770918367346937,
+      "loss": 0.049,
+      "step": 13569
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00010769387755102039,
+      "loss": 0.042,
+      "step": 13570
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010767857142857142,
+      "loss": 0.1208,
+      "step": 13571
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010766326530612244,
+      "loss": 0.1627,
+      "step": 13572
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010764795918367345,
+      "loss": 0.1558,
+      "step": 13573
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010763265306122449,
+      "loss": 0.0782,
+      "step": 13574
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.0001076173469387755,
+      "loss": 0.0683,
+      "step": 13575
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010760204081632653,
+      "loss": 0.0891,
+      "step": 13576
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010758673469387754,
+      "loss": 0.0844,
+      "step": 13577
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010757142857142857,
+      "loss": 0.0865,
+      "step": 13578
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010755612244897958,
+      "loss": 0.0913,
+      "step": 13579
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010754081632653061,
+      "loss": 0.0483,
+      "step": 13580
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010752551020408162,
+      "loss": 0.1286,
+      "step": 13581
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010751020408163265,
+      "loss": 0.0385,
+      "step": 13582
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010749489795918367,
+      "loss": 0.1191,
+      "step": 13583
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.0001074795918367347,
+      "loss": 0.2423,
+      "step": 13584
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.0001074642857142857,
+      "loss": 0.061,
+      "step": 13585
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010744897959183673,
+      "loss": 0.1283,
+      "step": 13586
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010743367346938775,
+      "loss": 0.0886,
+      "step": 13587
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010741836734693878,
+      "loss": 0.105,
+      "step": 13588
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010740306122448978,
+      "loss": 0.0716,
+      "step": 13589
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010738775510204081,
+      "loss": 0.1057,
+      "step": 13590
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010737244897959183,
+      "loss": 0.1341,
+      "step": 13591
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010735714285714286,
+      "loss": 0.0506,
+      "step": 13592
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010734183673469386,
+      "loss": 0.0452,
+      "step": 13593
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010732653061224489,
+      "loss": 0.048,
+      "step": 13594
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010731122448979591,
+      "loss": 0.1719,
+      "step": 13595
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010729591836734694,
+      "loss": 0.104,
+      "step": 13596
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010728061224489794,
+      "loss": 0.1311,
+      "step": 13597
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010726530612244897,
+      "loss": 0.0925,
+      "step": 13598
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010724999999999999,
+      "loss": 0.1314,
+      "step": 13599
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010723469387755101,
+      "loss": 0.0853,
+      "step": 13600
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010721938775510203,
+      "loss": 0.1276,
+      "step": 13601
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010720408163265306,
+      "loss": 0.0607,
+      "step": 13602
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010718877551020407,
+      "loss": 0.147,
+      "step": 13603
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010717346938775509,
+      "loss": 0.095,
+      "step": 13604
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010715816326530611,
+      "loss": 0.0416,
+      "step": 13605
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010714285714285714,
+      "loss": 0.0487,
+      "step": 13606
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010712755102040815,
+      "loss": 0.0788,
+      "step": 13607
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010711224489795917,
+      "loss": 0.0596,
+      "step": 13608
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010709693877551019,
+      "loss": 0.0767,
+      "step": 13609
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010708163265306122,
+      "loss": 0.1289,
+      "step": 13610
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010706632653061222,
+      "loss": 0.0447,
+      "step": 13611
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00010705102040816325,
+      "loss": 0.0873,
+      "step": 13612
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00010703571428571427,
+      "loss": 0.1196,
+      "step": 13613
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.0001070204081632653,
+      "loss": 0.0652,
+      "step": 13614
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.0001070051020408163,
+      "loss": 0.1541,
+      "step": 13615
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00010698979591836733,
+      "loss": 0.0598,
+      "step": 13616
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00010697448979591835,
+      "loss": 0.2128,
+      "step": 13617
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00010695918367346938,
+      "loss": 0.0524,
+      "step": 13618
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00010694387755102041,
+      "loss": 0.036,
+      "step": 13619
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00010692857142857142,
+      "loss": 0.0731,
+      "step": 13620
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00010691326530612245,
+      "loss": 0.1221,
+      "step": 13621
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010689795918367346,
+      "loss": 0.1384,
+      "step": 13622
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.0001068826530612245,
+      "loss": 0.1685,
+      "step": 13623
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.0001068673469387755,
+      "loss": 0.1202,
+      "step": 13624
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010685204081632653,
+      "loss": 0.0429,
+      "step": 13625
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010683673469387754,
+      "loss": 0.0464,
+      "step": 13626
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010682142857142858,
+      "loss": 0.1111,
+      "step": 13627
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010680612244897958,
+      "loss": 0.3077,
+      "step": 13628
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010679081632653061,
+      "loss": 0.0464,
+      "step": 13629
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010677551020408163,
+      "loss": 0.0835,
+      "step": 13630
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010676020408163264,
+      "loss": 0.0786,
+      "step": 13631
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010674489795918366,
+      "loss": 0.0435,
+      "step": 13632
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010672959183673469,
+      "loss": 0.0542,
+      "step": 13633
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010671428571428571,
+      "loss": 0.1391,
+      "step": 13634
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010669897959183673,
+      "loss": 0.1832,
+      "step": 13635
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010668367346938774,
+      "loss": 0.0519,
+      "step": 13636
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010666836734693877,
+      "loss": 0.0789,
+      "step": 13637
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010665306122448979,
+      "loss": 0.0615,
+      "step": 13638
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.0001066377551020408,
+      "loss": 0.0719,
+      "step": 13639
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010662244897959182,
+      "loss": 0.1522,
+      "step": 13640
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010660714285714285,
+      "loss": 0.1415,
+      "step": 13641
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010659183673469386,
+      "loss": 0.082,
+      "step": 13642
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010657653061224489,
+      "loss": 0.0435,
+      "step": 13643
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.0001065612244897959,
+      "loss": 0.0977,
+      "step": 13644
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010654591836734694,
+      "loss": 0.0335,
+      "step": 13645
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010653061224489794,
+      "loss": 0.0763,
+      "step": 13646
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010651530612244897,
+      "loss": 0.1678,
+      "step": 13647
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010649999999999999,
+      "loss": 0.2095,
+      "step": 13648
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010648469387755102,
+      "loss": 0.0988,
+      "step": 13649
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010646938775510202,
+      "loss": 0.1346,
+      "step": 13650
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010645408163265305,
+      "loss": 0.0695,
+      "step": 13651
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010643877551020407,
+      "loss": 0.0808,
+      "step": 13652
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.0001064234693877551,
+      "loss": 0.1776,
+      "step": 13653
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.0001064081632653061,
+      "loss": 0.0604,
+      "step": 13654
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00010639285714285713,
+      "loss": 0.0956,
+      "step": 13655
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00010637755102040815,
+      "loss": 0.0471,
+      "step": 13656
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00010636224489795918,
+      "loss": 0.0816,
+      "step": 13657
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00010634693877551018,
+      "loss": 0.0779,
+      "step": 13658
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00010633163265306121,
+      "loss": 0.1295,
+      "step": 13659
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00010631632653061223,
+      "loss": 0.0395,
+      "step": 13660
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00010630102040816326,
+      "loss": 0.0515,
+      "step": 13661
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00010628571428571427,
+      "loss": 0.0308,
+      "step": 13662
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.0001062704081632653,
+      "loss": 0.0462,
+      "step": 13663
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010625510204081631,
+      "loss": 0.0327,
+      "step": 13664
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010623979591836734,
+      "loss": 0.1397,
+      "step": 13665
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010622448979591836,
+      "loss": 0.1807,
+      "step": 13666
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010620918367346938,
+      "loss": 0.1277,
+      "step": 13667
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010619387755102041,
+      "loss": 0.1315,
+      "step": 13668
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010617857142857142,
+      "loss": 0.0371,
+      "step": 13669
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010616326530612244,
+      "loss": 0.0376,
+      "step": 13670
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010614795918367346,
+      "loss": 0.1348,
+      "step": 13671
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010613265306122449,
+      "loss": 0.2305,
+      "step": 13672
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00010611734693877549,
+      "loss": 0.1432,
+      "step": 13673
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010610204081632652,
+      "loss": 0.0794,
+      "step": 13674
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010608673469387754,
+      "loss": 0.1671,
+      "step": 13675
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010607142857142857,
+      "loss": 0.092,
+      "step": 13676
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010605612244897957,
+      "loss": 0.1389,
+      "step": 13677
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.0001060408163265306,
+      "loss": 0.1066,
+      "step": 13678
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010602551020408162,
+      "loss": 0.0372,
+      "step": 13679
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010601020408163265,
+      "loss": 0.1305,
+      "step": 13680
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010599489795918366,
+      "loss": 0.1056,
+      "step": 13681
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010597959183673469,
+      "loss": 0.0432,
+      "step": 13682
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.0001059642857142857,
+      "loss": 0.0498,
+      "step": 13683
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010594897959183673,
+      "loss": 0.1895,
+      "step": 13684
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010593367346938774,
+      "loss": 0.0951,
+      "step": 13685
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010591836734693877,
+      "loss": 0.0785,
+      "step": 13686
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010590306122448978,
+      "loss": 0.0894,
+      "step": 13687
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010588775510204082,
+      "loss": 0.0773,
+      "step": 13688
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010587244897959182,
+      "loss": 0.0737,
+      "step": 13689
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010585714285714285,
+      "loss": 0.2848,
+      "step": 13690
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010584183673469387,
+      "loss": 0.1058,
+      "step": 13691
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.0001058265306122449,
+      "loss": 0.0598,
+      "step": 13692
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.0001058112244897959,
+      "loss": 0.0961,
+      "step": 13693
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010579591836734693,
+      "loss": 0.0335,
+      "step": 13694
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010578061224489795,
+      "loss": 0.0816,
+      "step": 13695
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010576530612244898,
+      "loss": 0.034,
+      "step": 13696
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010574999999999998,
+      "loss": 0.0937,
+      "step": 13697
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010573469387755101,
+      "loss": 0.0469,
+      "step": 13698
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010571938775510203,
+      "loss": 0.0914,
+      "step": 13699
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010570408163265306,
+      "loss": 0.035,
+      "step": 13700
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010568877551020406,
+      "loss": 0.0801,
+      "step": 13701
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.0001056734693877551,
+      "loss": 0.0335,
+      "step": 13702
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010565816326530611,
+      "loss": 0.258,
+      "step": 13703
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010564285714285713,
+      "loss": 0.0592,
+      "step": 13704
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010562755102040814,
+      "loss": 0.0416,
+      "step": 13705
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010561224489795918,
+      "loss": 0.0459,
+      "step": 13706
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010559693877551019,
+      "loss": 0.0793,
+      "step": 13707
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010558163265306121,
+      "loss": 0.0341,
+      "step": 13708
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010556632653061223,
+      "loss": 0.1355,
+      "step": 13709
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010555102040816326,
+      "loss": 0.1347,
+      "step": 13710
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010553571428571429,
+      "loss": 0.0894,
+      "step": 13711
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010552040816326529,
+      "loss": 0.0946,
+      "step": 13712
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010550510204081632,
+      "loss": 0.164,
+      "step": 13713
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00010548979591836734,
+      "loss": 0.039,
+      "step": 13714
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010547448979591837,
+      "loss": 0.085,
+      "step": 13715
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010545918367346937,
+      "loss": 0.2699,
+      "step": 13716
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.0001054438775510204,
+      "loss": 0.0982,
+      "step": 13717
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010542857142857142,
+      "loss": 0.0891,
+      "step": 13718
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010541326530612245,
+      "loss": 0.04,
+      "step": 13719
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010539795918367345,
+      "loss": 0.0397,
+      "step": 13720
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010538265306122448,
+      "loss": 0.0841,
+      "step": 13721
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.0001053673469387755,
+      "loss": 0.1243,
+      "step": 13722
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010535204081632653,
+      "loss": 0.087,
+      "step": 13723
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010533673469387754,
+      "loss": 0.0593,
+      "step": 13724
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010532142857142857,
+      "loss": 0.1306,
+      "step": 13725
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010530612244897958,
+      "loss": 0.1605,
+      "step": 13726
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010529081632653061,
+      "loss": 0.0584,
+      "step": 13727
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010527551020408162,
+      "loss": 0.2214,
+      "step": 13728
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010526020408163265,
+      "loss": 0.1028,
+      "step": 13729
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010524489795918366,
+      "loss": 0.0867,
+      "step": 13730
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.0001052295918367347,
+      "loss": 0.0515,
+      "step": 13731
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.0001052142857142857,
+      "loss": 0.0577,
+      "step": 13732
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010519897959183673,
+      "loss": 0.0826,
+      "step": 13733
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010518367346938775,
+      "loss": 0.174,
+      "step": 13734
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010516836734693876,
+      "loss": 0.0987,
+      "step": 13735
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010515306122448978,
+      "loss": 0.0997,
+      "step": 13736
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010513775510204081,
+      "loss": 0.0398,
+      "step": 13737
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010512244897959183,
+      "loss": 0.0334,
+      "step": 13738
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010510714285714284,
+      "loss": 0.0394,
+      "step": 13739
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010509183673469386,
+      "loss": 0.1678,
+      "step": 13740
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010507653061224489,
+      "loss": 0.1224,
+      "step": 13741
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010506122448979591,
+      "loss": 0.0498,
+      "step": 13742
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010504591836734693,
+      "loss": 0.1209,
+      "step": 13743
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010503061224489794,
+      "loss": 0.12,
+      "step": 13744
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010501530612244897,
+      "loss": 0.0434,
+      "step": 13745
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010499999999999999,
+      "loss": 0.1971,
+      "step": 13746
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010498469387755101,
+      "loss": 0.1796,
+      "step": 13747
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010496938775510202,
+      "loss": 0.2101,
+      "step": 13748
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010495408163265306,
+      "loss": 0.0792,
+      "step": 13749
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010493877551020406,
+      "loss": 0.0774,
+      "step": 13750
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010492346938775509,
+      "loss": 0.0531,
+      "step": 13751
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.0001049081632653061,
+      "loss": 0.0764,
+      "step": 13752
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010489285714285714,
+      "loss": 0.0991,
+      "step": 13753
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010487755102040814,
+      "loss": 0.0956,
+      "step": 13754
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010486224489795917,
+      "loss": 0.1255,
+      "step": 13755
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.0001048469387755102,
+      "loss": 0.0846,
+      "step": 13756
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00010483163265306122,
+      "loss": 0.0842,
+      "step": 13757
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00010481632653061225,
+      "loss": 0.0812,
+      "step": 13758
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00010480102040816325,
+      "loss": 0.2242,
+      "step": 13759
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00010478571428571428,
+      "loss": 0.0903,
+      "step": 13760
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.0001047704081632653,
+      "loss": 0.0766,
+      "step": 13761
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00010475510204081633,
+      "loss": 0.1166,
+      "step": 13762
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00010473979591836733,
+      "loss": 0.0409,
+      "step": 13763
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00010472448979591836,
+      "loss": 0.1071,
+      "step": 13764
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00010470918367346938,
+      "loss": 0.0805,
+      "step": 13765
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010469387755102041,
+      "loss": 0.1189,
+      "step": 13766
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010467857142857142,
+      "loss": 0.1176,
+      "step": 13767
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010466326530612245,
+      "loss": 0.0808,
+      "step": 13768
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010464795918367346,
+      "loss": 0.0832,
+      "step": 13769
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010463265306122448,
+      "loss": 0.0779,
+      "step": 13770
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.0001046173469387755,
+      "loss": 0.1533,
+      "step": 13771
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010460204081632653,
+      "loss": 0.2287,
+      "step": 13772
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010458673469387754,
+      "loss": 0.0438,
+      "step": 13773
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010457142857142856,
+      "loss": 0.1321,
+      "step": 13774
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010455612244897958,
+      "loss": 0.0794,
+      "step": 13775
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010454081632653061,
+      "loss": 0.0354,
+      "step": 13776
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010452551020408163,
+      "loss": 0.157,
+      "step": 13777
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010451020408163264,
+      "loss": 0.1652,
+      "step": 13778
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010449489795918366,
+      "loss": 0.1276,
+      "step": 13779
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010447959183673469,
+      "loss": 0.0897,
+      "step": 13780
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.0001044642857142857,
+      "loss": 0.0815,
+      "step": 13781
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010444897959183672,
+      "loss": 0.0686,
+      "step": 13782
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010443367346938774,
+      "loss": 0.0315,
+      "step": 13783
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010441836734693877,
+      "loss": 0.1936,
+      "step": 13784
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010440306122448978,
+      "loss": 0.0585,
+      "step": 13785
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.0001043877551020408,
+      "loss": 0.1287,
+      "step": 13786
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010437244897959182,
+      "loss": 0.0984,
+      "step": 13787
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010435714285714285,
+      "loss": 0.0899,
+      "step": 13788
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010434183673469386,
+      "loss": 0.0431,
+      "step": 13789
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010432653061224489,
+      "loss": 0.1298,
+      "step": 13790
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.0001043112244897959,
+      "loss": 0.0856,
+      "step": 13791
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010429591836734694,
+      "loss": 0.0689,
+      "step": 13792
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010428061224489794,
+      "loss": 0.0742,
+      "step": 13793
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010426530612244897,
+      "loss": 0.0834,
+      "step": 13794
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010424999999999999,
+      "loss": 0.0788,
+      "step": 13795
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010423469387755102,
+      "loss": 0.1038,
+      "step": 13796
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010421938775510202,
+      "loss": 0.1133,
+      "step": 13797
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010420408163265305,
+      "loss": 0.057,
+      "step": 13798
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010418877551020407,
+      "loss": 0.0855,
+      "step": 13799
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.0001041734693877551,
+      "loss": 0.1298,
+      "step": 13800
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010415816326530612,
+      "loss": 0.118,
+      "step": 13801
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010414285714285713,
+      "loss": 0.1223,
+      "step": 13802
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010412755102040816,
+      "loss": 0.1622,
+      "step": 13803
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010411224489795918,
+      "loss": 0.1081,
+      "step": 13804
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.0001040969387755102,
+      "loss": 0.1187,
+      "step": 13805
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010408163265306121,
+      "loss": 0.1876,
+      "step": 13806
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00010406632653061224,
+      "loss": 0.0979,
+      "step": 13807
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010405102040816326,
+      "loss": 0.0387,
+      "step": 13808
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010403571428571428,
+      "loss": 0.1693,
+      "step": 13809
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.0001040204081632653,
+      "loss": 0.1985,
+      "step": 13810
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010400510204081633,
+      "loss": 0.0868,
+      "step": 13811
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010398979591836733,
+      "loss": 0.1705,
+      "step": 13812
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010397448979591836,
+      "loss": 0.0548,
+      "step": 13813
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010395918367346938,
+      "loss": 0.0407,
+      "step": 13814
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010394387755102041,
+      "loss": 0.0902,
+      "step": 13815
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010392857142857141,
+      "loss": 0.1846,
+      "step": 13816
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010391326530612244,
+      "loss": 0.1593,
+      "step": 13817
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010389795918367346,
+      "loss": 0.0782,
+      "step": 13818
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010388265306122449,
+      "loss": 0.085,
+      "step": 13819
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010386734693877549,
+      "loss": 0.1648,
+      "step": 13820
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010385204081632652,
+      "loss": 0.0447,
+      "step": 13821
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010383673469387754,
+      "loss": 0.1856,
+      "step": 13822
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010382142857142857,
+      "loss": 0.1069,
+      "step": 13823
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010380612244897957,
+      "loss": 0.0857,
+      "step": 13824
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.0001037908163265306,
+      "loss": 0.1008,
+      "step": 13825
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010377551020408162,
+      "loss": 0.0851,
+      "step": 13826
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010376020408163265,
+      "loss": 0.0382,
+      "step": 13827
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010374489795918366,
+      "loss": 0.0978,
+      "step": 13828
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010372959183673469,
+      "loss": 0.0503,
+      "step": 13829
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.0001037142857142857,
+      "loss": 0.1054,
+      "step": 13830
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010369897959183673,
+      "loss": 0.1206,
+      "step": 13831
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010368367346938774,
+      "loss": 0.1207,
+      "step": 13832
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010366836734693877,
+      "loss": 0.1801,
+      "step": 13833
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010365306122448978,
+      "loss": 0.0556,
+      "step": 13834
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010363775510204081,
+      "loss": 0.1459,
+      "step": 13835
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010362244897959182,
+      "loss": 0.1307,
+      "step": 13836
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010360714285714285,
+      "loss": 0.0951,
+      "step": 13837
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010359183673469387,
+      "loss": 0.0629,
+      "step": 13838
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.0001035765306122449,
+      "loss": 0.0585,
+      "step": 13839
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.0001035612244897959,
+      "loss": 0.0659,
+      "step": 13840
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010354591836734693,
+      "loss": 0.1922,
+      "step": 13841
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010353061224489795,
+      "loss": 0.1687,
+      "step": 13842
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010351530612244896,
+      "loss": 0.0894,
+      "step": 13843
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010349999999999998,
+      "loss": 0.0505,
+      "step": 13844
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010348469387755101,
+      "loss": 0.0502,
+      "step": 13845
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010346938775510204,
+      "loss": 0.0783,
+      "step": 13846
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010345408163265305,
+      "loss": 0.0484,
+      "step": 13847
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010343877551020408,
+      "loss": 0.0507,
+      "step": 13848
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.0001034234693877551,
+      "loss": 0.042,
+      "step": 13849
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00010340816326530612,
+      "loss": 0.0567,
+      "step": 13850
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00010339285714285713,
+      "loss": 0.0942,
+      "step": 13851
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00010337755102040816,
+      "loss": 0.1236,
+      "step": 13852
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00010336224489795918,
+      "loss": 0.15,
+      "step": 13853
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.0001033469387755102,
+      "loss": 0.0998,
+      "step": 13854
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00010333163265306121,
+      "loss": 0.0825,
+      "step": 13855
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00010331632653061224,
+      "loss": 0.1647,
+      "step": 13856
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00010330102040816326,
+      "loss": 0.1075,
+      "step": 13857
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00010328571428571429,
+      "loss": 0.0658,
+      "step": 13858
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010327040816326529,
+      "loss": 0.1003,
+      "step": 13859
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010325510204081632,
+      "loss": 0.0586,
+      "step": 13860
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010323979591836734,
+      "loss": 0.0525,
+      "step": 13861
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010322448979591837,
+      "loss": 0.048,
+      "step": 13862
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010320918367346937,
+      "loss": 0.1219,
+      "step": 13863
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.0001031938775510204,
+      "loss": 0.0304,
+      "step": 13864
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010317857142857142,
+      "loss": 0.116,
+      "step": 13865
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010316326530612245,
+      "loss": 0.1054,
+      "step": 13866
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010314795918367345,
+      "loss": 0.0475,
+      "step": 13867
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010313265306122448,
+      "loss": 0.1315,
+      "step": 13868
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.0001031173469387755,
+      "loss": 0.0873,
+      "step": 13869
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00010310204081632653,
+      "loss": 0.0488,
+      "step": 13870
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00010308673469387754,
+      "loss": 0.0653,
+      "step": 13871
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00010307142857142857,
+      "loss": 0.1063,
+      "step": 13872
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00010305612244897958,
+      "loss": 0.0563,
+      "step": 13873
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.0001030408163265306,
+      "loss": 0.0744,
+      "step": 13874
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00010302551020408162,
+      "loss": 0.1299,
+      "step": 13875
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00010301020408163265,
+      "loss": 0.0554,
+      "step": 13876
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00010299489795918366,
+      "loss": 0.0497,
+      "step": 13877
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00010297959183673468,
+      "loss": 0.0746,
+      "step": 13878
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.0001029642857142857,
+      "loss": 0.077,
+      "step": 13879
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010294897959183673,
+      "loss": 0.206,
+      "step": 13880
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010293367346938775,
+      "loss": 0.0443,
+      "step": 13881
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010291836734693876,
+      "loss": 0.1168,
+      "step": 13882
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010290306122448978,
+      "loss": 0.0337,
+      "step": 13883
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010288775510204081,
+      "loss": 0.1483,
+      "step": 13884
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010287244897959183,
+      "loss": 0.0526,
+      "step": 13885
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010285714285714284,
+      "loss": 0.059,
+      "step": 13886
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010284183673469386,
+      "loss": 0.0817,
+      "step": 13887
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010282653061224489,
+      "loss": 0.0974,
+      "step": 13888
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.0001028112244897959,
+      "loss": 0.1098,
+      "step": 13889
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010279591836734693,
+      "loss": 0.0935,
+      "step": 13890
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010278061224489794,
+      "loss": 0.0724,
+      "step": 13891
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010276530612244897,
+      "loss": 0.2114,
+      "step": 13892
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010275,
+      "loss": 0.1784,
+      "step": 13893
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010273469387755101,
+      "loss": 0.05,
+      "step": 13894
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010271938775510204,
+      "loss": 0.0508,
+      "step": 13895
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010270408163265305,
+      "loss": 0.0567,
+      "step": 13896
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010268877551020409,
+      "loss": 0.1176,
+      "step": 13897
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010267346938775509,
+      "loss": 0.1228,
+      "step": 13898
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010265816326530612,
+      "loss": 0.1836,
+      "step": 13899
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010264285714285714,
+      "loss": 0.2424,
+      "step": 13900
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010262755102040817,
+      "loss": 0.1536,
+      "step": 13901
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010261224489795917,
+      "loss": 0.1156,
+      "step": 13902
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.0001025969387755102,
+      "loss": 0.1353,
+      "step": 13903
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010258163265306122,
+      "loss": 0.0617,
+      "step": 13904
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010256632653061225,
+      "loss": 0.1095,
+      "step": 13905
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010255102040816325,
+      "loss": 0.0796,
+      "step": 13906
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010253571428571428,
+      "loss": 0.1341,
+      "step": 13907
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.0001025204081632653,
+      "loss": 0.0434,
+      "step": 13908
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010250510204081632,
+      "loss": 0.144,
+      "step": 13909
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00010248979591836733,
+      "loss": 0.0542,
+      "step": 13910
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00010247448979591836,
+      "loss": 0.0809,
+      "step": 13911
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00010245918367346938,
+      "loss": 0.0992,
+      "step": 13912
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.0001024438775510204,
+      "loss": 0.2064,
+      "step": 13913
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00010242857142857141,
+      "loss": 0.084,
+      "step": 13914
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00010241326530612245,
+      "loss": 0.0827,
+      "step": 13915
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00010239795918367346,
+      "loss": 0.1095,
+      "step": 13916
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00010238265306122448,
+      "loss": 0.1174,
+      "step": 13917
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.0001023673469387755,
+      "loss": 0.1673,
+      "step": 13918
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00010235204081632653,
+      "loss": 0.0469,
+      "step": 13919
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00010233673469387753,
+      "loss": 0.0545,
+      "step": 13920
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010232142857142856,
+      "loss": 0.0769,
+      "step": 13921
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010230612244897958,
+      "loss": 0.1952,
+      "step": 13922
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010229081632653061,
+      "loss": 0.1748,
+      "step": 13923
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010227551020408161,
+      "loss": 0.0405,
+      "step": 13924
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010226020408163264,
+      "loss": 0.1952,
+      "step": 13925
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010224489795918366,
+      "loss": 0.0594,
+      "step": 13926
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010222959183673469,
+      "loss": 0.1135,
+      "step": 13927
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.0001022142857142857,
+      "loss": 0.2714,
+      "step": 13928
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010219897959183672,
+      "loss": 0.1004,
+      "step": 13929
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00010218367346938774,
+      "loss": 0.0563,
+      "step": 13930
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010216836734693877,
+      "loss": 0.0802,
+      "step": 13931
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010215306122448978,
+      "loss": 0.1271,
+      "step": 13932
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.0001021377551020408,
+      "loss": 0.0872,
+      "step": 13933
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010212244897959182,
+      "loss": 0.1522,
+      "step": 13934
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010210714285714285,
+      "loss": 0.1353,
+      "step": 13935
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010209183673469386,
+      "loss": 0.0627,
+      "step": 13936
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010207653061224489,
+      "loss": 0.095,
+      "step": 13937
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010206122448979592,
+      "loss": 0.0462,
+      "step": 13938
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010204591836734693,
+      "loss": 0.0536,
+      "step": 13939
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010203061224489795,
+      "loss": 0.1372,
+      "step": 13940
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00010201530612244897,
+      "loss": 0.1683,
+      "step": 13941
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.000102,
+      "loss": 0.0487,
+      "step": 13942
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00010198469387755102,
+      "loss": 0.0934,
+      "step": 13943
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00010196938775510203,
+      "loss": 0.0925,
+      "step": 13944
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00010195408163265305,
+      "loss": 0.0742,
+      "step": 13945
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00010193877551020408,
+      "loss": 0.0824,
+      "step": 13946
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.0001019234693877551,
+      "loss": 0.1913,
+      "step": 13947
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00010190816326530611,
+      "loss": 0.216,
+      "step": 13948
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00010189285714285713,
+      "loss": 0.0831,
+      "step": 13949
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00010187755102040816,
+      "loss": 0.0952,
+      "step": 13950
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00010186224489795917,
+      "loss": 0.0933,
+      "step": 13951
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.0001018469387755102,
+      "loss": 0.1114,
+      "step": 13952
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00010183163265306121,
+      "loss": 0.0916,
+      "step": 13953
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00010181632653061224,
+      "loss": 0.1514,
+      "step": 13954
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00010180102040816325,
+      "loss": 0.0345,
+      "step": 13955
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00010178571428571428,
+      "loss": 0.1019,
+      "step": 13956
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.0001017704081632653,
+      "loss": 0.0623,
+      "step": 13957
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00010175510204081633,
+      "loss": 0.0587,
+      "step": 13958
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00010173979591836733,
+      "loss": 0.0417,
+      "step": 13959
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00010172448979591836,
+      "loss": 0.1413,
+      "step": 13960
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00010170918367346938,
+      "loss": 0.0875,
+      "step": 13961
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010169387755102041,
+      "loss": 0.0956,
+      "step": 13962
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010167857142857141,
+      "loss": 0.1288,
+      "step": 13963
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010166326530612244,
+      "loss": 0.0438,
+      "step": 13964
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010164795918367346,
+      "loss": 0.1125,
+      "step": 13965
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010163265306122449,
+      "loss": 0.0449,
+      "step": 13966
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010161734693877549,
+      "loss": 0.1103,
+      "step": 13967
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010160204081632652,
+      "loss": 0.0503,
+      "step": 13968
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010158673469387754,
+      "loss": 0.0512,
+      "step": 13969
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010157142857142857,
+      "loss": 0.1233,
+      "step": 13970
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00010155612244897957,
+      "loss": 0.1273,
+      "step": 13971
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.0001015408163265306,
+      "loss": 0.0616,
+      "step": 13972
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00010152551020408162,
+      "loss": 0.1208,
+      "step": 13973
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00010151020408163265,
+      "loss": 0.0506,
+      "step": 13974
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00010149489795918365,
+      "loss": 0.0349,
+      "step": 13975
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00010147959183673469,
+      "loss": 0.1541,
+      "step": 13976
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.0001014642857142857,
+      "loss": 0.1012,
+      "step": 13977
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00010144897959183673,
+      "loss": 0.2685,
+      "step": 13978
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00010143367346938774,
+      "loss": 0.1104,
+      "step": 13979
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00010141836734693877,
+      "loss": 0.0542,
+      "step": 13980
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00010140306122448978,
+      "loss": 0.0421,
+      "step": 13981
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.0001013877551020408,
+      "loss": 0.0397,
+      "step": 13982
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010137244897959183,
+      "loss": 0.0749,
+      "step": 13983
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010135714285714285,
+      "loss": 0.1473,
+      "step": 13984
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010134183673469388,
+      "loss": 0.089,
+      "step": 13985
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010132653061224488,
+      "loss": 0.1351,
+      "step": 13986
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010131122448979591,
+      "loss": 0.0893,
+      "step": 13987
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010129591836734693,
+      "loss": 0.0754,
+      "step": 13988
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010128061224489796,
+      "loss": 0.0412,
+      "step": 13989
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010126530612244896,
+      "loss": 0.102,
+      "step": 13990
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010125,
+      "loss": 0.0594,
+      "step": 13991
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00010123469387755101,
+      "loss": 0.134,
+      "step": 13992
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010121938775510204,
+      "loss": 0.0498,
+      "step": 13993
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010120408163265305,
+      "loss": 0.0475,
+      "step": 13994
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010118877551020408,
+      "loss": 0.0432,
+      "step": 13995
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010117346938775509,
+      "loss": 0.0536,
+      "step": 13996
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010115816326530612,
+      "loss": 0.1619,
+      "step": 13997
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010114285714285713,
+      "loss": 0.054,
+      "step": 13998
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010112755102040816,
+      "loss": 0.0515,
+      "step": 13999
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010111224489795917,
+      "loss": 0.0435,
+      "step": 14000
+    },
+    {
+      "epoch": 13.59,
+      "eval_cer": 0.25434841072496284,
+      "eval_loss": 1.0843515396118164,
+      "eval_runtime": 226.4616,
+      "eval_samples_per_second": 38.92,
+      "eval_steps_per_second": 4.866,
+      "eval_wer": 0.648046227026552,
+      "step": 14000
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.0001010969387755102,
+      "loss": 0.0875,
+      "step": 14001
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00010108163265306121,
+      "loss": 0.044,
+      "step": 14002
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010106632653061224,
+      "loss": 0.2015,
+      "step": 14003
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010105102040816326,
+      "loss": 0.0927,
+      "step": 14004
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010103571428571429,
+      "loss": 0.0706,
+      "step": 14005
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010102040816326529,
+      "loss": 0.0898,
+      "step": 14006
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010100510204081632,
+      "loss": 0.0319,
+      "step": 14007
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010098979591836734,
+      "loss": 0.0552,
+      "step": 14008
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010097448979591837,
+      "loss": 0.1476,
+      "step": 14009
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010095918367346937,
+      "loss": 0.1722,
+      "step": 14010
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.0001009438775510204,
+      "loss": 0.05,
+      "step": 14011
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010092857142857142,
+      "loss": 0.1579,
+      "step": 14012
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00010091326530612244,
+      "loss": 0.1192,
+      "step": 14013
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00010089795918367345,
+      "loss": 0.0499,
+      "step": 14014
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00010088265306122448,
+      "loss": 0.2236,
+      "step": 14015
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.0001008673469387755,
+      "loss": 0.1566,
+      "step": 14016
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00010085204081632652,
+      "loss": 0.0842,
+      "step": 14017
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00010083673469387753,
+      "loss": 0.0721,
+      "step": 14018
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00010082142857142857,
+      "loss": 0.0825,
+      "step": 14019
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00010080612244897958,
+      "loss": 0.0948,
+      "step": 14020
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.0001007908163265306,
+      "loss": 0.1277,
+      "step": 14021
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00010077551020408162,
+      "loss": 0.0637,
+      "step": 14022
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00010076020408163265,
+      "loss": 0.0925,
+      "step": 14023
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010074489795918365,
+      "loss": 0.1423,
+      "step": 14024
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010072959183673468,
+      "loss": 0.0925,
+      "step": 14025
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.0001007142857142857,
+      "loss": 0.0619,
+      "step": 14026
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010069897959183673,
+      "loss": 0.0629,
+      "step": 14027
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010068367346938776,
+      "loss": 0.0469,
+      "step": 14028
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010066836734693876,
+      "loss": 0.0952,
+      "step": 14029
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010065306122448979,
+      "loss": 0.0427,
+      "step": 14030
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010063775510204081,
+      "loss": 0.0806,
+      "step": 14031
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010062244897959184,
+      "loss": 0.0475,
+      "step": 14032
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00010060714285714284,
+      "loss": 0.0493,
+      "step": 14033
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010059183673469387,
+      "loss": 0.1933,
+      "step": 14034
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010057653061224489,
+      "loss": 0.1442,
+      "step": 14035
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010056122448979592,
+      "loss": 0.0903,
+      "step": 14036
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010054591836734693,
+      "loss": 0.0568,
+      "step": 14037
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010053061224489796,
+      "loss": 0.0872,
+      "step": 14038
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010051530612244897,
+      "loss": 0.0524,
+      "step": 14039
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.0001005,
+      "loss": 0.1987,
+      "step": 14040
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010048469387755101,
+      "loss": 0.0543,
+      "step": 14041
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010046938775510204,
+      "loss": 0.0391,
+      "step": 14042
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010045408163265305,
+      "loss": 0.0558,
+      "step": 14043
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00010043877551020407,
+      "loss": 0.1262,
+      "step": 14044
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010042346938775509,
+      "loss": 0.07,
+      "step": 14045
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010040816326530612,
+      "loss": 0.0642,
+      "step": 14046
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010039285714285714,
+      "loss": 0.0538,
+      "step": 14047
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010037755102040815,
+      "loss": 0.1293,
+      "step": 14048
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010036224489795917,
+      "loss": 0.0498,
+      "step": 14049
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.0001003469387755102,
+      "loss": 0.0574,
+      "step": 14050
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010033163265306122,
+      "loss": 0.085,
+      "step": 14051
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010031632653061223,
+      "loss": 0.0797,
+      "step": 14052
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010030102040816325,
+      "loss": 0.0913,
+      "step": 14053
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00010028571428571428,
+      "loss": 0.1228,
+      "step": 14054
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.0001002704081632653,
+      "loss": 0.1699,
+      "step": 14055
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00010025510204081632,
+      "loss": 0.122,
+      "step": 14056
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00010023979591836733,
+      "loss": 0.0727,
+      "step": 14057
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00010022448979591836,
+      "loss": 0.0432,
+      "step": 14058
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00010020918367346937,
+      "loss": 0.1372,
+      "step": 14059
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.0001001938775510204,
+      "loss": 0.0983,
+      "step": 14060
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00010017857142857141,
+      "loss": 0.155,
+      "step": 14061
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00010016326530612245,
+      "loss": 0.078,
+      "step": 14062
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00010014795918367345,
+      "loss": 0.0751,
+      "step": 14063
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00010013265306122448,
+      "loss": 0.0796,
+      "step": 14064
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.0001001173469387755,
+      "loss": 0.1114,
+      "step": 14065
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.00010010204081632653,
+      "loss": 0.0522,
+      "step": 14066
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.00010008673469387753,
+      "loss": 0.0512,
+      "step": 14067
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.00010007142857142856,
+      "loss": 0.1259,
+      "step": 14068
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.00010005612244897958,
+      "loss": 0.0826,
+      "step": 14069
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.00010004081632653061,
+      "loss": 0.0474,
+      "step": 14070
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.00010002551020408161,
+      "loss": 0.1028,
+      "step": 14071
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.00010001020408163264,
+      "loss": 0.3284,
+      "step": 14072
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 9.999489795918367e-05,
+      "loss": 0.0876,
+      "step": 14073
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 9.997959183673469e-05,
+      "loss": 0.0638,
+      "step": 14074
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.996428571428572e-05,
+      "loss": 0.0365,
+      "step": 14075
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.994897959183672e-05,
+      "loss": 0.0525,
+      "step": 14076
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.993367346938775e-05,
+      "loss": 0.1458,
+      "step": 14077
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.991836734693877e-05,
+      "loss": 0.1604,
+      "step": 14078
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.990306122448979e-05,
+      "loss": 0.1001,
+      "step": 14079
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.98877551020408e-05,
+      "loss": 0.0468,
+      "step": 14080
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.987244897959184e-05,
+      "loss": 0.0496,
+      "step": 14081
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.985714285714285e-05,
+      "loss": 0.0753,
+      "step": 14082
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.984183673469387e-05,
+      "loss": 0.0537,
+      "step": 14083
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.982653061224489e-05,
+      "loss": 0.2231,
+      "step": 14084
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.981122448979592e-05,
+      "loss": 0.0982,
+      "step": 14085
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.979591836734693e-05,
+      "loss": 0.1148,
+      "step": 14086
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.978061224489795e-05,
+      "loss": 0.0542,
+      "step": 14087
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.976530612244897e-05,
+      "loss": 0.0539,
+      "step": 14088
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.975e-05,
+      "loss": 0.03,
+      "step": 14089
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.9734693877551e-05,
+      "loss": 0.1129,
+      "step": 14090
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.971938775510203e-05,
+      "loss": 0.0937,
+      "step": 14091
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.970408163265305e-05,
+      "loss": 0.0474,
+      "step": 14092
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.968877551020408e-05,
+      "loss": 0.0779,
+      "step": 14093
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.967346938775508e-05,
+      "loss": 0.0459,
+      "step": 14094
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.965816326530611e-05,
+      "loss": 0.09,
+      "step": 14095
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.964285714285713e-05,
+      "loss": 0.0417,
+      "step": 14096
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.962755102040816e-05,
+      "loss": 0.2044,
+      "step": 14097
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.961224489795917e-05,
+      "loss": 0.0861,
+      "step": 14098
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.95969387755102e-05,
+      "loss": 0.0397,
+      "step": 14099
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.958163265306121e-05,
+      "loss": 0.0827,
+      "step": 14100
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.956632653061224e-05,
+      "loss": 0.0792,
+      "step": 14101
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.955102040816325e-05,
+      "loss": 0.0706,
+      "step": 14102
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.953571428571428e-05,
+      "loss": 0.3763,
+      "step": 14103
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.95204081632653e-05,
+      "loss": 0.0868,
+      "step": 14104
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.950510204081632e-05,
+      "loss": 0.0832,
+      "step": 14105
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.948979591836733e-05,
+      "loss": 0.0454,
+      "step": 14106
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.947448979591836e-05,
+      "loss": 0.0615,
+      "step": 14107
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.945918367346938e-05,
+      "loss": 0.0975,
+      "step": 14108
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.94438775510204e-05,
+      "loss": 0.1133,
+      "step": 14109
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.942857142857141e-05,
+      "loss": 0.0571,
+      "step": 14110
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.941326530612244e-05,
+      "loss": 0.0443,
+      "step": 14111
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.939795918367346e-05,
+      "loss": 0.0443,
+      "step": 14112
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.938265306122449e-05,
+      "loss": 0.0446,
+      "step": 14113
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.936734693877549e-05,
+      "loss": 0.1552,
+      "step": 14114
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.935204081632652e-05,
+      "loss": 0.1774,
+      "step": 14115
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.933673469387754e-05,
+      "loss": 0.1002,
+      "step": 14116
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.932142857142857e-05,
+      "loss": 0.0591,
+      "step": 14117
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.930612244897957e-05,
+      "loss": 0.1773,
+      "step": 14118
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.92908163265306e-05,
+      "loss": 0.2199,
+      "step": 14119
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.927551020408163e-05,
+      "loss": 0.1002,
+      "step": 14120
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.926020408163264e-05,
+      "loss": 0.0547,
+      "step": 14121
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.924489795918367e-05,
+      "loss": 0.1917,
+      "step": 14122
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.922959183673468e-05,
+      "loss": 0.0471,
+      "step": 14123
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.921428571428572e-05,
+      "loss": 0.1293,
+      "step": 14124
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.919897959183672e-05,
+      "loss": 0.132,
+      "step": 14125
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.918367346938775e-05,
+      "loss": 0.0428,
+      "step": 14126
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.916836734693877e-05,
+      "loss": 0.0534,
+      "step": 14127
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.91530612244898e-05,
+      "loss": 0.1767,
+      "step": 14128
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.91377551020408e-05,
+      "loss": 0.1245,
+      "step": 14129
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.912244897959183e-05,
+      "loss": 0.1114,
+      "step": 14130
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.910714285714285e-05,
+      "loss": 0.0814,
+      "step": 14131
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.909183673469388e-05,
+      "loss": 0.0839,
+      "step": 14132
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.907653061224488e-05,
+      "loss": 0.108,
+      "step": 14133
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.906122448979591e-05,
+      "loss": 0.1383,
+      "step": 14134
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.904591836734693e-05,
+      "loss": 0.0981,
+      "step": 14135
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.903061224489796e-05,
+      "loss": 0.1346,
+      "step": 14136
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.901530612244896e-05,
+      "loss": 0.0446,
+      "step": 14137
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.9e-05,
+      "loss": 0.0997,
+      "step": 14138
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.898469387755101e-05,
+      "loss": 0.0726,
+      "step": 14139
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.896938775510204e-05,
+      "loss": 0.1069,
+      "step": 14140
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.895408163265305e-05,
+      "loss": 0.1815,
+      "step": 14141
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.893877551020408e-05,
+      "loss": 0.0962,
+      "step": 14142
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.892346938775509e-05,
+      "loss": 0.0479,
+      "step": 14143
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.890816326530612e-05,
+      "loss": 0.0477,
+      "step": 14144
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.889285714285713e-05,
+      "loss": 0.0663,
+      "step": 14145
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.887755102040816e-05,
+      "loss": 0.0414,
+      "step": 14146
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.886224489795917e-05,
+      "loss": 0.1004,
+      "step": 14147
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.88469387755102e-05,
+      "loss": 0.0446,
+      "step": 14148
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.883163265306121e-05,
+      "loss": 0.0976,
+      "step": 14149
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.881632653061224e-05,
+      "loss": 0.0494,
+      "step": 14150
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.880102040816326e-05,
+      "loss": 0.0861,
+      "step": 14151
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.878571428571427e-05,
+      "loss": 0.0494,
+      "step": 14152
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.877040816326529e-05,
+      "loss": 0.1658,
+      "step": 14153
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.875510204081632e-05,
+      "loss": 0.0957,
+      "step": 14154
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.873979591836734e-05,
+      "loss": 0.1664,
+      "step": 14155
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.872448979591835e-05,
+      "loss": 0.0921,
+      "step": 14156
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.870918367346937e-05,
+      "loss": 0.1234,
+      "step": 14157
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.86938775510204e-05,
+      "loss": 0.0766,
+      "step": 14158
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.867857142857142e-05,
+      "loss": 0.1589,
+      "step": 14159
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.866326530612244e-05,
+      "loss": 0.2066,
+      "step": 14160
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.864795918367345e-05,
+      "loss": 0.0856,
+      "step": 14161
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.863265306122448e-05,
+      "loss": 0.1551,
+      "step": 14162
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.861734693877549e-05,
+      "loss": 0.1663,
+      "step": 14163
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.860204081632652e-05,
+      "loss": 0.1327,
+      "step": 14164
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.858673469387755e-05,
+      "loss": 0.1029,
+      "step": 14165
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.857142857142856e-05,
+      "loss": 0.2248,
+      "step": 14166
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.85561224489796e-05,
+      "loss": 0.0934,
+      "step": 14167
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.85408163265306e-05,
+      "loss": 0.1329,
+      "step": 14168
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.852551020408163e-05,
+      "loss": 0.0922,
+      "step": 14169
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.851020408163265e-05,
+      "loss": 0.0661,
+      "step": 14170
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.849489795918368e-05,
+      "loss": 0.0922,
+      "step": 14171
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.847959183673468e-05,
+      "loss": 0.2348,
+      "step": 14172
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.846428571428571e-05,
+      "loss": 0.0485,
+      "step": 14173
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.844897959183673e-05,
+      "loss": 0.0879,
+      "step": 14174
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.843367346938776e-05,
+      "loss": 0.0854,
+      "step": 14175
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.841836734693876e-05,
+      "loss": 0.0601,
+      "step": 14176
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.840306122448979e-05,
+      "loss": 0.0771,
+      "step": 14177
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.838775510204081e-05,
+      "loss": 0.186,
+      "step": 14178
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.837244897959184e-05,
+      "loss": 0.0582,
+      "step": 14179
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.835714285714284e-05,
+      "loss": 0.0751,
+      "step": 14180
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.834183673469387e-05,
+      "loss": 0.0445,
+      "step": 14181
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.832653061224489e-05,
+      "loss": 0.0466,
+      "step": 14182
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.831122448979591e-05,
+      "loss": 0.0951,
+      "step": 14183
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.829591836734692e-05,
+      "loss": 0.151,
+      "step": 14184
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.828061224489796e-05,
+      "loss": 0.1342,
+      "step": 14185
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.826530612244897e-05,
+      "loss": 0.0952,
+      "step": 14186
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.824999999999999e-05,
+      "loss": 0.0408,
+      "step": 14187
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.8234693877551e-05,
+      "loss": 0.1275,
+      "step": 14188
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.821938775510204e-05,
+      "loss": 0.0523,
+      "step": 14189
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.820408163265305e-05,
+      "loss": 0.0794,
+      "step": 14190
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.818877551020407e-05,
+      "loss": 0.1028,
+      "step": 14191
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.817346938775509e-05,
+      "loss": 0.0911,
+      "step": 14192
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.815816326530612e-05,
+      "loss": 0.1787,
+      "step": 14193
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.814285714285714e-05,
+      "loss": 0.0401,
+      "step": 14194
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.812755102040815e-05,
+      "loss": 0.0836,
+      "step": 14195
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.811224489795917e-05,
+      "loss": 0.0676,
+      "step": 14196
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.80969387755102e-05,
+      "loss": 0.1522,
+      "step": 14197
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.80816326530612e-05,
+      "loss": 0.0501,
+      "step": 14198
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.806632653061223e-05,
+      "loss": 0.0582,
+      "step": 14199
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.805102040816325e-05,
+      "loss": 0.1163,
+      "step": 14200
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.803571428571428e-05,
+      "loss": 0.0351,
+      "step": 14201
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.802040816326528e-05,
+      "loss": 0.1201,
+      "step": 14202
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.800510204081632e-05,
+      "loss": 0.0573,
+      "step": 14203
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.798979591836733e-05,
+      "loss": 0.0874,
+      "step": 14204
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.797448979591836e-05,
+      "loss": 0.1259,
+      "step": 14205
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.795918367346937e-05,
+      "loss": 0.0941,
+      "step": 14206
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.79438775510204e-05,
+      "loss": 0.1063,
+      "step": 14207
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.792857142857141e-05,
+      "loss": 0.0649,
+      "step": 14208
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.791326530612244e-05,
+      "loss": 0.1917,
+      "step": 14209
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.789795918367348e-05,
+      "loss": 0.1134,
+      "step": 14210
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.788265306122448e-05,
+      "loss": 0.0527,
+      "step": 14211
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.786734693877551e-05,
+      "loss": 0.0405,
+      "step": 14212
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.785204081632653e-05,
+      "loss": 0.0564,
+      "step": 14213
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.783673469387754e-05,
+      "loss": 0.0435,
+      "step": 14214
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.782142857142856e-05,
+      "loss": 0.184,
+      "step": 14215
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.780612244897959e-05,
+      "loss": 0.0984,
+      "step": 14216
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.779081632653061e-05,
+      "loss": 0.1367,
+      "step": 14217
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.777551020408162e-05,
+      "loss": 0.0421,
+      "step": 14218
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.776020408163264e-05,
+      "loss": 0.0398,
+      "step": 14219
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.774489795918367e-05,
+      "loss": 0.1196,
+      "step": 14220
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.772959183673469e-05,
+      "loss": 0.1765,
+      "step": 14221
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.77142857142857e-05,
+      "loss": 0.1749,
+      "step": 14222
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.769897959183672e-05,
+      "loss": 0.0605,
+      "step": 14223
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.768367346938775e-05,
+      "loss": 0.0824,
+      "step": 14224
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.766836734693877e-05,
+      "loss": 0.0744,
+      "step": 14225
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.765306122448979e-05,
+      "loss": 0.0806,
+      "step": 14226
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.76377551020408e-05,
+      "loss": 0.0681,
+      "step": 14227
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.762244897959184e-05,
+      "loss": 0.1513,
+      "step": 14228
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.760714285714284e-05,
+      "loss": 0.0994,
+      "step": 14229
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.759183673469387e-05,
+      "loss": 0.0783,
+      "step": 14230
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.757653061224489e-05,
+      "loss": 0.0419,
+      "step": 14231
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.756122448979592e-05,
+      "loss": 0.1907,
+      "step": 14232
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.754591836734692e-05,
+      "loss": 0.0508,
+      "step": 14233
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.753061224489795e-05,
+      "loss": 0.2328,
+      "step": 14234
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.751530612244897e-05,
+      "loss": 0.1392,
+      "step": 14235
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.75e-05,
+      "loss": 0.0926,
+      "step": 14236
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.7484693877551e-05,
+      "loss": 0.0382,
+      "step": 14237
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.746938775510203e-05,
+      "loss": 0.0909,
+      "step": 14238
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 9.745408163265305e-05,
+      "loss": 0.0654,
+      "step": 14239
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.743877551020408e-05,
+      "loss": 0.0417,
+      "step": 14240
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.742346938775508e-05,
+      "loss": 0.0846,
+      "step": 14241
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.740816326530611e-05,
+      "loss": 0.0563,
+      "step": 14242
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.739285714285713e-05,
+      "loss": 0.0315,
+      "step": 14243
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.737755102040816e-05,
+      "loss": 0.052,
+      "step": 14244
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.736224489795916e-05,
+      "loss": 0.0779,
+      "step": 14245
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.73469387755102e-05,
+      "loss": 0.1019,
+      "step": 14246
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.733163265306121e-05,
+      "loss": 0.1274,
+      "step": 14247
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.731632653061224e-05,
+      "loss": 0.1064,
+      "step": 14248
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.730102040816325e-05,
+      "loss": 0.0502,
+      "step": 14249
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.728571428571428e-05,
+      "loss": 0.1425,
+      "step": 14250
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.72704081632653e-05,
+      "loss": 0.055,
+      "step": 14251
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.725510204081632e-05,
+      "loss": 0.0414,
+      "step": 14252
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.723979591836733e-05,
+      "loss": 0.1906,
+      "step": 14253
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.722448979591836e-05,
+      "loss": 0.1509,
+      "step": 14254
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.720918367346939e-05,
+      "loss": 0.0585,
+      "step": 14255
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.71938775510204e-05,
+      "loss": 0.128,
+      "step": 14256
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.717857142857142e-05,
+      "loss": 0.1513,
+      "step": 14257
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.716326530612244e-05,
+      "loss": 0.083,
+      "step": 14258
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.714795918367347e-05,
+      "loss": 0.2265,
+      "step": 14259
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.713265306122447e-05,
+      "loss": 0.0488,
+      "step": 14260
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.71173469387755e-05,
+      "loss": 0.0877,
+      "step": 14261
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.710204081632652e-05,
+      "loss": 0.0348,
+      "step": 14262
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.708673469387755e-05,
+      "loss": 0.0835,
+      "step": 14263
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.707142857142856e-05,
+      "loss": 0.0519,
+      "step": 14264
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.705612244897959e-05,
+      "loss": 0.1473,
+      "step": 14265
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.70408163265306e-05,
+      "loss": 0.2175,
+      "step": 14266
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.702551020408163e-05,
+      "loss": 0.09,
+      "step": 14267
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.702551020408163e-05,
+      "loss": 0.0697,
+      "step": 14268
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.701020408163264e-05,
+      "loss": 0.0908,
+      "step": 14269
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.699489795918367e-05,
+      "loss": 0.032,
+      "step": 14270
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.697959183673468e-05,
+      "loss": 0.1171,
+      "step": 14271
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.696428571428571e-05,
+      "loss": 0.1257,
+      "step": 14272
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.694897959183672e-05,
+      "loss": 0.1417,
+      "step": 14273
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.693367346938775e-05,
+      "loss": 0.1002,
+      "step": 14274
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.691836734693877e-05,
+      "loss": 0.0525,
+      "step": 14275
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.69030612244898e-05,
+      "loss": 0.0526,
+      "step": 14276
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.68877551020408e-05,
+      "loss": 0.1152,
+      "step": 14277
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.687244897959183e-05,
+      "loss": 0.1515,
+      "step": 14278
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.685714285714285e-05,
+      "loss": 0.1422,
+      "step": 14279
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.684183673469388e-05,
+      "loss": 0.0951,
+      "step": 14280
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.682653061224488e-05,
+      "loss": 0.0876,
+      "step": 14281
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.681122448979591e-05,
+      "loss": 0.0489,
+      "step": 14282
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.679591836734693e-05,
+      "loss": 0.05,
+      "step": 14283
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.678061224489796e-05,
+      "loss": 0.0953,
+      "step": 14284
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.676530612244896e-05,
+      "loss": 0.1019,
+      "step": 14285
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.675e-05,
+      "loss": 0.0565,
+      "step": 14286
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.673469387755101e-05,
+      "loss": 0.0874,
+      "step": 14287
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.671938775510204e-05,
+      "loss": 0.1028,
+      "step": 14288
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.670408163265304e-05,
+      "loss": 0.0323,
+      "step": 14289
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.668877551020408e-05,
+      "loss": 0.048,
+      "step": 14290
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.667346938775509e-05,
+      "loss": 0.1358,
+      "step": 14291
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.665816326530611e-05,
+      "loss": 0.0493,
+      "step": 14292
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.664285714285713e-05,
+      "loss": 0.1374,
+      "step": 14293
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.662755102040816e-05,
+      "loss": 0.1234,
+      "step": 14294
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.661224489795917e-05,
+      "loss": 0.0356,
+      "step": 14295
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.659693877551019e-05,
+      "loss": 0.1434,
+      "step": 14296
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.658163265306121e-05,
+      "loss": 0.0541,
+      "step": 14297
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.656632653061224e-05,
+      "loss": 0.0887,
+      "step": 14298
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.655102040816326e-05,
+      "loss": 0.0447,
+      "step": 14299
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.653571428571427e-05,
+      "loss": 0.0414,
+      "step": 14300
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.65204081632653e-05,
+      "loss": 0.08,
+      "step": 14301
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.650510204081632e-05,
+      "loss": 0.0772,
+      "step": 14302
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.648979591836735e-05,
+      "loss": 0.2757,
+      "step": 14303
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.647448979591835e-05,
+      "loss": 0.0504,
+      "step": 14304
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.645918367346938e-05,
+      "loss": 0.0785,
+      "step": 14305
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.64438775510204e-05,
+      "loss": 0.1589,
+      "step": 14306
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.642857142857143e-05,
+      "loss": 0.0975,
+      "step": 14307
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.641326530612244e-05,
+      "loss": 0.1206,
+      "step": 14308
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.639795918367347e-05,
+      "loss": 0.2363,
+      "step": 14309
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.638265306122448e-05,
+      "loss": 0.0609,
+      "step": 14310
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.636734693877551e-05,
+      "loss": 0.0962,
+      "step": 14311
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.635204081632652e-05,
+      "loss": 0.0761,
+      "step": 14312
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.633673469387755e-05,
+      "loss": 0.0806,
+      "step": 14313
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.632142857142856e-05,
+      "loss": 0.032,
+      "step": 14314
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.63061224489796e-05,
+      "loss": 0.058,
+      "step": 14315
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.62908163265306e-05,
+      "loss": 0.1802,
+      "step": 14316
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.627551020408163e-05,
+      "loss": 0.1036,
+      "step": 14317
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.626020408163265e-05,
+      "loss": 0.0377,
+      "step": 14318
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.624489795918368e-05,
+      "loss": 0.1242,
+      "step": 14319
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.622959183673468e-05,
+      "loss": 0.0807,
+      "step": 14320
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.621428571428571e-05,
+      "loss": 0.0799,
+      "step": 14321
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.619897959183673e-05,
+      "loss": 0.1324,
+      "step": 14322
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.618367346938774e-05,
+      "loss": 0.1503,
+      "step": 14323
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.616836734693876e-05,
+      "loss": 0.1305,
+      "step": 14324
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.615306122448979e-05,
+      "loss": 0.0475,
+      "step": 14325
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.613775510204081e-05,
+      "loss": 0.1067,
+      "step": 14326
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.612244897959183e-05,
+      "loss": 0.0966,
+      "step": 14327
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.610714285714284e-05,
+      "loss": 0.1329,
+      "step": 14328
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.609183673469387e-05,
+      "loss": 0.1268,
+      "step": 14329
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.607653061224489e-05,
+      "loss": 0.0557,
+      "step": 14330
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.606122448979591e-05,
+      "loss": 0.0337,
+      "step": 14331
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 9.604591836734692e-05,
+      "loss": 0.131,
+      "step": 14332
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.603061224489795e-05,
+      "loss": 0.1041,
+      "step": 14333
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.601530612244896e-05,
+      "loss": 0.0757,
+      "step": 14334
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.599999999999999e-05,
+      "loss": 0.0576,
+      "step": 14335
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.5984693877551e-05,
+      "loss": 0.0772,
+      "step": 14336
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.596938775510204e-05,
+      "loss": 0.0862,
+      "step": 14337
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.595408163265304e-05,
+      "loss": 0.0476,
+      "step": 14338
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.593877551020407e-05,
+      "loss": 0.0409,
+      "step": 14339
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.592346938775509e-05,
+      "loss": 0.0783,
+      "step": 14340
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.590816326530612e-05,
+      "loss": 0.2556,
+      "step": 14341
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.589285714285712e-05,
+      "loss": 0.0919,
+      "step": 14342
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.587755102040815e-05,
+      "loss": 0.0347,
+      "step": 14343
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.586224489795917e-05,
+      "loss": 0.0854,
+      "step": 14344
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.58469387755102e-05,
+      "loss": 0.0494,
+      "step": 14345
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.583163265306123e-05,
+      "loss": 0.0331,
+      "step": 14346
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.581632653061223e-05,
+      "loss": 0.2202,
+      "step": 14347
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.580102040816326e-05,
+      "loss": 0.1043,
+      "step": 14348
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.578571428571428e-05,
+      "loss": 0.0891,
+      "step": 14349
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.577040816326531e-05,
+      "loss": 0.0739,
+      "step": 14350
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.575510204081631e-05,
+      "loss": 0.0724,
+      "step": 14351
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.573979591836735e-05,
+      "loss": 0.0679,
+      "step": 14352
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.572448979591836e-05,
+      "loss": 0.3967,
+      "step": 14353
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.570918367346938e-05,
+      "loss": 0.0985,
+      "step": 14354
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.56938775510204e-05,
+      "loss": 0.0918,
+      "step": 14355
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.567857142857143e-05,
+      "loss": 0.1487,
+      "step": 14356
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.566326530612244e-05,
+      "loss": 0.0493,
+      "step": 14357
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.564795918367346e-05,
+      "loss": 0.0348,
+      "step": 14358
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.563265306122448e-05,
+      "loss": 0.0786,
+      "step": 14359
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.561734693877551e-05,
+      "loss": 0.0607,
+      "step": 14360
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.560204081632653e-05,
+      "loss": 0.0595,
+      "step": 14361
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.558673469387754e-05,
+      "loss": 0.0703,
+      "step": 14362
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.557142857142856e-05,
+      "loss": 0.0368,
+      "step": 14363
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.555612244897959e-05,
+      "loss": 0.0669,
+      "step": 14364
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.554081632653061e-05,
+      "loss": 0.1641,
+      "step": 14365
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.552551020408162e-05,
+      "loss": 0.0423,
+      "step": 14366
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.551020408163264e-05,
+      "loss": 0.059,
+      "step": 14367
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.549489795918367e-05,
+      "loss": 0.0913,
+      "step": 14368
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.547959183673468e-05,
+      "loss": 0.0444,
+      "step": 14369
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.54642857142857e-05,
+      "loss": 0.0997,
+      "step": 14370
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.544897959183672e-05,
+      "loss": 0.106,
+      "step": 14371
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.543367346938775e-05,
+      "loss": 0.1673,
+      "step": 14372
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.541836734693876e-05,
+      "loss": 0.1806,
+      "step": 14373
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.540306122448979e-05,
+      "loss": 0.0847,
+      "step": 14374
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.53877551020408e-05,
+      "loss": 0.0887,
+      "step": 14375
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.537244897959183e-05,
+      "loss": 0.1246,
+      "step": 14376
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.535714285714284e-05,
+      "loss": 0.0402,
+      "step": 14377
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.534183673469387e-05,
+      "loss": 0.1967,
+      "step": 14378
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.532653061224489e-05,
+      "loss": 0.1021,
+      "step": 14379
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.531122448979592e-05,
+      "loss": 0.0931,
+      "step": 14380
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.529591836734692e-05,
+      "loss": 0.0546,
+      "step": 14381
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.528061224489795e-05,
+      "loss": 0.0467,
+      "step": 14382
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 9.526530612244897e-05,
+      "loss": 0.0484,
+      "step": 14383
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.525e-05,
+      "loss": 0.0589,
+      "step": 14384
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.5234693877551e-05,
+      "loss": 0.0925,
+      "step": 14385
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.521938775510203e-05,
+      "loss": 0.0391,
+      "step": 14386
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.520408163265305e-05,
+      "loss": 0.0896,
+      "step": 14387
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.518877551020408e-05,
+      "loss": 0.112,
+      "step": 14388
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.517346938775508e-05,
+      "loss": 0.0856,
+      "step": 14389
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.515816326530611e-05,
+      "loss": 0.2904,
+      "step": 14390
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.514285714285713e-05,
+      "loss": 0.1377,
+      "step": 14391
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.512755102040816e-05,
+      "loss": 0.1294,
+      "step": 14392
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.511224489795918e-05,
+      "loss": 0.1197,
+      "step": 14393
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.50969387755102e-05,
+      "loss": 0.0382,
+      "step": 14394
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.508163265306123e-05,
+      "loss": 0.0623,
+      "step": 14395
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.506632653061224e-05,
+      "loss": 0.0316,
+      "step": 14396
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.505102040816326e-05,
+      "loss": 0.1816,
+      "step": 14397
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.503571428571428e-05,
+      "loss": 0.0929,
+      "step": 14398
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.502040816326531e-05,
+      "loss": 0.0443,
+      "step": 14399
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.500510204081631e-05,
+      "loss": 0.1239,
+      "step": 14400
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.498979591836734e-05,
+      "loss": 0.0655,
+      "step": 14401
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.497448979591836e-05,
+      "loss": 0.1095,
+      "step": 14402
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.495918367346939e-05,
+      "loss": 0.1787,
+      "step": 14403
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.494387755102039e-05,
+      "loss": 0.0547,
+      "step": 14404
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.492857142857142e-05,
+      "loss": 0.1269,
+      "step": 14405
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.491326530612244e-05,
+      "loss": 0.1038,
+      "step": 14406
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.489795918367347e-05,
+      "loss": 0.1752,
+      "step": 14407
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.488265306122447e-05,
+      "loss": 0.0659,
+      "step": 14408
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.48673469387755e-05,
+      "loss": 0.0694,
+      "step": 14409
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.485204081632652e-05,
+      "loss": 0.0822,
+      "step": 14410
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.483673469387755e-05,
+      "loss": 0.0444,
+      "step": 14411
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.482142857142855e-05,
+      "loss": 0.043,
+      "step": 14412
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.480612244897959e-05,
+      "loss": 0.0966,
+      "step": 14413
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.47908163265306e-05,
+      "loss": 0.0963,
+      "step": 14414
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.477551020408163e-05,
+      "loss": 0.0804,
+      "step": 14415
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.476020408163264e-05,
+      "loss": 0.0467,
+      "step": 14416
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.474489795918367e-05,
+      "loss": 0.0585,
+      "step": 14417
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.472959183673468e-05,
+      "loss": 0.0358,
+      "step": 14418
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.471428571428571e-05,
+      "loss": 0.0766,
+      "step": 14419
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.469897959183672e-05,
+      "loss": 0.2133,
+      "step": 14420
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.468367346938775e-05,
+      "loss": 0.1933,
+      "step": 14421
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.466836734693877e-05,
+      "loss": 0.0821,
+      "step": 14422
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.46530612244898e-05,
+      "loss": 0.0747,
+      "step": 14423
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.46377551020408e-05,
+      "loss": 0.0463,
+      "step": 14424
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.462244897959183e-05,
+      "loss": 0.0372,
+      "step": 14425
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.460714285714285e-05,
+      "loss": 0.032,
+      "step": 14426
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.459183673469388e-05,
+      "loss": 0.0956,
+      "step": 14427
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.457653061224488e-05,
+      "loss": 0.1678,
+      "step": 14428
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.456122448979591e-05,
+      "loss": 0.0374,
+      "step": 14429
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.454591836734693e-05,
+      "loss": 0.1085,
+      "step": 14430
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.453061224489795e-05,
+      "loss": 0.0829,
+      "step": 14431
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.451530612244896e-05,
+      "loss": 0.0431,
+      "step": 14432
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.449999999999999e-05,
+      "loss": 0.2584,
+      "step": 14433
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.448469387755101e-05,
+      "loss": 0.0774,
+      "step": 14434
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.446938775510203e-05,
+      "loss": 0.1143,
+      "step": 14435
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.445408163265304e-05,
+      "loss": 0.0298,
+      "step": 14436
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.443877551020407e-05,
+      "loss": 0.0766,
+      "step": 14437
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.44234693877551e-05,
+      "loss": 0.057,
+      "step": 14438
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.440816326530611e-05,
+      "loss": 0.0551,
+      "step": 14439
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.439285714285714e-05,
+      "loss": 0.1646,
+      "step": 14440
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.437755102040816e-05,
+      "loss": 0.0348,
+      "step": 14441
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.436224489795919e-05,
+      "loss": 0.0419,
+      "step": 14442
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.434693877551019e-05,
+      "loss": 0.074,
+      "step": 14443
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.433163265306122e-05,
+      "loss": 0.0511,
+      "step": 14444
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.431632653061224e-05,
+      "loss": 0.135,
+      "step": 14445
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.430102040816327e-05,
+      "loss": 0.1315,
+      "step": 14446
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.428571428571427e-05,
+      "loss": 0.0399,
+      "step": 14447
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.42704081632653e-05,
+      "loss": 0.0403,
+      "step": 14448
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.425510204081632e-05,
+      "loss": 0.0742,
+      "step": 14449
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.423979591836735e-05,
+      "loss": 0.029,
+      "step": 14450
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.422448979591835e-05,
+      "loss": 0.076,
+      "step": 14451
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.420918367346938e-05,
+      "loss": 0.2412,
+      "step": 14452
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.41938775510204e-05,
+      "loss": 0.1181,
+      "step": 14453
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.417857142857143e-05,
+      "loss": 0.0565,
+      "step": 14454
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.416326530612243e-05,
+      "loss": 0.1258,
+      "step": 14455
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.414795918367347e-05,
+      "loss": 0.0448,
+      "step": 14456
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.413265306122448e-05,
+      "loss": 0.1112,
+      "step": 14457
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.411734693877551e-05,
+      "loss": 0.1603,
+      "step": 14458
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.410204081632652e-05,
+      "loss": 0.1339,
+      "step": 14459
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.408673469387755e-05,
+      "loss": 0.0471,
+      "step": 14460
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.407142857142856e-05,
+      "loss": 0.0745,
+      "step": 14461
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.405612244897958e-05,
+      "loss": 0.1564,
+      "step": 14462
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.40408163265306e-05,
+      "loss": 0.0735,
+      "step": 14463
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.402551020408163e-05,
+      "loss": 0.1475,
+      "step": 14464
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.401020408163265e-05,
+      "loss": 0.0762,
+      "step": 14465
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.399489795918366e-05,
+      "loss": 0.0755,
+      "step": 14466
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.397959183673468e-05,
+      "loss": 0.0388,
+      "step": 14467
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.396428571428571e-05,
+      "loss": 0.1537,
+      "step": 14468
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.394897959183673e-05,
+      "loss": 0.0849,
+      "step": 14469
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.393367346938774e-05,
+      "loss": 0.1402,
+      "step": 14470
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.391836734693876e-05,
+      "loss": 0.0863,
+      "step": 14471
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.390306122448979e-05,
+      "loss": 0.124,
+      "step": 14472
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.38877551020408e-05,
+      "loss": 0.0801,
+      "step": 14473
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.387244897959183e-05,
+      "loss": 0.0409,
+      "step": 14474
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.385714285714284e-05,
+      "loss": 0.0699,
+      "step": 14475
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.384183673469387e-05,
+      "loss": 0.0463,
+      "step": 14476
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.382653061224488e-05,
+      "loss": 0.2002,
+      "step": 14477
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.381122448979591e-05,
+      "loss": 0.0777,
+      "step": 14478
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.379591836734692e-05,
+      "loss": 0.031,
+      "step": 14479
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.378061224489795e-05,
+      "loss": 0.0476,
+      "step": 14480
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.376530612244896e-05,
+      "loss": 0.1257,
+      "step": 14481
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.374999999999999e-05,
+      "loss": 0.0596,
+      "step": 14482
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.373469387755102e-05,
+      "loss": 0.1222,
+      "step": 14483
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.371938775510204e-05,
+      "loss": 0.0455,
+      "step": 14484
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.370408163265307e-05,
+      "loss": 0.0829,
+      "step": 14485
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.368877551020407e-05,
+      "loss": 0.1216,
+      "step": 14486
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.36734693877551e-05,
+      "loss": 0.0717,
+      "step": 14487
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.365816326530612e-05,
+      "loss": 0.1296,
+      "step": 14488
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.364285714285715e-05,
+      "loss": 0.1403,
+      "step": 14489
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.362755102040815e-05,
+      "loss": 0.0436,
+      "step": 14490
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.361224489795918e-05,
+      "loss": 0.1168,
+      "step": 14491
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.35969387755102e-05,
+      "loss": 0.1643,
+      "step": 14492
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.358163265306122e-05,
+      "loss": 0.1206,
+      "step": 14493
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.356632653061223e-05,
+      "loss": 0.0503,
+      "step": 14494
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.355102040816326e-05,
+      "loss": 0.1922,
+      "step": 14495
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.353571428571428e-05,
+      "loss": 0.0865,
+      "step": 14496
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.35204081632653e-05,
+      "loss": 0.0844,
+      "step": 14497
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.350510204081631e-05,
+      "loss": 0.0428,
+      "step": 14498
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.348979591836735e-05,
+      "loss": 0.0748,
+      "step": 14499
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.347448979591836e-05,
+      "loss": 0.0216,
+      "step": 14500
+    },
+    {
+      "epoch": 14.08,
+      "eval_cer": 0.2520524672688804,
+      "eval_loss": 1.1083904504776,
+      "eval_runtime": 227.4265,
+      "eval_samples_per_second": 38.755,
+      "eval_steps_per_second": 4.846,
+      "eval_wer": 0.6511770974085181,
+      "step": 14500
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.345918367346938e-05,
+      "loss": 0.1364,
+      "step": 14501
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.34438775510204e-05,
+      "loss": 0.0872,
+      "step": 14502
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.342857142857143e-05,
+      "loss": 0.0456,
+      "step": 14503
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.341326530612243e-05,
+      "loss": 0.042,
+      "step": 14504
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.339795918367346e-05,
+      "loss": 0.0464,
+      "step": 14505
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.338265306122448e-05,
+      "loss": 0.0335,
+      "step": 14506
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.336734693877551e-05,
+      "loss": 0.0367,
+      "step": 14507
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.335204081632651e-05,
+      "loss": 0.0881,
+      "step": 14508
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.333673469387754e-05,
+      "loss": 0.0895,
+      "step": 14509
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.332142857142856e-05,
+      "loss": 0.0757,
+      "step": 14510
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.330612244897959e-05,
+      "loss": 0.0378,
+      "step": 14511
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.329081632653059e-05,
+      "loss": 0.0434,
+      "step": 14512
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.327551020408162e-05,
+      "loss": 0.0378,
+      "step": 14513
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.326020408163264e-05,
+      "loss": 0.2015,
+      "step": 14514
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.324489795918367e-05,
+      "loss": 0.1252,
+      "step": 14515
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.322959183673467e-05,
+      "loss": 0.034,
+      "step": 14516
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.32142857142857e-05,
+      "loss": 0.1134,
+      "step": 14517
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.319897959183672e-05,
+      "loss": 0.0341,
+      "step": 14518
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.318367346938775e-05,
+      "loss": 0.0673,
+      "step": 14519
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.316836734693876e-05,
+      "loss": 0.1626,
+      "step": 14520
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.315306122448979e-05,
+      "loss": 0.0615,
+      "step": 14521
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.31377551020408e-05,
+      "loss": 0.115,
+      "step": 14522
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.312244897959183e-05,
+      "loss": 0.0614,
+      "step": 14523
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.310714285714284e-05,
+      "loss": 0.0383,
+      "step": 14524
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.309183673469387e-05,
+      "loss": 0.0506,
+      "step": 14525
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.307653061224489e-05,
+      "loss": 0.1192,
+      "step": 14526
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.306122448979592e-05,
+      "loss": 0.0526,
+      "step": 14527
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.304591836734693e-05,
+      "loss": 0.0926,
+      "step": 14528
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.303061224489795e-05,
+      "loss": 0.156,
+      "step": 14529
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.301530612244898e-05,
+      "loss": 0.0727,
+      "step": 14530
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.3e-05,
+      "loss": 0.0507,
+      "step": 14531
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.298469387755101e-05,
+      "loss": 0.0456,
+      "step": 14532
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.296938775510203e-05,
+      "loss": 0.126,
+      "step": 14533
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.295408163265306e-05,
+      "loss": 0.0528,
+      "step": 14534
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.293877551020408e-05,
+      "loss": 0.0805,
+      "step": 14535
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.29234693877551e-05,
+      "loss": 0.0386,
+      "step": 14536
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.290816326530611e-05,
+      "loss": 0.0478,
+      "step": 14537
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.289285714285714e-05,
+      "loss": 0.1079,
+      "step": 14538
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.287755102040815e-05,
+      "loss": 0.17,
+      "step": 14539
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.286224489795918e-05,
+      "loss": 0.0875,
+      "step": 14540
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.28469387755102e-05,
+      "loss": 0.0298,
+      "step": 14541
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.283163265306122e-05,
+      "loss": 0.041,
+      "step": 14542
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.281632653061223e-05,
+      "loss": 0.0728,
+      "step": 14543
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.280102040816326e-05,
+      "loss": 0.0482,
+      "step": 14544
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.278571428571428e-05,
+      "loss": 0.1266,
+      "step": 14545
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.27704081632653e-05,
+      "loss": 0.0469,
+      "step": 14546
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.275510204081631e-05,
+      "loss": 0.0847,
+      "step": 14547
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.273979591836734e-05,
+      "loss": 0.0369,
+      "step": 14548
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.272448979591836e-05,
+      "loss": 0.049,
+      "step": 14549
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.270918367346939e-05,
+      "loss": 0.0299,
+      "step": 14550
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.269387755102039e-05,
+      "loss": 0.1015,
+      "step": 14551
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.267857142857142e-05,
+      "loss": 0.1006,
+      "step": 14552
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.266326530612244e-05,
+      "loss": 0.1308,
+      "step": 14553
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.264795918367347e-05,
+      "loss": 0.0473,
+      "step": 14554
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.263265306122447e-05,
+      "loss": 0.088,
+      "step": 14555
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.26173469387755e-05,
+      "loss": 0.1175,
+      "step": 14556
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.260204081632652e-05,
+      "loss": 0.0968,
+      "step": 14557
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.258673469387755e-05,
+      "loss": 0.1358,
+      "step": 14558
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.257142857142855e-05,
+      "loss": 0.0874,
+      "step": 14559
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.255612244897958e-05,
+      "loss": 0.1223,
+      "step": 14560
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.25408163265306e-05,
+      "loss": 0.084,
+      "step": 14561
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.252551020408163e-05,
+      "loss": 0.0342,
+      "step": 14562
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.251020408163264e-05,
+      "loss": 0.0764,
+      "step": 14563
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.249489795918367e-05,
+      "loss": 0.0874,
+      "step": 14564
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.247959183673468e-05,
+      "loss": 0.1233,
+      "step": 14565
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.246428571428571e-05,
+      "loss": 0.1589,
+      "step": 14566
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.244897959183672e-05,
+      "loss": 0.1555,
+      "step": 14567
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.243367346938775e-05,
+      "loss": 0.1798,
+      "step": 14568
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.241836734693877e-05,
+      "loss": 0.057,
+      "step": 14569
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.240306122448978e-05,
+      "loss": 0.1976,
+      "step": 14570
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.23877551020408e-05,
+      "loss": 0.1586,
+      "step": 14571
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.237244897959183e-05,
+      "loss": 0.2399,
+      "step": 14572
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.235714285714286e-05,
+      "loss": 0.0646,
+      "step": 14573
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.234183673469386e-05,
+      "loss": 0.0811,
+      "step": 14574
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.23265306122449e-05,
+      "loss": 0.0388,
+      "step": 14575
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.231122448979591e-05,
+      "loss": 0.0758,
+      "step": 14576
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.229591836734694e-05,
+      "loss": 0.183,
+      "step": 14577
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.228061224489795e-05,
+      "loss": 0.0434,
+      "step": 14578
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.226530612244898e-05,
+      "loss": 0.0334,
+      "step": 14579
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.224999999999999e-05,
+      "loss": 0.0373,
+      "step": 14580
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.223469387755102e-05,
+      "loss": 0.0375,
+      "step": 14581
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.221938775510203e-05,
+      "loss": 0.0555,
+      "step": 14582
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.220408163265306e-05,
+      "loss": 0.2174,
+      "step": 14583
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.218877551020407e-05,
+      "loss": 0.0795,
+      "step": 14584
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.21734693877551e-05,
+      "loss": 0.0445,
+      "step": 14585
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.215816326530611e-05,
+      "loss": 0.0739,
+      "step": 14586
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.214285714285714e-05,
+      "loss": 0.1413,
+      "step": 14587
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.212755102040816e-05,
+      "loss": 0.1098,
+      "step": 14588
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.211224489795919e-05,
+      "loss": 0.0862,
+      "step": 14589
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.209693877551019e-05,
+      "loss": 0.2062,
+      "step": 14590
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.208163265306122e-05,
+      "loss": 0.0534,
+      "step": 14591
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.206632653061224e-05,
+      "loss": 0.0397,
+      "step": 14592
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.205102040816327e-05,
+      "loss": 0.0289,
+      "step": 14593
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.203571428571427e-05,
+      "loss": 0.0345,
+      "step": 14594
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.20204081632653e-05,
+      "loss": 0.2357,
+      "step": 14595
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.200510204081632e-05,
+      "loss": 0.0979,
+      "step": 14596
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.198979591836735e-05,
+      "loss": 0.0412,
+      "step": 14597
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.197448979591835e-05,
+      "loss": 0.0867,
+      "step": 14598
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.195918367346938e-05,
+      "loss": 0.0382,
+      "step": 14599
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.19438775510204e-05,
+      "loss": 0.043,
+      "step": 14600
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.192857142857142e-05,
+      "loss": 0.1351,
+      "step": 14601
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.191326530612243e-05,
+      "loss": 0.1676,
+      "step": 14602
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.189795918367346e-05,
+      "loss": 0.0462,
+      "step": 14603
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.188265306122448e-05,
+      "loss": 0.0422,
+      "step": 14604
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.18673469387755e-05,
+      "loss": 0.0824,
+      "step": 14605
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.185204081632652e-05,
+      "loss": 0.0776,
+      "step": 14606
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.183673469387755e-05,
+      "loss": 0.0487,
+      "step": 14607
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.182142857142856e-05,
+      "loss": 0.1121,
+      "step": 14608
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.180612244897958e-05,
+      "loss": 0.0824,
+      "step": 14609
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.17908163265306e-05,
+      "loss": 0.0772,
+      "step": 14610
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.177551020408163e-05,
+      "loss": 0.0784,
+      "step": 14611
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.176020408163263e-05,
+      "loss": 0.0427,
+      "step": 14612
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.174489795918366e-05,
+      "loss": 0.0708,
+      "step": 14613
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.172959183673468e-05,
+      "loss": 0.2296,
+      "step": 14614
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.171428571428571e-05,
+      "loss": 0.0509,
+      "step": 14615
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.169897959183671e-05,
+      "loss": 0.1311,
+      "step": 14616
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.168367346938774e-05,
+      "loss": 0.0448,
+      "step": 14617
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.166836734693876e-05,
+      "loss": 0.0387,
+      "step": 14618
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.165306122448979e-05,
+      "loss": 0.0387,
+      "step": 14619
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.163775510204082e-05,
+      "loss": 0.2015,
+      "step": 14620
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.162244897959182e-05,
+      "loss": 0.0843,
+      "step": 14621
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.160714285714286e-05,
+      "loss": 0.1189,
+      "step": 14622
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.159183673469387e-05,
+      "loss": 0.0867,
+      "step": 14623
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.15765306122449e-05,
+      "loss": 0.0384,
+      "step": 14624
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.15612244897959e-05,
+      "loss": 0.192,
+      "step": 14625
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.154591836734694e-05,
+      "loss": 0.0352,
+      "step": 14626
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.153061224489795e-05,
+      "loss": 0.0576,
+      "step": 14627
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.151530612244898e-05,
+      "loss": 0.0467,
+      "step": 14628
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.149999999999999e-05,
+      "loss": 0.0322,
+      "step": 14629
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.148469387755102e-05,
+      "loss": 0.1056,
+      "step": 14630
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.146938775510204e-05,
+      "loss": 0.0365,
+      "step": 14631
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.145408163265305e-05,
+      "loss": 0.0951,
+      "step": 14632
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.143877551020407e-05,
+      "loss": 0.0859,
+      "step": 14633
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.14234693877551e-05,
+      "loss": 0.1241,
+      "step": 14634
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.140816326530612e-05,
+      "loss": 0.1016,
+      "step": 14635
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.139285714285713e-05,
+      "loss": 0.0473,
+      "step": 14636
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.137755102040815e-05,
+      "loss": 0.0759,
+      "step": 14637
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.136224489795918e-05,
+      "loss": 0.0323,
+      "step": 14638
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.13469387755102e-05,
+      "loss": 0.1858,
+      "step": 14639
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.133163265306122e-05,
+      "loss": 0.0872,
+      "step": 14640
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.131632653061223e-05,
+      "loss": 0.0874,
+      "step": 14641
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.130102040816326e-05,
+      "loss": 0.0885,
+      "step": 14642
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.128571428571427e-05,
+      "loss": 0.0285,
+      "step": 14643
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.12704081632653e-05,
+      "loss": 0.0364,
+      "step": 14644
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.125510204081631e-05,
+      "loss": 0.0482,
+      "step": 14645
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.123979591836734e-05,
+      "loss": 0.0878,
+      "step": 14646
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.122448979591835e-05,
+      "loss": 0.0545,
+      "step": 14647
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.120918367346938e-05,
+      "loss": 0.0399,
+      "step": 14648
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.11938775510204e-05,
+      "loss": 0.0492,
+      "step": 14649
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.117857142857143e-05,
+      "loss": 0.1269,
+      "step": 14650
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.116326530612243e-05,
+      "loss": 0.0505,
+      "step": 14651
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.114795918367346e-05,
+      "loss": 0.1683,
+      "step": 14652
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.113265306122448e-05,
+      "loss": 0.1715,
+      "step": 14653
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.111734693877551e-05,
+      "loss": 0.0719,
+      "step": 14654
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.110204081632651e-05,
+      "loss": 0.0697,
+      "step": 14655
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.108673469387754e-05,
+      "loss": 0.0624,
+      "step": 14656
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.107142857142856e-05,
+      "loss": 0.0747,
+      "step": 14657
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.105612244897959e-05,
+      "loss": 0.0593,
+      "step": 14658
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.104081632653059e-05,
+      "loss": 0.095,
+      "step": 14659
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.102551020408162e-05,
+      "loss": 0.1828,
+      "step": 14660
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.101020408163264e-05,
+      "loss": 0.1905,
+      "step": 14661
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.099489795918367e-05,
+      "loss": 0.045,
+      "step": 14662
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.097959183673467e-05,
+      "loss": 0.0365,
+      "step": 14663
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.09642857142857e-05,
+      "loss": 0.1056,
+      "step": 14664
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.094897959183674e-05,
+      "loss": 0.0441,
+      "step": 14665
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.093367346938775e-05,
+      "loss": 0.13,
+      "step": 14666
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.091836734693877e-05,
+      "loss": 0.0855,
+      "step": 14667
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.090306122448979e-05,
+      "loss": 0.0945,
+      "step": 14668
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.088775510204082e-05,
+      "loss": 0.0451,
+      "step": 14669
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.087244897959183e-05,
+      "loss": 0.0399,
+      "step": 14670
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.085714285714285e-05,
+      "loss": 0.0861,
+      "step": 14671
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.084183673469387e-05,
+      "loss": 0.0394,
+      "step": 14672
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.08265306122449e-05,
+      "loss": 0.0413,
+      "step": 14673
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.08112244897959e-05,
+      "loss": 0.0987,
+      "step": 14674
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.079591836734693e-05,
+      "loss": 0.033,
+      "step": 14675
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.078061224489795e-05,
+      "loss": 0.0765,
+      "step": 14676
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.076530612244898e-05,
+      "loss": 0.2099,
+      "step": 14677
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.074999999999998e-05,
+      "loss": 0.0865,
+      "step": 14678
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.073469387755101e-05,
+      "loss": 0.0406,
+      "step": 14679
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.071938775510203e-05,
+      "loss": 0.0356,
+      "step": 14680
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.070408163265306e-05,
+      "loss": 0.0509,
+      "step": 14681
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.068877551020406e-05,
+      "loss": 0.0404,
+      "step": 14682
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.06734693877551e-05,
+      "loss": 0.1124,
+      "step": 14683
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.065816326530611e-05,
+      "loss": 0.0492,
+      "step": 14684
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.064285714285714e-05,
+      "loss": 0.0745,
+      "step": 14685
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.062755102040815e-05,
+      "loss": 0.1879,
+      "step": 14686
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.061224489795918e-05,
+      "loss": 0.115,
+      "step": 14687
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.05969387755102e-05,
+      "loss": 0.0375,
+      "step": 14688
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.058163265306122e-05,
+      "loss": 0.1398,
+      "step": 14689
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.056632653061223e-05,
+      "loss": 0.065,
+      "step": 14690
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.055102040816326e-05,
+      "loss": 0.0768,
+      "step": 14691
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.053571428571428e-05,
+      "loss": 0.0442,
+      "step": 14692
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.05204081632653e-05,
+      "loss": 0.0746,
+      "step": 14693
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.050510204081631e-05,
+      "loss": 0.0345,
+      "step": 14694
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.048979591836734e-05,
+      "loss": 0.0799,
+      "step": 14695
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.047448979591836e-05,
+      "loss": 0.2135,
+      "step": 14696
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.045918367346939e-05,
+      "loss": 0.1012,
+      "step": 14697
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.044387755102039e-05,
+      "loss": 0.0373,
+      "step": 14698
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.042857142857142e-05,
+      "loss": 0.1111,
+      "step": 14699
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.041326530612244e-05,
+      "loss": 0.034,
+      "step": 14700
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.039795918367347e-05,
+      "loss": 0.0326,
+      "step": 14701
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.038265306122447e-05,
+      "loss": 0.3563,
+      "step": 14702
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.03673469387755e-05,
+      "loss": 0.0869,
+      "step": 14703
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.035204081632652e-05,
+      "loss": 0.0458,
+      "step": 14704
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.033673469387755e-05,
+      "loss": 0.035,
+      "step": 14705
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.032142857142855e-05,
+      "loss": 0.0697,
+      "step": 14706
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.030612244897958e-05,
+      "loss": 0.1013,
+      "step": 14707
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.02908163265306e-05,
+      "loss": 0.1244,
+      "step": 14708
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.027551020408162e-05,
+      "loss": 0.0391,
+      "step": 14709
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.026020408163265e-05,
+      "loss": 0.0477,
+      "step": 14710
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.024489795918367e-05,
+      "loss": 0.0589,
+      "step": 14711
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.02295918367347e-05,
+      "loss": 0.0405,
+      "step": 14712
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.02142857142857e-05,
+      "loss": 0.0466,
+      "step": 14713
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.019897959183673e-05,
+      "loss": 0.2483,
+      "step": 14714
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.018367346938775e-05,
+      "loss": 0.1678,
+      "step": 14715
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.016836734693878e-05,
+      "loss": 0.0829,
+      "step": 14716
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.015306122448978e-05,
+      "loss": 0.0469,
+      "step": 14717
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.013775510204081e-05,
+      "loss": 0.0745,
+      "step": 14718
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.012244897959183e-05,
+      "loss": 0.0517,
+      "step": 14719
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.010714285714286e-05,
+      "loss": 0.0297,
+      "step": 14720
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.009183673469386e-05,
+      "loss": 0.081,
+      "step": 14721
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.00765306122449e-05,
+      "loss": 0.167,
+      "step": 14722
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.006122448979591e-05,
+      "loss": 0.043,
+      "step": 14723
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 9.004591836734694e-05,
+      "loss": 0.0523,
+      "step": 14724
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 9.003061224489794e-05,
+      "loss": 0.0395,
+      "step": 14725
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 9.001530612244898e-05,
+      "loss": 0.1195,
+      "step": 14726
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 0.1479,
+      "step": 14727
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.998469387755102e-05,
+      "loss": 0.0914,
+      "step": 14728
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.996938775510203e-05,
+      "loss": 0.069,
+      "step": 14729
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.995408163265306e-05,
+      "loss": 0.03,
+      "step": 14730
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.993877551020407e-05,
+      "loss": 0.1122,
+      "step": 14731
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.99234693877551e-05,
+      "loss": 0.0678,
+      "step": 14732
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.990816326530611e-05,
+      "loss": 0.0674,
+      "step": 14733
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.989285714285714e-05,
+      "loss": 0.061,
+      "step": 14734
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.987755102040816e-05,
+      "loss": 0.0672,
+      "step": 14735
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.986224489795919e-05,
+      "loss": 0.0485,
+      "step": 14736
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.984693877551019e-05,
+      "loss": 0.0297,
+      "step": 14737
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.983163265306122e-05,
+      "loss": 0.0439,
+      "step": 14738
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.981632653061224e-05,
+      "loss": 0.1379,
+      "step": 14739
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.980102040816325e-05,
+      "loss": 0.0504,
+      "step": 14740
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.978571428571427e-05,
+      "loss": 0.0879,
+      "step": 14741
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.97704081632653e-05,
+      "loss": 0.081,
+      "step": 14742
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.975510204081632e-05,
+      "loss": 0.1099,
+      "step": 14743
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.973979591836734e-05,
+      "loss": 0.0689,
+      "step": 14744
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.972448979591835e-05,
+      "loss": 0.0763,
+      "step": 14745
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.970918367346938e-05,
+      "loss": 0.1258,
+      "step": 14746
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.96938775510204e-05,
+      "loss": 0.0742,
+      "step": 14747
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.967857142857142e-05,
+      "loss": 0.112,
+      "step": 14748
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.966326530612243e-05,
+      "loss": 0.0693,
+      "step": 14749
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.964795918367346e-05,
+      "loss": 0.0823,
+      "step": 14750
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.963265306122447e-05,
+      "loss": 0.0315,
+      "step": 14751
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.96173469387755e-05,
+      "loss": 0.0461,
+      "step": 14752
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.960204081632652e-05,
+      "loss": 0.1605,
+      "step": 14753
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.958673469387755e-05,
+      "loss": 0.0774,
+      "step": 14754
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.957142857142858e-05,
+      "loss": 0.0953,
+      "step": 14755
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.955612244897958e-05,
+      "loss": 0.0877,
+      "step": 14756
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.954081632653061e-05,
+      "loss": 0.0574,
+      "step": 14757
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.952551020408163e-05,
+      "loss": 0.3538,
+      "step": 14758
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.951020408163266e-05,
+      "loss": 0.0753,
+      "step": 14759
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.949489795918366e-05,
+      "loss": 0.127,
+      "step": 14760
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.947959183673469e-05,
+      "loss": 0.0799,
+      "step": 14761
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.946428571428571e-05,
+      "loss": 0.0925,
+      "step": 14762
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.944897959183674e-05,
+      "loss": 0.0493,
+      "step": 14763
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.943367346938774e-05,
+      "loss": 0.2048,
+      "step": 14764
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.941836734693877e-05,
+      "loss": 0.1726,
+      "step": 14765
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.940306122448979e-05,
+      "loss": 0.0755,
+      "step": 14766
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.938775510204082e-05,
+      "loss": 0.1055,
+      "step": 14767
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.937244897959182e-05,
+      "loss": 0.0775,
+      "step": 14768
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.935714285714285e-05,
+      "loss": 0.0338,
+      "step": 14769
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.934183673469387e-05,
+      "loss": 0.0951,
+      "step": 14770
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.932653061224489e-05,
+      "loss": 0.0891,
+      "step": 14771
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.93112244897959e-05,
+      "loss": 0.074,
+      "step": 14772
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.929591836734694e-05,
+      "loss": 0.045,
+      "step": 14773
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.928061224489795e-05,
+      "loss": 0.1178,
+      "step": 14774
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.926530612244897e-05,
+      "loss": 0.1473,
+      "step": 14775
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.924999999999999e-05,
+      "loss": 0.0976,
+      "step": 14776
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.923469387755102e-05,
+      "loss": 0.206,
+      "step": 14777
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.921938775510203e-05,
+      "loss": 0.0862,
+      "step": 14778
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.920408163265305e-05,
+      "loss": 0.0439,
+      "step": 14779
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.918877551020407e-05,
+      "loss": 0.0868,
+      "step": 14780
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.91734693877551e-05,
+      "loss": 0.0311,
+      "step": 14781
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.91581632653061e-05,
+      "loss": 0.1335,
+      "step": 14782
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.914285714285713e-05,
+      "loss": 0.0709,
+      "step": 14783
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.912755102040815e-05,
+      "loss": 0.0856,
+      "step": 14784
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.911224489795918e-05,
+      "loss": 0.1759,
+      "step": 14785
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.909693877551018e-05,
+      "loss": 0.0338,
+      "step": 14786
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.908163265306122e-05,
+      "loss": 0.0822,
+      "step": 14787
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.906632653061223e-05,
+      "loss": 0.0706,
+      "step": 14788
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.905102040816326e-05,
+      "loss": 0.1175,
+      "step": 14789
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.903571428571427e-05,
+      "loss": 0.046,
+      "step": 14790
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.90204081632653e-05,
+      "loss": 0.0385,
+      "step": 14791
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.900510204081631e-05,
+      "loss": 0.0772,
+      "step": 14792
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.898979591836734e-05,
+      "loss": 0.1196,
+      "step": 14793
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.897448979591835e-05,
+      "loss": 0.0911,
+      "step": 14794
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.895918367346938e-05,
+      "loss": 0.0399,
+      "step": 14795
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.89438775510204e-05,
+      "loss": 0.1511,
+      "step": 14796
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.892857142857143e-05,
+      "loss": 0.0721,
+      "step": 14797
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.891326530612243e-05,
+      "loss": 0.0399,
+      "step": 14798
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.889795918367346e-05,
+      "loss": 0.08,
+      "step": 14799
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.888265306122449e-05,
+      "loss": 0.0343,
+      "step": 14800
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.886734693877551e-05,
+      "loss": 0.0547,
+      "step": 14801
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.885204081632652e-05,
+      "loss": 0.1259,
+      "step": 14802
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.883673469387754e-05,
+      "loss": 0.0803,
+      "step": 14803
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.882142857142857e-05,
+      "loss": 0.0881,
+      "step": 14804
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.880612244897959e-05,
+      "loss": 0.1365,
+      "step": 14805
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.87908163265306e-05,
+      "loss": 0.0869,
+      "step": 14806
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.877551020408162e-05,
+      "loss": 0.0317,
+      "step": 14807
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.876020408163265e-05,
+      "loss": 0.1273,
+      "step": 14808
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.874489795918367e-05,
+      "loss": 0.0848,
+      "step": 14809
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.872959183673469e-05,
+      "loss": 0.1226,
+      "step": 14810
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.87142857142857e-05,
+      "loss": 0.1213,
+      "step": 14811
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.869897959183673e-05,
+      "loss": 0.0905,
+      "step": 14812
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.868367346938774e-05,
+      "loss": 0.0564,
+      "step": 14813
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.866836734693877e-05,
+      "loss": 0.0887,
+      "step": 14814
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.865306122448979e-05,
+      "loss": 0.0804,
+      "step": 14815
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.863775510204082e-05,
+      "loss": 0.0377,
+      "step": 14816
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.862244897959182e-05,
+      "loss": 0.0752,
+      "step": 14817
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.860714285714285e-05,
+      "loss": 0.0382,
+      "step": 14818
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.859183673469387e-05,
+      "loss": 0.0388,
+      "step": 14819
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.85765306122449e-05,
+      "loss": 0.1738,
+      "step": 14820
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.85612244897959e-05,
+      "loss": 0.1135,
+      "step": 14821
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.854591836734693e-05,
+      "loss": 0.0805,
+      "step": 14822
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.853061224489795e-05,
+      "loss": 0.0764,
+      "step": 14823
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.851530612244898e-05,
+      "loss": 0.0794,
+      "step": 14824
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.849999999999998e-05,
+      "loss": 0.0341,
+      "step": 14825
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.848469387755101e-05,
+      "loss": 0.0877,
+      "step": 14826
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.846938775510203e-05,
+      "loss": 0.126,
+      "step": 14827
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.845408163265306e-05,
+      "loss": 0.0402,
+      "step": 14828
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.843877551020406e-05,
+      "loss": 0.0427,
+      "step": 14829
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.84234693877551e-05,
+      "loss": 0.0343,
+      "step": 14830
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.840816326530611e-05,
+      "loss": 0.0286,
+      "step": 14831
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.839285714285714e-05,
+      "loss": 0.1341,
+      "step": 14832
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.837755102040815e-05,
+      "loss": 0.1732,
+      "step": 14833
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.836224489795918e-05,
+      "loss": 0.0421,
+      "step": 14834
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.83469387755102e-05,
+      "loss": 0.0898,
+      "step": 14835
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.833163265306122e-05,
+      "loss": 0.0848,
+      "step": 14836
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.831632653061223e-05,
+      "loss": 0.0431,
+      "step": 14837
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.830102040816326e-05,
+      "loss": 0.1271,
+      "step": 14838
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.828571428571427e-05,
+      "loss": 0.1335,
+      "step": 14839
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.82704081632653e-05,
+      "loss": 0.1348,
+      "step": 14840
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.825510204081631e-05,
+      "loss": 0.0857,
+      "step": 14841
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.823979591836734e-05,
+      "loss": 0.0537,
+      "step": 14842
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.822448979591836e-05,
+      "loss": 0.1245,
+      "step": 14843
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.820918367346939e-05,
+      "loss": 0.0719,
+      "step": 14844
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.81938775510204e-05,
+      "loss": 0.1122,
+      "step": 14845
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.817857142857142e-05,
+      "loss": 0.0531,
+      "step": 14846
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.816326530612245e-05,
+      "loss": 0.1922,
+      "step": 14847
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.814795918367345e-05,
+      "loss": 0.0371,
+      "step": 14848
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.813265306122449e-05,
+      "loss": 0.0306,
+      "step": 14849
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.81173469387755e-05,
+      "loss": 0.0379,
+      "step": 14850
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.810204081632653e-05,
+      "loss": 0.0634,
+      "step": 14851
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.808673469387754e-05,
+      "loss": 0.1491,
+      "step": 14852
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.807142857142857e-05,
+      "loss": 0.1246,
+      "step": 14853
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.805612244897958e-05,
+      "loss": 0.0621,
+      "step": 14854
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.804081632653061e-05,
+      "loss": 0.1253,
+      "step": 14855
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.802551020408162e-05,
+      "loss": 0.1162,
+      "step": 14856
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 8.801020408163265e-05,
+      "loss": 0.1128,
+      "step": 14857
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.799489795918367e-05,
+      "loss": 0.0716,
+      "step": 14858
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.79795918367347e-05,
+      "loss": 0.0467,
+      "step": 14859
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.79642857142857e-05,
+      "loss": 0.0886,
+      "step": 14860
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.794897959183673e-05,
+      "loss": 0.1166,
+      "step": 14861
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.793367346938775e-05,
+      "loss": 0.1532,
+      "step": 14862
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.791836734693878e-05,
+      "loss": 0.1191,
+      "step": 14863
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.790306122448978e-05,
+      "loss": 0.1066,
+      "step": 14864
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.788775510204081e-05,
+      "loss": 0.0824,
+      "step": 14865
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.787244897959183e-05,
+      "loss": 0.0797,
+      "step": 14866
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.785714285714286e-05,
+      "loss": 0.049,
+      "step": 14867
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.784183673469386e-05,
+      "loss": 0.0572,
+      "step": 14868
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.782653061224489e-05,
+      "loss": 0.1233,
+      "step": 14869
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.781122448979591e-05,
+      "loss": 0.1189,
+      "step": 14870
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.779591836734694e-05,
+      "loss": 0.0916,
+      "step": 14871
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.778061224489794e-05,
+      "loss": 0.082,
+      "step": 14872
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.776530612244897e-05,
+      "loss": 0.0398,
+      "step": 14873
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.774999999999999e-05,
+      "loss": 0.0751,
+      "step": 14874
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.773469387755102e-05,
+      "loss": 0.0762,
+      "step": 14875
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.771938775510203e-05,
+      "loss": 0.0464,
+      "step": 14876
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.770408163265306e-05,
+      "loss": 0.0562,
+      "step": 14877
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.768877551020407e-05,
+      "loss": 0.0834,
+      "step": 14878
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.767346938775509e-05,
+      "loss": 0.1247,
+      "step": 14879
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.765816326530611e-05,
+      "loss": 0.0471,
+      "step": 14880
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.764285714285714e-05,
+      "loss": 0.0839,
+      "step": 14881
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.762755102040815e-05,
+      "loss": 0.1846,
+      "step": 14882
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.761224489795917e-05,
+      "loss": 0.1302,
+      "step": 14883
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.759693877551019e-05,
+      "loss": 0.1623,
+      "step": 14884
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.758163265306122e-05,
+      "loss": 0.0478,
+      "step": 14885
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.756632653061224e-05,
+      "loss": 0.0666,
+      "step": 14886
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.755102040816325e-05,
+      "loss": 0.1317,
+      "step": 14887
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.753571428571427e-05,
+      "loss": 0.0383,
+      "step": 14888
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.75204081632653e-05,
+      "loss": 0.1069,
+      "step": 14889
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.75051020408163e-05,
+      "loss": 0.0836,
+      "step": 14890
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.748979591836733e-05,
+      "loss": 0.0513,
+      "step": 14891
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.747448979591837e-05,
+      "loss": 0.0799,
+      "step": 14892
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.745918367346938e-05,
+      "loss": 0.0522,
+      "step": 14893
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.744387755102041e-05,
+      "loss": 0.0761,
+      "step": 14894
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.742857142857142e-05,
+      "loss": 0.1637,
+      "step": 14895
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.741326530612245e-05,
+      "loss": 0.0749,
+      "step": 14896
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.739795918367346e-05,
+      "loss": 0.1146,
+      "step": 14897
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.73826530612245e-05,
+      "loss": 0.0369,
+      "step": 14898
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.73673469387755e-05,
+      "loss": 0.0731,
+      "step": 14899
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.735204081632653e-05,
+      "loss": 0.0496,
+      "step": 14900
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.733673469387755e-05,
+      "loss": 0.0673,
+      "step": 14901
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.732142857142858e-05,
+      "loss": 0.18,
+      "step": 14902
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.730612244897958e-05,
+      "loss": 0.1174,
+      "step": 14903
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.729081632653061e-05,
+      "loss": 0.0834,
+      "step": 14904
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.727551020408163e-05,
+      "loss": 0.108,
+      "step": 14905
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.726020408163266e-05,
+      "loss": 0.0406,
+      "step": 14906
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.724489795918366e-05,
+      "loss": 0.0483,
+      "step": 14907
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.722959183673469e-05,
+      "loss": 0.1201,
+      "step": 14908
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 8.721428571428571e-05,
+      "loss": 0.0402,
+      "step": 14909
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.719897959183673e-05,
+      "loss": 0.0397,
+      "step": 14910
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.718367346938774e-05,
+      "loss": 0.0382,
+      "step": 14911
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.716836734693877e-05,
+      "loss": 0.0727,
+      "step": 14912
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.715306122448979e-05,
+      "loss": 0.086,
+      "step": 14913
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.713775510204081e-05,
+      "loss": 0.1785,
+      "step": 14914
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.712244897959182e-05,
+      "loss": 0.089,
+      "step": 14915
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.710714285714285e-05,
+      "loss": 0.0481,
+      "step": 14916
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.709183673469387e-05,
+      "loss": 0.0325,
+      "step": 14917
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.707653061224489e-05,
+      "loss": 0.0338,
+      "step": 14918
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.70612244897959e-05,
+      "loss": 0.0968,
+      "step": 14919
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.704591836734694e-05,
+      "loss": 0.1569,
+      "step": 14920
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.703061224489794e-05,
+      "loss": 0.2343,
+      "step": 14921
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.701530612244897e-05,
+      "loss": 0.1654,
+      "step": 14922
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.699999999999999e-05,
+      "loss": 0.0768,
+      "step": 14923
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.698469387755102e-05,
+      "loss": 0.0419,
+      "step": 14924
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.696938775510202e-05,
+      "loss": 0.0563,
+      "step": 14925
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.695408163265305e-05,
+      "loss": 0.0458,
+      "step": 14926
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.693877551020407e-05,
+      "loss": 0.2129,
+      "step": 14927
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.69234693877551e-05,
+      "loss": 0.0755,
+      "step": 14928
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.69081632653061e-05,
+      "loss": 0.045,
+      "step": 14929
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.689285714285713e-05,
+      "loss": 0.0592,
+      "step": 14930
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.687755102040815e-05,
+      "loss": 0.0806,
+      "step": 14931
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.686224489795918e-05,
+      "loss": 0.11,
+      "step": 14932
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.684693877551018e-05,
+      "loss": 0.1479,
+      "step": 14933
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.683163265306121e-05,
+      "loss": 0.1315,
+      "step": 14934
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.681632653061223e-05,
+      "loss": 0.1082,
+      "step": 14935
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.680102040816326e-05,
+      "loss": 0.1656,
+      "step": 14936
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.678571428571429e-05,
+      "loss": 0.0477,
+      "step": 14937
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.67704081632653e-05,
+      "loss": 0.0683,
+      "step": 14938
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.675510204081633e-05,
+      "loss": 0.1341,
+      "step": 14939
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.673979591836734e-05,
+      "loss": 0.0438,
+      "step": 14940
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.672448979591836e-05,
+      "loss": 0.0835,
+      "step": 14941
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.670918367346938e-05,
+      "loss": 0.0475,
+      "step": 14942
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.669387755102041e-05,
+      "loss": 0.0484,
+      "step": 14943
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.667857142857143e-05,
+      "loss": 0.0742,
+      "step": 14944
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.666326530612244e-05,
+      "loss": 0.0799,
+      "step": 14945
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.664795918367346e-05,
+      "loss": 0.106,
+      "step": 14946
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.663265306122449e-05,
+      "loss": 0.1262,
+      "step": 14947
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.66173469387755e-05,
+      "loss": 0.0316,
+      "step": 14948
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.660204081632652e-05,
+      "loss": 0.0781,
+      "step": 14949
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.658673469387754e-05,
+      "loss": 0.0838,
+      "step": 14950
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.657142857142857e-05,
+      "loss": 0.1074,
+      "step": 14951
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.655612244897957e-05,
+      "loss": 0.133,
+      "step": 14952
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.65408163265306e-05,
+      "loss": 0.0394,
+      "step": 14953
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.652551020408162e-05,
+      "loss": 0.0808,
+      "step": 14954
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.651020408163265e-05,
+      "loss": 0.0308,
+      "step": 14955
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.649489795918366e-05,
+      "loss": 0.0669,
+      "step": 14956
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.647959183673469e-05,
+      "loss": 0.0995,
+      "step": 14957
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.64642857142857e-05,
+      "loss": 0.1241,
+      "step": 14958
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.644897959183673e-05,
+      "loss": 0.0368,
+      "step": 14959
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 8.643367346938774e-05,
+      "loss": 0.0821,
+      "step": 14960
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.641836734693877e-05,
+      "loss": 0.1571,
+      "step": 14961
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.640306122448979e-05,
+      "loss": 0.0761,
+      "step": 14962
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.638775510204082e-05,
+      "loss": 0.0704,
+      "step": 14963
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.637244897959182e-05,
+      "loss": 0.1051,
+      "step": 14964
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.635714285714285e-05,
+      "loss": 0.0858,
+      "step": 14965
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.634183673469387e-05,
+      "loss": 0.0876,
+      "step": 14966
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.63265306122449e-05,
+      "loss": 0.0369,
+      "step": 14967
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.63112244897959e-05,
+      "loss": 0.0436,
+      "step": 14968
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.629591836734693e-05,
+      "loss": 0.0483,
+      "step": 14969
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.628061224489795e-05,
+      "loss": 0.1243,
+      "step": 14970
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.626530612244898e-05,
+      "loss": 0.0524,
+      "step": 14971
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.624999999999998e-05,
+      "loss": 0.1606,
+      "step": 14972
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.623469387755101e-05,
+      "loss": 0.0773,
+      "step": 14973
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.621938775510203e-05,
+      "loss": 0.0521,
+      "step": 14974
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.620408163265306e-05,
+      "loss": 0.0518,
+      "step": 14975
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.618877551020406e-05,
+      "loss": 0.0393,
+      "step": 14976
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.61734693877551e-05,
+      "loss": 0.1463,
+      "step": 14977
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.615816326530611e-05,
+      "loss": 0.1931,
+      "step": 14978
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.614285714285714e-05,
+      "loss": 0.0732,
+      "step": 14979
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.612755102040815e-05,
+      "loss": 0.1108,
+      "step": 14980
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.611224489795918e-05,
+      "loss": 0.03,
+      "step": 14981
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.60969387755102e-05,
+      "loss": 0.0387,
+      "step": 14982
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.608163265306121e-05,
+      "loss": 0.0495,
+      "step": 14983
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.606632653061224e-05,
+      "loss": 0.0857,
+      "step": 14984
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.605102040816326e-05,
+      "loss": 0.1306,
+      "step": 14985
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.603571428571429e-05,
+      "loss": 0.0376,
+      "step": 14986
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.602040816326529e-05,
+      "loss": 0.0741,
+      "step": 14987
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.600510204081632e-05,
+      "loss": 0.0328,
+      "step": 14988
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.598979591836734e-05,
+      "loss": 0.0401,
+      "step": 14989
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.597448979591837e-05,
+      "loss": 0.0374,
+      "step": 14990
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.595918367346937e-05,
+      "loss": 0.1129,
+      "step": 14991
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.59438775510204e-05,
+      "loss": 0.037,
+      "step": 14992
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.592857142857142e-05,
+      "loss": 0.0733,
+      "step": 14993
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.591326530612245e-05,
+      "loss": 0.0868,
+      "step": 14994
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.589795918367345e-05,
+      "loss": 0.074,
+      "step": 14995
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.588265306122448e-05,
+      "loss": 0.1355,
+      "step": 14996
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.58673469387755e-05,
+      "loss": 0.1168,
+      "step": 14997
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.585204081632653e-05,
+      "loss": 0.0425,
+      "step": 14998
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.583673469387754e-05,
+      "loss": 0.0364,
+      "step": 14999
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.582142857142857e-05,
+      "loss": 0.0265,
+      "step": 15000
+    },
+    {
+      "epoch": 14.56,
+      "eval_cer": 0.2563266819439452,
+      "eval_loss": 1.1152018308639526,
+      "eval_runtime": 228.4383,
+      "eval_samples_per_second": 38.584,
+      "eval_steps_per_second": 4.824,
+      "eval_wer": 0.6606685781454262,
+      "step": 15000
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.580612244897958e-05,
+      "loss": 0.0955,
+      "step": 15001
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.579081632653061e-05,
+      "loss": 0.2536,
+      "step": 15002
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.577551020408162e-05,
+      "loss": 0.0871,
+      "step": 15003
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.576020408163265e-05,
+      "loss": 0.0378,
+      "step": 15004
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.574489795918367e-05,
+      "loss": 0.108,
+      "step": 15005
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.57295918367347e-05,
+      "loss": 0.0945,
+      "step": 15006
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.57142857142857e-05,
+      "loss": 0.0569,
+      "step": 15007
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.569897959183673e-05,
+      "loss": 0.0952,
+      "step": 15008
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.568367346938775e-05,
+      "loss": 0.089,
+      "step": 15009
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.566836734693878e-05,
+      "loss": 0.0864,
+      "step": 15010
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.565306122448978e-05,
+      "loss": 0.0858,
+      "step": 15011
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.563775510204081e-05,
+      "loss": 0.041,
+      "step": 15012
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.562244897959183e-05,
+      "loss": 0.1215,
+      "step": 15013
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.560714285714286e-05,
+      "loss": 0.114,
+      "step": 15014
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.559183673469386e-05,
+      "loss": 0.167,
+      "step": 15015
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.557653061224489e-05,
+      "loss": 0.117,
+      "step": 15016
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.556122448979591e-05,
+      "loss": 0.042,
+      "step": 15017
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.554591836734693e-05,
+      "loss": 0.047,
+      "step": 15018
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.553061224489794e-05,
+      "loss": 0.1102,
+      "step": 15019
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.551530612244897e-05,
+      "loss": 0.126,
+      "step": 15020
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.549999999999999e-05,
+      "loss": 0.1329,
+      "step": 15021
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.548469387755101e-05,
+      "loss": 0.0829,
+      "step": 15022
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.546938775510203e-05,
+      "loss": 0.0318,
+      "step": 15023
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.545408163265306e-05,
+      "loss": 0.0757,
+      "step": 15024
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.543877551020407e-05,
+      "loss": 0.1141,
+      "step": 15025
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.542346938775509e-05,
+      "loss": 0.0683,
+      "step": 15026
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.540816326530612e-05,
+      "loss": 0.2384,
+      "step": 15027
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.539285714285714e-05,
+      "loss": 0.083,
+      "step": 15028
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.537755102040817e-05,
+      "loss": 0.0444,
+      "step": 15029
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.536224489795917e-05,
+      "loss": 0.1138,
+      "step": 15030
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.53469387755102e-05,
+      "loss": 0.1518,
+      "step": 15031
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.533163265306122e-05,
+      "loss": 0.0784,
+      "step": 15032
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.531632653061225e-05,
+      "loss": 0.0769,
+      "step": 15033
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.530102040816325e-05,
+      "loss": 0.0484,
+      "step": 15034
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.528571428571428e-05,
+      "loss": 0.0462,
+      "step": 15035
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.52704081632653e-05,
+      "loss": 0.0859,
+      "step": 15036
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.525510204081633e-05,
+      "loss": 0.1076,
+      "step": 15037
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.523979591836733e-05,
+      "loss": 0.0383,
+      "step": 15038
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.522448979591836e-05,
+      "loss": 0.1931,
+      "step": 15039
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.520918367346938e-05,
+      "loss": 0.0904,
+      "step": 15040
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.519387755102041e-05,
+      "loss": 0.126,
+      "step": 15041
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.517857142857142e-05,
+      "loss": 0.1094,
+      "step": 15042
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.516326530612245e-05,
+      "loss": 0.0638,
+      "step": 15043
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.514795918367346e-05,
+      "loss": 0.0859,
+      "step": 15044
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.51326530612245e-05,
+      "loss": 0.0381,
+      "step": 15045
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.51173469387755e-05,
+      "loss": 0.0823,
+      "step": 15046
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.510204081632653e-05,
+      "loss": 0.1168,
+      "step": 15047
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.508673469387754e-05,
+      "loss": 0.1017,
+      "step": 15048
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.507142857142856e-05,
+      "loss": 0.0295,
+      "step": 15049
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.505612244897958e-05,
+      "loss": 0.1735,
+      "step": 15050
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.504081632653061e-05,
+      "loss": 0.0456,
+      "step": 15051
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.502551020408163e-05,
+      "loss": 0.2246,
+      "step": 15052
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.501020408163264e-05,
+      "loss": 0.0471,
+      "step": 15053
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.499489795918366e-05,
+      "loss": 0.1116,
+      "step": 15054
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.497959183673469e-05,
+      "loss": 0.0322,
+      "step": 15055
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.496428571428571e-05,
+      "loss": 0.0527,
+      "step": 15056
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.494897959183672e-05,
+      "loss": 0.0503,
+      "step": 15057
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.493367346938774e-05,
+      "loss": 0.0844,
+      "step": 15058
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.491836734693877e-05,
+      "loss": 0.0489,
+      "step": 15059
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.490306122448978e-05,
+      "loss": 0.0386,
+      "step": 15060
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.48877551020408e-05,
+      "loss": 0.126,
+      "step": 15061
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.487244897959182e-05,
+      "loss": 0.0665,
+      "step": 15062
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.485714285714285e-05,
+      "loss": 0.0416,
+      "step": 15063
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.484183673469386e-05,
+      "loss": 0.0289,
+      "step": 15064
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.482653061224489e-05,
+      "loss": 0.1409,
+      "step": 15065
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.48112244897959e-05,
+      "loss": 0.0696,
+      "step": 15066
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.479591836734694e-05,
+      "loss": 0.0814,
+      "step": 15067
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.478061224489794e-05,
+      "loss": 0.048,
+      "step": 15068
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.476530612244897e-05,
+      "loss": 0.0499,
+      "step": 15069
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.474999999999999e-05,
+      "loss": 0.1211,
+      "step": 15070
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.473469387755102e-05,
+      "loss": 0.0815,
+      "step": 15071
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.471938775510205e-05,
+      "loss": 0.0855,
+      "step": 15072
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.470408163265305e-05,
+      "loss": 0.0361,
+      "step": 15073
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.468877551020408e-05,
+      "loss": 0.0488,
+      "step": 15074
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.46734693877551e-05,
+      "loss": 0.109,
+      "step": 15075
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.465816326530613e-05,
+      "loss": 0.0817,
+      "step": 15076
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.464285714285713e-05,
+      "loss": 0.0909,
+      "step": 15077
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.462755102040816e-05,
+      "loss": 0.0802,
+      "step": 15078
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.461224489795918e-05,
+      "loss": 0.087,
+      "step": 15079
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.45969387755102e-05,
+      "loss": 0.104,
+      "step": 15080
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.458163265306121e-05,
+      "loss": 0.0317,
+      "step": 15081
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.456632653061224e-05,
+      "loss": 0.0501,
+      "step": 15082
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.455102040816326e-05,
+      "loss": 0.0584,
+      "step": 15083
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.453571428571428e-05,
+      "loss": 0.0848,
+      "step": 15084
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.45204081632653e-05,
+      "loss": 0.0498,
+      "step": 15085
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.450510204081633e-05,
+      "loss": 0.1205,
+      "step": 15086
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.448979591836734e-05,
+      "loss": 0.0284,
+      "step": 15087
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.447448979591836e-05,
+      "loss": 0.0398,
+      "step": 15088
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.445918367346938e-05,
+      "loss": 0.1641,
+      "step": 15089
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.444387755102041e-05,
+      "loss": 0.1251,
+      "step": 15090
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.442857142857141e-05,
+      "loss": 0.0792,
+      "step": 15091
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.441326530612244e-05,
+      "loss": 0.0378,
+      "step": 15092
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.439795918367346e-05,
+      "loss": 0.1145,
+      "step": 15093
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 8.438265306122449e-05,
+      "loss": 0.0592,
+      "step": 15094
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.436734693877549e-05,
+      "loss": 0.1525,
+      "step": 15095
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.435204081632652e-05,
+      "loss": 0.2116,
+      "step": 15096
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.433673469387754e-05,
+      "loss": 0.12,
+      "step": 15097
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.432142857142857e-05,
+      "loss": 0.0766,
+      "step": 15098
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.430612244897957e-05,
+      "loss": 0.0809,
+      "step": 15099
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.42908163265306e-05,
+      "loss": 0.0511,
+      "step": 15100
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.427551020408162e-05,
+      "loss": 0.1286,
+      "step": 15101
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.426020408163265e-05,
+      "loss": 0.0878,
+      "step": 15102
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.424489795918366e-05,
+      "loss": 0.0859,
+      "step": 15103
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.422959183673469e-05,
+      "loss": 0.1789,
+      "step": 15104
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.42142857142857e-05,
+      "loss": 0.0496,
+      "step": 15105
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.419897959183673e-05,
+      "loss": 0.0425,
+      "step": 15106
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.418367346938774e-05,
+      "loss": 0.0757,
+      "step": 15107
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.416836734693877e-05,
+      "loss": 0.1311,
+      "step": 15108
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.415306122448978e-05,
+      "loss": 0.0405,
+      "step": 15109
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.413775510204082e-05,
+      "loss": 0.1584,
+      "step": 15110
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.412244897959182e-05,
+      "loss": 0.0743,
+      "step": 15111
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.410714285714285e-05,
+      "loss": 0.0739,
+      "step": 15112
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.409183673469387e-05,
+      "loss": 0.0392,
+      "step": 15113
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.40765306122449e-05,
+      "loss": 0.18,
+      "step": 15114
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.40612244897959e-05,
+      "loss": 0.124,
+      "step": 15115
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.404591836734693e-05,
+      "loss": 0.2739,
+      "step": 15116
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.403061224489795e-05,
+      "loss": 0.1448,
+      "step": 15117
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.401530612244898e-05,
+      "loss": 0.1169,
+      "step": 15118
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.4e-05,
+      "loss": 0.0772,
+      "step": 15119
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.398469387755101e-05,
+      "loss": 0.1516,
+      "step": 15120
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.396938775510204e-05,
+      "loss": 0.0967,
+      "step": 15121
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.395408163265305e-05,
+      "loss": 0.1165,
+      "step": 15122
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.393877551020408e-05,
+      "loss": 0.0404,
+      "step": 15123
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.39234693877551e-05,
+      "loss": 0.0819,
+      "step": 15124
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.390816326530612e-05,
+      "loss": 0.0314,
+      "step": 15125
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.389285714285713e-05,
+      "loss": 0.0851,
+      "step": 15126
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.387755102040816e-05,
+      "loss": 0.1448,
+      "step": 15127
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.386224489795918e-05,
+      "loss": 0.0768,
+      "step": 15128
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.38469387755102e-05,
+      "loss": 0.0367,
+      "step": 15129
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.383163265306121e-05,
+      "loss": 0.0715,
+      "step": 15130
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.381632653061224e-05,
+      "loss": 0.0823,
+      "step": 15131
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.380102040816326e-05,
+      "loss": 0.0979,
+      "step": 15132
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.378571428571429e-05,
+      "loss": 0.0552,
+      "step": 15133
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.377040816326529e-05,
+      "loss": 0.09,
+      "step": 15134
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.375510204081632e-05,
+      "loss": 0.0345,
+      "step": 15135
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.373979591836734e-05,
+      "loss": 0.0795,
+      "step": 15136
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.372448979591837e-05,
+      "loss": 0.1234,
+      "step": 15137
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.370918367346937e-05,
+      "loss": 0.0341,
+      "step": 15138
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.36938775510204e-05,
+      "loss": 0.1291,
+      "step": 15139
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.367857142857142e-05,
+      "loss": 0.049,
+      "step": 15140
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.366326530612245e-05,
+      "loss": 0.1194,
+      "step": 15141
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.364795918367345e-05,
+      "loss": 0.0564,
+      "step": 15142
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.363265306122448e-05,
+      "loss": 0.1083,
+      "step": 15143
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.36173469387755e-05,
+      "loss": 0.0751,
+      "step": 15144
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.360204081632653e-05,
+      "loss": 0.1808,
+      "step": 15145
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 8.358673469387754e-05,
+      "loss": 0.1242,
+      "step": 15146
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.357142857142857e-05,
+      "loss": 0.0716,
+      "step": 15147
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.355612244897958e-05,
+      "loss": 0.0812,
+      "step": 15148
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.354081632653061e-05,
+      "loss": 0.0423,
+      "step": 15149
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.352551020408162e-05,
+      "loss": 0.0294,
+      "step": 15150
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.351020408163265e-05,
+      "loss": 0.0762,
+      "step": 15151
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.349489795918366e-05,
+      "loss": 0.1211,
+      "step": 15152
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.347959183673468e-05,
+      "loss": 0.1611,
+      "step": 15153
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.34642857142857e-05,
+      "loss": 0.0699,
+      "step": 15154
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.344897959183673e-05,
+      "loss": 0.0721,
+      "step": 15155
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.343367346938775e-05,
+      "loss": 0.1081,
+      "step": 15156
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.341836734693876e-05,
+      "loss": 0.1055,
+      "step": 15157
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.340306122448978e-05,
+      "loss": 0.0844,
+      "step": 15158
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.338775510204081e-05,
+      "loss": 0.0444,
+      "step": 15159
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.337244897959183e-05,
+      "loss": 0.0864,
+      "step": 15160
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.335714285714284e-05,
+      "loss": 0.0421,
+      "step": 15161
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.334183673469386e-05,
+      "loss": 0.0541,
+      "step": 15162
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.332653061224489e-05,
+      "loss": 0.0903,
+      "step": 15163
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.331122448979592e-05,
+      "loss": 0.2449,
+      "step": 15164
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.329591836734693e-05,
+      "loss": 0.1276,
+      "step": 15165
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.328061224489796e-05,
+      "loss": 0.0818,
+      "step": 15166
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.326530612244897e-05,
+      "loss": 0.0774,
+      "step": 15167
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.325e-05,
+      "loss": 0.0272,
+      "step": 15168
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.323469387755101e-05,
+      "loss": 0.1069,
+      "step": 15169
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.321938775510204e-05,
+      "loss": 0.1063,
+      "step": 15170
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.320408163265306e-05,
+      "loss": 0.0485,
+      "step": 15171
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.318877551020409e-05,
+      "loss": 0.0647,
+      "step": 15172
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.317346938775509e-05,
+      "loss": 0.082,
+      "step": 15173
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.315816326530612e-05,
+      "loss": 0.0448,
+      "step": 15174
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.314285714285714e-05,
+      "loss": 0.0836,
+      "step": 15175
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.312755102040817e-05,
+      "loss": 0.0533,
+      "step": 15176
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.311224489795917e-05,
+      "loss": 0.0845,
+      "step": 15177
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.30969387755102e-05,
+      "loss": 0.0411,
+      "step": 15178
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.308163265306122e-05,
+      "loss": 0.0387,
+      "step": 15179
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.306632653061225e-05,
+      "loss": 0.08,
+      "step": 15180
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.305102040816325e-05,
+      "loss": 0.111,
+      "step": 15181
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.303571428571428e-05,
+      "loss": 0.0351,
+      "step": 15182
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.30204081632653e-05,
+      "loss": 0.1755,
+      "step": 15183
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.300510204081633e-05,
+      "loss": 0.0488,
+      "step": 15184
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.298979591836733e-05,
+      "loss": 0.0386,
+      "step": 15185
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.297448979591836e-05,
+      "loss": 0.0825,
+      "step": 15186
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.295918367346938e-05,
+      "loss": 0.0523,
+      "step": 15187
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.29438775510204e-05,
+      "loss": 0.0405,
+      "step": 15188
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.292857142857142e-05,
+      "loss": 0.2552,
+      "step": 15189
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.291326530612245e-05,
+      "loss": 0.1016,
+      "step": 15190
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.289795918367346e-05,
+      "loss": 0.039,
+      "step": 15191
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.288265306122448e-05,
+      "loss": 0.1057,
+      "step": 15192
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.28673469387755e-05,
+      "loss": 0.0916,
+      "step": 15193
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.285204081632653e-05,
+      "loss": 0.1029,
+      "step": 15194
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.283673469387754e-05,
+      "loss": 0.0976,
+      "step": 15195
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.282142857142856e-05,
+      "loss": 0.0524,
+      "step": 15196
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 8.280612244897958e-05,
+      "loss": 0.1169,
+      "step": 15197
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.279081632653061e-05,
+      "loss": 0.0878,
+      "step": 15198
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.277551020408161e-05,
+      "loss": 0.2557,
+      "step": 15199
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.276020408163264e-05,
+      "loss": 0.0783,
+      "step": 15200
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.274489795918366e-05,
+      "loss": 0.0696,
+      "step": 15201
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.272959183673469e-05,
+      "loss": 0.118,
+      "step": 15202
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.27142857142857e-05,
+      "loss": 0.1192,
+      "step": 15203
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.269897959183672e-05,
+      "loss": 0.0322,
+      "step": 15204
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.268367346938774e-05,
+      "loss": 0.0775,
+      "step": 15205
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.266836734693877e-05,
+      "loss": 0.0904,
+      "step": 15206
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.265306122448978e-05,
+      "loss": 0.0778,
+      "step": 15207
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.26377551020408e-05,
+      "loss": 0.3382,
+      "step": 15208
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.262244897959184e-05,
+      "loss": 0.0503,
+      "step": 15209
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.260714285714285e-05,
+      "loss": 0.0426,
+      "step": 15210
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.259183673469388e-05,
+      "loss": 0.128,
+      "step": 15211
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.257653061224489e-05,
+      "loss": 0.0855,
+      "step": 15212
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.256122448979592e-05,
+      "loss": 0.0694,
+      "step": 15213
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.254591836734694e-05,
+      "loss": 0.0889,
+      "step": 15214
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.253061224489797e-05,
+      "loss": 0.1144,
+      "step": 15215
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.251530612244897e-05,
+      "loss": 0.1124,
+      "step": 15216
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.25e-05,
+      "loss": 0.1187,
+      "step": 15217
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.248469387755102e-05,
+      "loss": 0.0482,
+      "step": 15218
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.246938775510203e-05,
+      "loss": 0.0559,
+      "step": 15219
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.245408163265305e-05,
+      "loss": 0.1385,
+      "step": 15220
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.243877551020408e-05,
+      "loss": 0.1792,
+      "step": 15221
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.24234693877551e-05,
+      "loss": 0.0452,
+      "step": 15222
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.240816326530612e-05,
+      "loss": 0.0835,
+      "step": 15223
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.239285714285713e-05,
+      "loss": 0.1439,
+      "step": 15224
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.237755102040816e-05,
+      "loss": 0.0948,
+      "step": 15225
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.236224489795918e-05,
+      "loss": 0.1611,
+      "step": 15226
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.23469387755102e-05,
+      "loss": 0.1427,
+      "step": 15227
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.233163265306121e-05,
+      "loss": 0.05,
+      "step": 15228
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.231632653061224e-05,
+      "loss": 0.1404,
+      "step": 15229
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.230102040816325e-05,
+      "loss": 0.033,
+      "step": 15230
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.228571428571428e-05,
+      "loss": 0.0395,
+      "step": 15231
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.22704081632653e-05,
+      "loss": 0.0609,
+      "step": 15232
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.225510204081633e-05,
+      "loss": 0.2062,
+      "step": 15233
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.223979591836733e-05,
+      "loss": 0.092,
+      "step": 15234
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.222448979591836e-05,
+      "loss": 0.0441,
+      "step": 15235
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.220918367346938e-05,
+      "loss": 0.0752,
+      "step": 15236
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.219387755102041e-05,
+      "loss": 0.0784,
+      "step": 15237
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 8.217857142857141e-05,
+      "loss": 0.0666,
+      "step": 15238
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.216326530612244e-05,
+      "loss": 0.1692,
+      "step": 15239
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.214795918367346e-05,
+      "loss": 0.1296,
+      "step": 15240
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.213265306122449e-05,
+      "loss": 0.0474,
+      "step": 15241
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.211734693877549e-05,
+      "loss": 0.0386,
+      "step": 15242
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.210204081632652e-05,
+      "loss": 0.1149,
+      "step": 15243
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.208673469387754e-05,
+      "loss": 0.0575,
+      "step": 15244
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.207142857142857e-05,
+      "loss": 0.1333,
+      "step": 15245
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.205612244897957e-05,
+      "loss": 0.1079,
+      "step": 15246
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.20408163265306e-05,
+      "loss": 0.0759,
+      "step": 15247
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.202551020408162e-05,
+      "loss": 0.0859,
+      "step": 15248
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.201020408163265e-05,
+      "loss": 0.0364,
+      "step": 15249
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.199489795918366e-05,
+      "loss": 0.0667,
+      "step": 15250
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.197959183673469e-05,
+      "loss": 0.0746,
+      "step": 15251
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.19642857142857e-05,
+      "loss": 0.0642,
+      "step": 15252
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.194897959183673e-05,
+      "loss": 0.0761,
+      "step": 15253
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.193367346938775e-05,
+      "loss": 0.042,
+      "step": 15254
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.191836734693877e-05,
+      "loss": 0.0411,
+      "step": 15255
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.19030612244898e-05,
+      "loss": 0.0798,
+      "step": 15256
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.188775510204081e-05,
+      "loss": 0.1435,
+      "step": 15257
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.187244897959183e-05,
+      "loss": 0.1402,
+      "step": 15258
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.185714285714285e-05,
+      "loss": 0.115,
+      "step": 15259
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.184183673469388e-05,
+      "loss": 0.0944,
+      "step": 15260
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.182653061224488e-05,
+      "loss": 0.0329,
+      "step": 15261
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.181122448979591e-05,
+      "loss": 0.1201,
+      "step": 15262
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.179591836734693e-05,
+      "loss": 0.0265,
+      "step": 15263
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.178061224489796e-05,
+      "loss": 0.1436,
+      "step": 15264
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.176530612244896e-05,
+      "loss": 0.1331,
+      "step": 15265
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.175e-05,
+      "loss": 0.162,
+      "step": 15266
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.173469387755101e-05,
+      "loss": 0.0718,
+      "step": 15267
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.171938775510204e-05,
+      "loss": 0.1239,
+      "step": 15268
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 8.170408163265305e-05,
+      "loss": 0.097,
+      "step": 15269
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.168877551020408e-05,
+      "loss": 0.1656,
+      "step": 15270
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.16734693877551e-05,
+      "loss": 0.1263,
+      "step": 15271
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.165816326530612e-05,
+      "loss": 0.092,
+      "step": 15272
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.164285714285713e-05,
+      "loss": 0.1047,
+      "step": 15273
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.162755102040816e-05,
+      "loss": 0.1134,
+      "step": 15274
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.161224489795917e-05,
+      "loss": 0.1526,
+      "step": 15275
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.15969387755102e-05,
+      "loss": 0.0725,
+      "step": 15276
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.158163265306121e-05,
+      "loss": 0.0589,
+      "step": 15277
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.156632653061224e-05,
+      "loss": 0.0916,
+      "step": 15278
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.155102040816326e-05,
+      "loss": 0.1157,
+      "step": 15279
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 8.153571428571429e-05,
+      "loss": 0.0961,
+      "step": 15280
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.152040816326529e-05,
+      "loss": 0.059,
+      "step": 15281
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.150510204081632e-05,
+      "loss": 0.0364,
+      "step": 15282
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.148979591836734e-05,
+      "loss": 0.1312,
+      "step": 15283
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.147448979591837e-05,
+      "loss": 0.1706,
+      "step": 15284
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.145918367346937e-05,
+      "loss": 0.0341,
+      "step": 15285
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.14438775510204e-05,
+      "loss": 0.1791,
+      "step": 15286
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.142857142857142e-05,
+      "loss": 0.0416,
+      "step": 15287
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.141326530612245e-05,
+      "loss": 0.0619,
+      "step": 15288
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.139795918367345e-05,
+      "loss": 0.1523,
+      "step": 15289
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 8.138265306122448e-05,
+      "loss": 0.0473,
+      "step": 15290
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.13673469387755e-05,
+      "loss": 0.0849,
+      "step": 15291
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.135204081632652e-05,
+      "loss": 0.048,
+      "step": 15292
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.133673469387754e-05,
+      "loss": 0.0663,
+      "step": 15293
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.132142857142857e-05,
+      "loss": 0.0349,
+      "step": 15294
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.130612244897958e-05,
+      "loss": 0.0894,
+      "step": 15295
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.12908163265306e-05,
+      "loss": 0.1741,
+      "step": 15296
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.127551020408162e-05,
+      "loss": 0.0991,
+      "step": 15297
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.126020408163265e-05,
+      "loss": 0.0455,
+      "step": 15298
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.124489795918368e-05,
+      "loss": 0.0489,
+      "step": 15299
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 8.122959183673468e-05,
+      "loss": 0.0909,
+      "step": 15300
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.121428571428571e-05,
+      "loss": 0.1833,
+      "step": 15301
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.119897959183673e-05,
+      "loss": 0.1596,
+      "step": 15302
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.118367346938776e-05,
+      "loss": 0.1148,
+      "step": 15303
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.116836734693876e-05,
+      "loss": 0.1216,
+      "step": 15304
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.115306122448979e-05,
+      "loss": 0.1405,
+      "step": 15305
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.113775510204081e-05,
+      "loss": 0.1463,
+      "step": 15306
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.112244897959184e-05,
+      "loss": 0.157,
+      "step": 15307
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.110714285714284e-05,
+      "loss": 0.0868,
+      "step": 15308
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.109183673469387e-05,
+      "loss": 0.0409,
+      "step": 15309
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 8.107653061224489e-05,
+      "loss": 0.0365,
+      "step": 15310
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.106122448979592e-05,
+      "loss": 0.0788,
+      "step": 15311
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.104591836734693e-05,
+      "loss": 0.0856,
+      "step": 15312
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.103061224489796e-05,
+      "loss": 0.0862,
+      "step": 15313
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.101530612244897e-05,
+      "loss": 0.2346,
+      "step": 15314
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.1e-05,
+      "loss": 0.0495,
+      "step": 15315
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.098469387755101e-05,
+      "loss": 0.0333,
+      "step": 15316
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.096938775510204e-05,
+      "loss": 0.0368,
+      "step": 15317
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.095408163265305e-05,
+      "loss": 0.1209,
+      "step": 15318
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.093877551020409e-05,
+      "loss": 0.0645,
+      "step": 15319
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.092346938775509e-05,
+      "loss": 0.1069,
+      "step": 15320
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 8.090816326530612e-05,
+      "loss": 0.0494,
+      "step": 15321
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.089285714285714e-05,
+      "loss": 0.1217,
+      "step": 15322
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.087755102040817e-05,
+      "loss": 0.0843,
+      "step": 15323
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.086224489795917e-05,
+      "loss": 0.0997,
+      "step": 15324
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.08469387755102e-05,
+      "loss": 0.0462,
+      "step": 15325
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.083163265306122e-05,
+      "loss": 0.133,
+      "step": 15326
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.081632653061223e-05,
+      "loss": 0.0613,
+      "step": 15327
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.080102040816325e-05,
+      "loss": 0.0849,
+      "step": 15328
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.078571428571428e-05,
+      "loss": 0.0479,
+      "step": 15329
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.07704081632653e-05,
+      "loss": 0.0755,
+      "step": 15330
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 8.075510204081632e-05,
+      "loss": 0.0548,
+      "step": 15331
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.073979591836733e-05,
+      "loss": 0.0958,
+      "step": 15332
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.072448979591836e-05,
+      "loss": 0.2552,
+      "step": 15333
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.070918367346938e-05,
+      "loss": 0.0941,
+      "step": 15334
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.06938775510204e-05,
+      "loss": 0.0427,
+      "step": 15335
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.067857142857141e-05,
+      "loss": 0.0623,
+      "step": 15336
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.066326530612245e-05,
+      "loss": 0.1014,
+      "step": 15337
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.064795918367345e-05,
+      "loss": 0.074,
+      "step": 15338
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.063265306122448e-05,
+      "loss": 0.1331,
+      "step": 15339
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.06173469387755e-05,
+      "loss": 0.1203,
+      "step": 15340
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 8.060204081632653e-05,
+      "loss": 0.0951,
+      "step": 15341
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.058673469387753e-05,
+      "loss": 0.1635,
+      "step": 15342
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.057142857142856e-05,
+      "loss": 0.0809,
+      "step": 15343
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.055612244897958e-05,
+      "loss": 0.0666,
+      "step": 15344
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.054081632653061e-05,
+      "loss": 0.251,
+      "step": 15345
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.052551020408164e-05,
+      "loss": 0.1384,
+      "step": 15346
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.051020408163264e-05,
+      "loss": 0.0445,
+      "step": 15347
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.049489795918367e-05,
+      "loss": 0.0341,
+      "step": 15348
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.047959183673469e-05,
+      "loss": 0.1728,
+      "step": 15349
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.046428571428572e-05,
+      "loss": 0.0676,
+      "step": 15350
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.044897959183672e-05,
+      "loss": 0.0847,
+      "step": 15351
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.043367346938775e-05,
+      "loss": 0.1718,
+      "step": 15352
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.041836734693877e-05,
+      "loss": 0.168,
+      "step": 15353
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.04030612244898e-05,
+      "loss": 0.1129,
+      "step": 15354
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.03877551020408e-05,
+      "loss": 0.0753,
+      "step": 15355
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.037244897959184e-05,
+      "loss": 0.1082,
+      "step": 15356
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.035714285714285e-05,
+      "loss": 0.1037,
+      "step": 15357
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.034183673469387e-05,
+      "loss": 0.2011,
+      "step": 15358
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.032653061224489e-05,
+      "loss": 0.0453,
+      "step": 15359
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.031122448979592e-05,
+      "loss": 0.1049,
+      "step": 15360
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.029591836734693e-05,
+      "loss": 0.0552,
+      "step": 15361
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 8.028061224489795e-05,
+      "loss": 0.1494,
+      "step": 15362
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.026530612244897e-05,
+      "loss": 0.033,
+      "step": 15363
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.025e-05,
+      "loss": 0.1379,
+      "step": 15364
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.023469387755102e-05,
+      "loss": 0.078,
+      "step": 15365
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.021938775510203e-05,
+      "loss": 0.082,
+      "step": 15366
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.020408163265305e-05,
+      "loss": 0.0704,
+      "step": 15367
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.018877551020408e-05,
+      "loss": 0.0703,
+      "step": 15368
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.017346938775508e-05,
+      "loss": 0.0483,
+      "step": 15369
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.015816326530611e-05,
+      "loss": 0.134,
+      "step": 15370
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.014285714285713e-05,
+      "loss": 0.1733,
+      "step": 15371
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 8.012755102040816e-05,
+      "loss": 0.1237,
+      "step": 15372
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.011224489795917e-05,
+      "loss": 0.1264,
+      "step": 15373
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.00969387755102e-05,
+      "loss": 0.0404,
+      "step": 15374
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.008163265306121e-05,
+      "loss": 0.0879,
+      "step": 15375
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.006632653061224e-05,
+      "loss": 0.0908,
+      "step": 15376
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.005102040816325e-05,
+      "loss": 0.2058,
+      "step": 15377
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.003571428571428e-05,
+      "loss": 0.1278,
+      "step": 15378
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.00204081632653e-05,
+      "loss": 0.0533,
+      "step": 15379
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.000510204081633e-05,
+      "loss": 0.0736,
+      "step": 15380
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 7.998979591836733e-05,
+      "loss": 0.1654,
+      "step": 15381
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 7.997448979591836e-05,
+      "loss": 0.0474,
+      "step": 15382
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 7.995918367346938e-05,
+      "loss": 0.1301,
+      "step": 15383
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.99438775510204e-05,
+      "loss": 0.2116,
+      "step": 15384
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.992857142857141e-05,
+      "loss": 0.122,
+      "step": 15385
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.991326530612244e-05,
+      "loss": 0.0595,
+      "step": 15386
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.989795918367346e-05,
+      "loss": 0.0405,
+      "step": 15387
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.988265306122449e-05,
+      "loss": 0.0328,
+      "step": 15388
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.986734693877549e-05,
+      "loss": 0.0824,
+      "step": 15389
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.985204081632652e-05,
+      "loss": 0.0444,
+      "step": 15390
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.983673469387755e-05,
+      "loss": 0.1115,
+      "step": 15391
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.982142857142857e-05,
+      "loss": 0.0753,
+      "step": 15392
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.980612244897959e-05,
+      "loss": 0.1187,
+      "step": 15393
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.97908163265306e-05,
+      "loss": 0.0858,
+      "step": 15394
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.977551020408163e-05,
+      "loss": 0.0512,
+      "step": 15395
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.976020408163265e-05,
+      "loss": 0.104,
+      "step": 15396
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.974489795918367e-05,
+      "loss": 0.0397,
+      "step": 15397
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.972959183673469e-05,
+      "loss": 0.0426,
+      "step": 15398
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.971428571428572e-05,
+      "loss": 0.1166,
+      "step": 15399
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.969897959183672e-05,
+      "loss": 0.0351,
+      "step": 15400
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.968367346938775e-05,
+      "loss": 0.0786,
+      "step": 15401
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.966836734693877e-05,
+      "loss": 0.0397,
+      "step": 15402
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.96530612244898e-05,
+      "loss": 0.0865,
+      "step": 15403
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.96377551020408e-05,
+      "loss": 0.0883,
+      "step": 15404
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.962244897959183e-05,
+      "loss": 0.1256,
+      "step": 15405
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.960714285714285e-05,
+      "loss": 0.0947,
+      "step": 15406
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.959183673469388e-05,
+      "loss": 0.1556,
+      "step": 15407
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.957653061224488e-05,
+      "loss": 0.2831,
+      "step": 15408
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.956122448979591e-05,
+      "loss": 0.0912,
+      "step": 15409
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.954591836734693e-05,
+      "loss": 0.0751,
+      "step": 15410
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.953061224489796e-05,
+      "loss": 0.0443,
+      "step": 15411
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.951530612244896e-05,
+      "loss": 0.0451,
+      "step": 15412
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.95e-05,
+      "loss": 0.0518,
+      "step": 15413
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.948469387755101e-05,
+      "loss": 0.1259,
+      "step": 15414
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.946938775510204e-05,
+      "loss": 0.1231,
+      "step": 15415
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.945408163265305e-05,
+      "loss": 0.0555,
+      "step": 15416
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.943877551020408e-05,
+      "loss": 0.0455,
+      "step": 15417
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.942346938775509e-05,
+      "loss": 0.1258,
+      "step": 15418
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.940816326530612e-05,
+      "loss": 0.0322,
+      "step": 15419
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.939285714285713e-05,
+      "loss": 0.1145,
+      "step": 15420
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.937755102040816e-05,
+      "loss": 0.1006,
+      "step": 15421
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.936224489795917e-05,
+      "loss": 0.1293,
+      "step": 15422
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.93469387755102e-05,
+      "loss": 0.1023,
+      "step": 15423
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.933163265306121e-05,
+      "loss": 0.0375,
+      "step": 15424
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.931632653061224e-05,
+      "loss": 0.0789,
+      "step": 15425
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.930102040816326e-05,
+      "loss": 0.0778,
+      "step": 15426
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.928571428571429e-05,
+      "loss": 0.1703,
+      "step": 15427
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.927040816326529e-05,
+      "loss": 0.0412,
+      "step": 15428
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.925510204081632e-05,
+      "loss": 0.0557,
+      "step": 15429
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.923979591836734e-05,
+      "loss": 0.0412,
+      "step": 15430
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.922448979591835e-05,
+      "loss": 0.1184,
+      "step": 15431
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.920918367346937e-05,
+      "loss": 0.0687,
+      "step": 15432
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.91938775510204e-05,
+      "loss": 0.1868,
+      "step": 15433
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 7.917857142857142e-05,
+      "loss": 0.0724,
+      "step": 15434
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.916326530612244e-05,
+      "loss": 0.084,
+      "step": 15435
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.914795918367347e-05,
+      "loss": 0.0757,
+      "step": 15436
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.913265306122448e-05,
+      "loss": 0.0815,
+      "step": 15437
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.911734693877551e-05,
+      "loss": 0.0535,
+      "step": 15438
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.910204081632652e-05,
+      "loss": 0.1442,
+      "step": 15439
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.908673469387755e-05,
+      "loss": 0.0429,
+      "step": 15440
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.907142857142857e-05,
+      "loss": 0.0459,
+      "step": 15441
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.90561224489796e-05,
+      "loss": 0.1596,
+      "step": 15442
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.90408163265306e-05,
+      "loss": 0.0754,
+      "step": 15443
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.902551020408163e-05,
+      "loss": 0.0502,
+      "step": 15444
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.901020408163265e-05,
+      "loss": 0.1138,
+      "step": 15445
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.899489795918368e-05,
+      "loss": 0.2413,
+      "step": 15446
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.897959183673468e-05,
+      "loss": 0.0857,
+      "step": 15447
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.896428571428571e-05,
+      "loss": 0.0876,
+      "step": 15448
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.894897959183673e-05,
+      "loss": 0.0819,
+      "step": 15449
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.893367346938776e-05,
+      "loss": 0.1188,
+      "step": 15450
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.891836734693876e-05,
+      "loss": 0.0897,
+      "step": 15451
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.890306122448979e-05,
+      "loss": 0.1136,
+      "step": 15452
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.888775510204081e-05,
+      "loss": 0.0384,
+      "step": 15453
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.887244897959184e-05,
+      "loss": 0.0793,
+      "step": 15454
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.885714285714284e-05,
+      "loss": 0.0327,
+      "step": 15455
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.884183673469387e-05,
+      "loss": 0.0831,
+      "step": 15456
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.882653061224489e-05,
+      "loss": 0.0937,
+      "step": 15457
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.881122448979592e-05,
+      "loss": 0.1277,
+      "step": 15458
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.879591836734693e-05,
+      "loss": 0.0783,
+      "step": 15459
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.878061224489796e-05,
+      "loss": 0.1463,
+      "step": 15460
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.876530612244897e-05,
+      "loss": 0.0722,
+      "step": 15461
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.874999999999999e-05,
+      "loss": 0.0563,
+      "step": 15462
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.8734693877551e-05,
+      "loss": 0.1343,
+      "step": 15463
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.871938775510204e-05,
+      "loss": 0.0761,
+      "step": 15464
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.870408163265305e-05,
+      "loss": 0.0831,
+      "step": 15465
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.868877551020407e-05,
+      "loss": 0.0324,
+      "step": 15466
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.867346938775509e-05,
+      "loss": 0.1162,
+      "step": 15467
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.865816326530612e-05,
+      "loss": 0.0273,
+      "step": 15468
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.864285714285714e-05,
+      "loss": 0.1128,
+      "step": 15469
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.862755102040815e-05,
+      "loss": 0.0786,
+      "step": 15470
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.861224489795917e-05,
+      "loss": 0.119,
+      "step": 15471
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.85969387755102e-05,
+      "loss": 0.0366,
+      "step": 15472
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.858163265306122e-05,
+      "loss": 0.081,
+      "step": 15473
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.856632653061223e-05,
+      "loss": 0.0455,
+      "step": 15474
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 7.855102040816325e-05,
+      "loss": 0.1506,
+      "step": 15475
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.853571428571428e-05,
+      "loss": 0.1238,
+      "step": 15476
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.852040816326529e-05,
+      "loss": 0.0423,
+      "step": 15477
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.850510204081632e-05,
+      "loss": 0.0822,
+      "step": 15478
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.848979591836733e-05,
+      "loss": 0.0273,
+      "step": 15479
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.847448979591836e-05,
+      "loss": 0.0621,
+      "step": 15480
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.84591836734694e-05,
+      "loss": 0.0325,
+      "step": 15481
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.84438775510204e-05,
+      "loss": 0.1031,
+      "step": 15482
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.842857142857143e-05,
+      "loss": 0.163,
+      "step": 15483
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.841326530612244e-05,
+      "loss": 0.0378,
+      "step": 15484
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.839795918367348e-05,
+      "loss": 0.0736,
+      "step": 15485
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.838265306122448e-05,
+      "loss": 0.055,
+      "step": 15486
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.836734693877551e-05,
+      "loss": 0.0465,
+      "step": 15487
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.835204081632653e-05,
+      "loss": 0.081,
+      "step": 15488
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.833673469387756e-05,
+      "loss": 0.0349,
+      "step": 15489
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.832142857142856e-05,
+      "loss": 0.1207,
+      "step": 15490
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.830612244897959e-05,
+      "loss": 0.0796,
+      "step": 15491
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.829081632653061e-05,
+      "loss": 0.0301,
+      "step": 15492
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.827551020408164e-05,
+      "loss": 0.0765,
+      "step": 15493
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.826020408163264e-05,
+      "loss": 0.1114,
+      "step": 15494
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.824489795918367e-05,
+      "loss": 0.0802,
+      "step": 15495
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.822959183673469e-05,
+      "loss": 0.0859,
+      "step": 15496
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.82142857142857e-05,
+      "loss": 0.0298,
+      "step": 15497
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.819897959183672e-05,
+      "loss": 0.0336,
+      "step": 15498
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.818367346938775e-05,
+      "loss": 0.1103,
+      "step": 15499
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.816836734693877e-05,
+      "loss": 0.0975,
+      "step": 15500
+    },
+    {
+      "epoch": 15.05,
+      "eval_cer": 0.24707328132792217,
+      "eval_loss": 1.1059867143630981,
+      "eval_runtime": 228.3545,
+      "eval_samples_per_second": 38.598,
+      "eval_steps_per_second": 4.826,
+      "eval_wer": 0.6455580089861472,
+      "step": 15500
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.815306122448979e-05,
+      "loss": 0.0293,
+      "step": 15501
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.81377551020408e-05,
+      "loss": 0.0433,
+      "step": 15502
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.812244897959184e-05,
+      "loss": 0.0814,
+      "step": 15503
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.810714285714285e-05,
+      "loss": 0.0946,
+      "step": 15504
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.809183673469387e-05,
+      "loss": 0.0715,
+      "step": 15505
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.807653061224489e-05,
+      "loss": 0.0959,
+      "step": 15506
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.806122448979592e-05,
+      "loss": 0.3032,
+      "step": 15507
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.804591836734692e-05,
+      "loss": 0.084,
+      "step": 15508
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.803061224489795e-05,
+      "loss": 0.0845,
+      "step": 15509
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.801530612244897e-05,
+      "loss": 0.0371,
+      "step": 15510
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.8e-05,
+      "loss": 0.0334,
+      "step": 15511
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.7984693877551e-05,
+      "loss": 0.0624,
+      "step": 15512
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.796938775510203e-05,
+      "loss": 0.0818,
+      "step": 15513
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.795408163265305e-05,
+      "loss": 0.0696,
+      "step": 15514
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.793877551020408e-05,
+      "loss": 0.073,
+      "step": 15515
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.792346938775508e-05,
+      "loss": 0.0441,
+      "step": 15516
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.790816326530611e-05,
+      "loss": 0.0729,
+      "step": 15517
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.789285714285713e-05,
+      "loss": 0.0728,
+      "step": 15518
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.787755102040816e-05,
+      "loss": 0.1682,
+      "step": 15519
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.786224489795917e-05,
+      "loss": 0.1243,
+      "step": 15520
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.78469387755102e-05,
+      "loss": 0.0877,
+      "step": 15521
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.783163265306121e-05,
+      "loss": 0.0723,
+      "step": 15522
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.781632653061224e-05,
+      "loss": 0.0832,
+      "step": 15523
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.780102040816325e-05,
+      "loss": 0.0509,
+      "step": 15524
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.778571428571428e-05,
+      "loss": 0.0968,
+      "step": 15525
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.777040816326531e-05,
+      "loss": 0.0969,
+      "step": 15526
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.775510204081632e-05,
+      "loss": 0.1868,
+      "step": 15527
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.773979591836734e-05,
+      "loss": 0.0334,
+      "step": 15528
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.772448979591836e-05,
+      "loss": 0.0404,
+      "step": 15529
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.770918367346939e-05,
+      "loss": 0.0267,
+      "step": 15530
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.76938775510204e-05,
+      "loss": 0.0968,
+      "step": 15531
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.767857142857142e-05,
+      "loss": 0.1674,
+      "step": 15532
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.766326530612244e-05,
+      "loss": 0.0721,
+      "step": 15533
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.764795918367347e-05,
+      "loss": 0.053,
+      "step": 15534
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.763265306122449e-05,
+      "loss": 0.0789,
+      "step": 15535
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.76173469387755e-05,
+      "loss": 0.0583,
+      "step": 15536
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.760204081632652e-05,
+      "loss": 0.0535,
+      "step": 15537
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.758673469387755e-05,
+      "loss": 0.1343,
+      "step": 15538
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.757142857142856e-05,
+      "loss": 0.0519,
+      "step": 15539
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.755612244897959e-05,
+      "loss": 0.0345,
+      "step": 15540
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.75408163265306e-05,
+      "loss": 0.0755,
+      "step": 15541
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.752551020408163e-05,
+      "loss": 0.1499,
+      "step": 15542
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.751020408163264e-05,
+      "loss": 0.0985,
+      "step": 15543
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.749489795918367e-05,
+      "loss": 0.1331,
+      "step": 15544
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.747959183673468e-05,
+      "loss": 0.0804,
+      "step": 15545
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.746428571428572e-05,
+      "loss": 0.0756,
+      "step": 15546
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.744897959183672e-05,
+      "loss": 0.044,
+      "step": 15547
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.743367346938775e-05,
+      "loss": 0.0768,
+      "step": 15548
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.741836734693877e-05,
+      "loss": 0.1337,
+      "step": 15549
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.74030612244898e-05,
+      "loss": 0.188,
+      "step": 15550
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.73877551020408e-05,
+      "loss": 0.1018,
+      "step": 15551
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.737244897959183e-05,
+      "loss": 0.1319,
+      "step": 15552
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.735714285714285e-05,
+      "loss": 0.037,
+      "step": 15553
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.734183673469388e-05,
+      "loss": 0.0367,
+      "step": 15554
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.732653061224488e-05,
+      "loss": 0.0578,
+      "step": 15555
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.731122448979591e-05,
+      "loss": 0.0454,
+      "step": 15556
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.729591836734693e-05,
+      "loss": 0.1462,
+      "step": 15557
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.728061224489796e-05,
+      "loss": 0.2132,
+      "step": 15558
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.726530612244896e-05,
+      "loss": 0.0424,
+      "step": 15559
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.725e-05,
+      "loss": 0.0397,
+      "step": 15560
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.723469387755101e-05,
+      "loss": 0.0369,
+      "step": 15561
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.721938775510204e-05,
+      "loss": 0.0422,
+      "step": 15562
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.720408163265304e-05,
+      "loss": 0.1231,
+      "step": 15563
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.718877551020408e-05,
+      "loss": 0.281,
+      "step": 15564
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.717346938775509e-05,
+      "loss": 0.1584,
+      "step": 15565
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.715816326530612e-05,
+      "loss": 0.0319,
+      "step": 15566
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.714285714285713e-05,
+      "loss": 0.0313,
+      "step": 15567
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.712755102040816e-05,
+      "loss": 0.0423,
+      "step": 15568
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.711224489795917e-05,
+      "loss": 0.178,
+      "step": 15569
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.709693877551019e-05,
+      "loss": 0.118,
+      "step": 15570
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.708163265306122e-05,
+      "loss": 0.0739,
+      "step": 15571
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.706632653061224e-05,
+      "loss": 0.109,
+      "step": 15572
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.705102040816327e-05,
+      "loss": 0.0793,
+      "step": 15573
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.703571428571427e-05,
+      "loss": 0.0684,
+      "step": 15574
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.70204081632653e-05,
+      "loss": 0.1032,
+      "step": 15575
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.700510204081632e-05,
+      "loss": 0.0898,
+      "step": 15576
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.698979591836735e-05,
+      "loss": 0.1286,
+      "step": 15577
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 7.697448979591835e-05,
+      "loss": 0.0384,
+      "step": 15578
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.695918367346938e-05,
+      "loss": 0.0441,
+      "step": 15579
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.69438775510204e-05,
+      "loss": 0.0757,
+      "step": 15580
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.692857142857143e-05,
+      "loss": 0.032,
+      "step": 15581
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.691326530612244e-05,
+      "loss": 0.0824,
+      "step": 15582
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.689795918367347e-05,
+      "loss": 0.0772,
+      "step": 15583
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.688265306122448e-05,
+      "loss": 0.0383,
+      "step": 15584
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.686734693877551e-05,
+      "loss": 0.078,
+      "step": 15585
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.685204081632652e-05,
+      "loss": 0.0397,
+      "step": 15586
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.683673469387755e-05,
+      "loss": 0.0809,
+      "step": 15587
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.682142857142856e-05,
+      "loss": 0.2937,
+      "step": 15588
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.68061224489796e-05,
+      "loss": 0.0398,
+      "step": 15589
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.67908163265306e-05,
+      "loss": 0.0873,
+      "step": 15590
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.677551020408163e-05,
+      "loss": 0.1101,
+      "step": 15591
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.676020408163265e-05,
+      "loss": 0.0446,
+      "step": 15592
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.674489795918368e-05,
+      "loss": 0.0394,
+      "step": 15593
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.672959183673468e-05,
+      "loss": 0.049,
+      "step": 15594
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.671428571428571e-05,
+      "loss": 0.0765,
+      "step": 15595
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.669897959183673e-05,
+      "loss": 0.0917,
+      "step": 15596
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.668367346938776e-05,
+      "loss": 0.0402,
+      "step": 15597
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.666836734693876e-05,
+      "loss": 0.1195,
+      "step": 15598
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.665306122448979e-05,
+      "loss": 0.0464,
+      "step": 15599
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.663775510204081e-05,
+      "loss": 0.0821,
+      "step": 15600
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.662244897959183e-05,
+      "loss": 0.0856,
+      "step": 15601
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.660714285714284e-05,
+      "loss": 0.067,
+      "step": 15602
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.659183673469387e-05,
+      "loss": 0.1166,
+      "step": 15603
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.657653061224489e-05,
+      "loss": 0.1117,
+      "step": 15604
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.656122448979591e-05,
+      "loss": 0.043,
+      "step": 15605
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.654591836734692e-05,
+      "loss": 0.0682,
+      "step": 15606
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.653061224489796e-05,
+      "loss": 0.0737,
+      "step": 15607
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.651530612244897e-05,
+      "loss": 0.1706,
+      "step": 15608
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.649999999999999e-05,
+      "loss": 0.0816,
+      "step": 15609
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.6484693877551e-05,
+      "loss": 0.039,
+      "step": 15610
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.646938775510204e-05,
+      "loss": 0.0817,
+      "step": 15611
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.645408163265304e-05,
+      "loss": 0.0696,
+      "step": 15612
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.643877551020407e-05,
+      "loss": 0.1366,
+      "step": 15613
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.642346938775509e-05,
+      "loss": 0.0854,
+      "step": 15614
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.640816326530612e-05,
+      "loss": 0.0822,
+      "step": 15615
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.639285714285712e-05,
+      "loss": 0.0406,
+      "step": 15616
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.637755102040815e-05,
+      "loss": 0.0311,
+      "step": 15617
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.636224489795918e-05,
+      "loss": 0.0793,
+      "step": 15618
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.63469387755102e-05,
+      "loss": 0.1057,
+      "step": 15619
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.633163265306123e-05,
+      "loss": 0.0555,
+      "step": 15620
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.631632653061223e-05,
+      "loss": 0.1486,
+      "step": 15621
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.630102040816326e-05,
+      "loss": 0.0361,
+      "step": 15622
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.628571428571428e-05,
+      "loss": 0.0795,
+      "step": 15623
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.627040816326531e-05,
+      "loss": 0.1281,
+      "step": 15624
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.625510204081632e-05,
+      "loss": 0.0308,
+      "step": 15625
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.623979591836735e-05,
+      "loss": 0.0508,
+      "step": 15626
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.622448979591836e-05,
+      "loss": 0.1288,
+      "step": 15627
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.62091836734694e-05,
+      "loss": 0.1005,
+      "step": 15628
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.61938775510204e-05,
+      "loss": 0.0721,
+      "step": 15629
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.617857142857143e-05,
+      "loss": 0.0699,
+      "step": 15630
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.616326530612244e-05,
+      "loss": 0.1765,
+      "step": 15631
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.614795918367346e-05,
+      "loss": 0.1925,
+      "step": 15632
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.613265306122448e-05,
+      "loss": 0.1976,
+      "step": 15633
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.611734693877551e-05,
+      "loss": 0.0798,
+      "step": 15634
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.610204081632653e-05,
+      "loss": 0.1328,
+      "step": 15635
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.608673469387754e-05,
+      "loss": 0.0509,
+      "step": 15636
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.607142857142856e-05,
+      "loss": 0.0255,
+      "step": 15637
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.605612244897959e-05,
+      "loss": 0.1294,
+      "step": 15638
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.604081632653061e-05,
+      "loss": 0.0809,
+      "step": 15639
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.602551020408162e-05,
+      "loss": 0.0721,
+      "step": 15640
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.601020408163264e-05,
+      "loss": 0.0484,
+      "step": 15641
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.599489795918367e-05,
+      "loss": 0.0306,
+      "step": 15642
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.597959183673469e-05,
+      "loss": 0.0404,
+      "step": 15643
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.59642857142857e-05,
+      "loss": 0.1739,
+      "step": 15644
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.594897959183672e-05,
+      "loss": 0.0657,
+      "step": 15645
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.593367346938775e-05,
+      "loss": 0.0353,
+      "step": 15646
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.591836734693876e-05,
+      "loss": 0.0835,
+      "step": 15647
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.590306122448979e-05,
+      "loss": 0.1558,
+      "step": 15648
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.58877551020408e-05,
+      "loss": 0.0547,
+      "step": 15649
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.587244897959184e-05,
+      "loss": 0.0968,
+      "step": 15650
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.585714285714284e-05,
+      "loss": 0.107,
+      "step": 15651
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.584183673469387e-05,
+      "loss": 0.1114,
+      "step": 15652
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.582653061224489e-05,
+      "loss": 0.0427,
+      "step": 15653
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.581122448979592e-05,
+      "loss": 0.14,
+      "step": 15654
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.579591836734692e-05,
+      "loss": 0.0783,
+      "step": 15655
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.578061224489795e-05,
+      "loss": 0.0369,
+      "step": 15656
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.576530612244897e-05,
+      "loss": 0.0943,
+      "step": 15657
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.575e-05,
+      "loss": 0.0508,
+      "step": 15658
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.5734693877551e-05,
+      "loss": 0.1252,
+      "step": 15659
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.571938775510203e-05,
+      "loss": 0.1159,
+      "step": 15660
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.570408163265305e-05,
+      "loss": 0.0331,
+      "step": 15661
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.568877551020408e-05,
+      "loss": 0.0432,
+      "step": 15662
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.567346938775511e-05,
+      "loss": 0.1875,
+      "step": 15663
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.565816326530611e-05,
+      "loss": 0.0442,
+      "step": 15664
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.564285714285714e-05,
+      "loss": 0.0394,
+      "step": 15665
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.562755102040816e-05,
+      "loss": 0.0752,
+      "step": 15666
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.561224489795918e-05,
+      "loss": 0.036,
+      "step": 15667
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.55969387755102e-05,
+      "loss": 0.1588,
+      "step": 15668
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.558163265306123e-05,
+      "loss": 0.1298,
+      "step": 15669
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.556632653061224e-05,
+      "loss": 0.0773,
+      "step": 15670
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 7.555102040816326e-05,
+      "loss": 0.0952,
+      "step": 15671
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.553571428571428e-05,
+      "loss": 0.0748,
+      "step": 15672
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.552040816326531e-05,
+      "loss": 0.0694,
+      "step": 15673
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.550510204081632e-05,
+      "loss": 0.0445,
+      "step": 15674
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.548979591836734e-05,
+      "loss": 0.1106,
+      "step": 15675
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.547448979591836e-05,
+      "loss": 0.0775,
+      "step": 15676
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.545918367346939e-05,
+      "loss": 0.0687,
+      "step": 15677
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.544387755102039e-05,
+      "loss": 0.1138,
+      "step": 15678
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.542857142857142e-05,
+      "loss": 0.079,
+      "step": 15679
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.541326530612244e-05,
+      "loss": 0.075,
+      "step": 15680
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.539795918367347e-05,
+      "loss": 0.0948,
+      "step": 15681
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.538265306122447e-05,
+      "loss": 0.149,
+      "step": 15682
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.53673469387755e-05,
+      "loss": 0.2034,
+      "step": 15683
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.535204081632652e-05,
+      "loss": 0.0339,
+      "step": 15684
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.533673469387755e-05,
+      "loss": 0.0804,
+      "step": 15685
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.532142857142856e-05,
+      "loss": 0.0897,
+      "step": 15686
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.530612244897959e-05,
+      "loss": 0.0981,
+      "step": 15687
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.52908163265306e-05,
+      "loss": 0.1163,
+      "step": 15688
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.527551020408163e-05,
+      "loss": 0.1732,
+      "step": 15689
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.526020408163264e-05,
+      "loss": 0.0782,
+      "step": 15690
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.524489795918367e-05,
+      "loss": 0.0439,
+      "step": 15691
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.522959183673468e-05,
+      "loss": 0.0657,
+      "step": 15692
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.521428571428571e-05,
+      "loss": 0.1122,
+      "step": 15693
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.519897959183672e-05,
+      "loss": 0.1333,
+      "step": 15694
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.518367346938775e-05,
+      "loss": 0.0526,
+      "step": 15695
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.516836734693877e-05,
+      "loss": 0.1693,
+      "step": 15696
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.51530612244898e-05,
+      "loss": 0.078,
+      "step": 15697
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.51377551020408e-05,
+      "loss": 0.1012,
+      "step": 15698
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.512244897959183e-05,
+      "loss": 0.067,
+      "step": 15699
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.510714285714285e-05,
+      "loss": 0.1339,
+      "step": 15700
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.509183673469388e-05,
+      "loss": 0.1304,
+      "step": 15701
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.507653061224488e-05,
+      "loss": 0.0826,
+      "step": 15702
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.506122448979591e-05,
+      "loss": 0.0448,
+      "step": 15703
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.504591836734693e-05,
+      "loss": 0.1162,
+      "step": 15704
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.503061224489796e-05,
+      "loss": 0.1328,
+      "step": 15705
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.501530612244896e-05,
+      "loss": 0.1181,
+      "step": 15706
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.5e-05,
+      "loss": 0.1204,
+      "step": 15707
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.498469387755101e-05,
+      "loss": 0.0443,
+      "step": 15708
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.496938775510203e-05,
+      "loss": 0.0324,
+      "step": 15709
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.495408163265306e-05,
+      "loss": 0.0319,
+      "step": 15710
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.493877551020407e-05,
+      "loss": 0.0441,
+      "step": 15711
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.492346938775509e-05,
+      "loss": 0.1001,
+      "step": 15712
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.490816326530611e-05,
+      "loss": 0.0884,
+      "step": 15713
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.489285714285714e-05,
+      "loss": 0.0457,
+      "step": 15714
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.487755102040816e-05,
+      "loss": 0.0441,
+      "step": 15715
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.486224489795917e-05,
+      "loss": 0.0735,
+      "step": 15716
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.484693877551019e-05,
+      "loss": 0.1172,
+      "step": 15717
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.483163265306122e-05,
+      "loss": 0.0502,
+      "step": 15718
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.481632653061224e-05,
+      "loss": 0.1758,
+      "step": 15719
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.480102040816326e-05,
+      "loss": 0.0389,
+      "step": 15720
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.478571428571427e-05,
+      "loss": 0.0767,
+      "step": 15721
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 7.47704081632653e-05,
+      "loss": 0.033,
+      "step": 15722
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.475510204081632e-05,
+      "loss": 0.0543,
+      "step": 15723
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.473979591836734e-05,
+      "loss": 0.0505,
+      "step": 15724
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.472448979591835e-05,
+      "loss": 0.3041,
+      "step": 15725
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.470918367346938e-05,
+      "loss": 0.1318,
+      "step": 15726
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.46938775510204e-05,
+      "loss": 0.0371,
+      "step": 15727
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.467857142857142e-05,
+      "loss": 0.0253,
+      "step": 15728
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.466326530612244e-05,
+      "loss": 0.0665,
+      "step": 15729
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.464795918367347e-05,
+      "loss": 0.0352,
+      "step": 15730
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.463265306122448e-05,
+      "loss": 0.0614,
+      "step": 15731
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.461734693877551e-05,
+      "loss": 0.111,
+      "step": 15732
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.460204081632653e-05,
+      "loss": 0.0409,
+      "step": 15733
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.458673469387755e-05,
+      "loss": 0.0926,
+      "step": 15734
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.457142857142856e-05,
+      "loss": 0.0832,
+      "step": 15735
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.45561224489796e-05,
+      "loss": 0.0377,
+      "step": 15736
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.454081632653061e-05,
+      "loss": 0.0979,
+      "step": 15737
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.452551020408163e-05,
+      "loss": 0.1553,
+      "step": 15738
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.451020408163265e-05,
+      "loss": 0.0483,
+      "step": 15739
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.449489795918366e-05,
+      "loss": 0.0857,
+      "step": 15740
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.447959183673469e-05,
+      "loss": 0.0334,
+      "step": 15741
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.446428571428571e-05,
+      "loss": 0.081,
+      "step": 15742
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.444897959183673e-05,
+      "loss": 0.066,
+      "step": 15743
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.443367346938774e-05,
+      "loss": 0.2048,
+      "step": 15744
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.441836734693877e-05,
+      "loss": 0.0794,
+      "step": 15745
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.440306122448979e-05,
+      "loss": 0.0759,
+      "step": 15746
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.438775510204081e-05,
+      "loss": 0.0753,
+      "step": 15747
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.437244897959183e-05,
+      "loss": 0.1017,
+      "step": 15748
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.435714285714286e-05,
+      "loss": 0.0588,
+      "step": 15749
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.434183673469387e-05,
+      "loss": 0.081,
+      "step": 15750
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.432653061224489e-05,
+      "loss": 0.1183,
+      "step": 15751
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.431122448979591e-05,
+      "loss": 0.0415,
+      "step": 15752
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.429591836734694e-05,
+      "loss": 0.0677,
+      "step": 15753
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.428061224489795e-05,
+      "loss": 0.0792,
+      "step": 15754
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.426530612244897e-05,
+      "loss": 0.1125,
+      "step": 15755
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.424999999999999e-05,
+      "loss": 0.0485,
+      "step": 15756
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.423469387755102e-05,
+      "loss": 0.0813,
+      "step": 15757
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.421938775510204e-05,
+      "loss": 0.0902,
+      "step": 15758
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.420408163265305e-05,
+      "loss": 0.0411,
+      "step": 15759
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.418877551020407e-05,
+      "loss": 0.1117,
+      "step": 15760
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.417346938775509e-05,
+      "loss": 0.0365,
+      "step": 15761
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.415816326530612e-05,
+      "loss": 0.0314,
+      "step": 15762
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.414285714285713e-05,
+      "loss": 0.0581,
+      "step": 15763
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 7.412755102040815e-05,
+      "loss": 0.1138,
+      "step": 15764
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.411224489795917e-05,
+      "loss": 0.0757,
+      "step": 15765
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.40969387755102e-05,
+      "loss": 0.0481,
+      "step": 15766
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.408163265306122e-05,
+      "loss": 0.0396,
+      "step": 15767
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.406632653061223e-05,
+      "loss": 0.0348,
+      "step": 15768
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.405102040816325e-05,
+      "loss": 0.1007,
+      "step": 15769
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.403571428571428e-05,
+      "loss": 0.082,
+      "step": 15770
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.40204081632653e-05,
+      "loss": 0.0389,
+      "step": 15771
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.400510204081631e-05,
+      "loss": 0.0387,
+      "step": 15772
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.398979591836733e-05,
+      "loss": 0.1035,
+      "step": 15773
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.397448979591836e-05,
+      "loss": 0.0531,
+      "step": 15774
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.395918367346938e-05,
+      "loss": 0.1307,
+      "step": 15775
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.394387755102041e-05,
+      "loss": 0.1128,
+      "step": 15776
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.392857142857143e-05,
+      "loss": 0.0323,
+      "step": 15777
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.391326530612244e-05,
+      "loss": 0.0352,
+      "step": 15778
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.389795918367346e-05,
+      "loss": 0.205,
+      "step": 15779
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.388265306122449e-05,
+      "loss": 0.1566,
+      "step": 15780
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.386734693877551e-05,
+      "loss": 0.1502,
+      "step": 15781
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.385204081632653e-05,
+      "loss": 0.2435,
+      "step": 15782
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.383673469387754e-05,
+      "loss": 0.0343,
+      "step": 15783
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.382142857142857e-05,
+      "loss": 0.0334,
+      "step": 15784
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.380612244897959e-05,
+      "loss": 0.1901,
+      "step": 15785
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.379081632653061e-05,
+      "loss": 0.0835,
+      "step": 15786
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.377551020408162e-05,
+      "loss": 0.0345,
+      "step": 15787
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.376020408163265e-05,
+      "loss": 0.1484,
+      "step": 15788
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.374489795918367e-05,
+      "loss": 0.091,
+      "step": 15789
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.372959183673469e-05,
+      "loss": 0.112,
+      "step": 15790
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.37142857142857e-05,
+      "loss": 0.1104,
+      "step": 15791
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.369897959183674e-05,
+      "loss": 0.1396,
+      "step": 15792
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.368367346938775e-05,
+      "loss": 0.0263,
+      "step": 15793
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.366836734693877e-05,
+      "loss": 0.1011,
+      "step": 15794
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.365306122448979e-05,
+      "loss": 0.0418,
+      "step": 15795
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.36377551020408e-05,
+      "loss": 0.1242,
+      "step": 15796
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.362244897959183e-05,
+      "loss": 0.0371,
+      "step": 15797
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.360714285714285e-05,
+      "loss": 0.0785,
+      "step": 15798
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.359183673469387e-05,
+      "loss": 0.0912,
+      "step": 15799
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.357653061224489e-05,
+      "loss": 0.0494,
+      "step": 15800
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.356122448979592e-05,
+      "loss": 0.1718,
+      "step": 15801
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.354591836734693e-05,
+      "loss": 0.0873,
+      "step": 15802
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.353061224489795e-05,
+      "loss": 0.0448,
+      "step": 15803
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.351530612244897e-05,
+      "loss": 0.0917,
+      "step": 15804
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.35e-05,
+      "loss": 0.1086,
+      "step": 15805
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.348469387755101e-05,
+      "loss": 0.0805,
+      "step": 15806
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.346938775510203e-05,
+      "loss": 0.1398,
+      "step": 15807
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.345408163265305e-05,
+      "loss": 0.0762,
+      "step": 15808
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.343877551020408e-05,
+      "loss": 0.0426,
+      "step": 15809
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.34234693877551e-05,
+      "loss": 0.0818,
+      "step": 15810
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.340816326530611e-05,
+      "loss": 0.0839,
+      "step": 15811
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.339285714285713e-05,
+      "loss": 0.101,
+      "step": 15812
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.337755102040816e-05,
+      "loss": 0.1425,
+      "step": 15813
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.336224489795918e-05,
+      "loss": 0.0859,
+      "step": 15814
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.33469387755102e-05,
+      "loss": 0.0749,
+      "step": 15815
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.333163265306121e-05,
+      "loss": 0.0314,
+      "step": 15816
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.331632653061223e-05,
+      "loss": 0.0829,
+      "step": 15817
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.330102040816326e-05,
+      "loss": 0.0549,
+      "step": 15818
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.328571428571428e-05,
+      "loss": 0.1651,
+      "step": 15819
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.327040816326529e-05,
+      "loss": 0.0518,
+      "step": 15820
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.325510204081632e-05,
+      "loss": 0.1131,
+      "step": 15821
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.323979591836734e-05,
+      "loss": 0.0256,
+      "step": 15822
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.322448979591837e-05,
+      "loss": 0.0399,
+      "step": 15823
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.320918367346939e-05,
+      "loss": 0.062,
+      "step": 15824
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.31938775510204e-05,
+      "loss": 0.1034,
+      "step": 15825
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.317857142857142e-05,
+      "loss": 0.1268,
+      "step": 15826
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.316326530612244e-05,
+      "loss": 0.0818,
+      "step": 15827
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.314795918367347e-05,
+      "loss": 0.0742,
+      "step": 15828
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.313265306122449e-05,
+      "loss": 0.0869,
+      "step": 15829
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.31173469387755e-05,
+      "loss": 0.0906,
+      "step": 15830
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.310204081632652e-05,
+      "loss": 0.0896,
+      "step": 15831
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.308673469387755e-05,
+      "loss": 0.0684,
+      "step": 15832
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.307142857142857e-05,
+      "loss": 0.1188,
+      "step": 15833
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.305612244897959e-05,
+      "loss": 0.0409,
+      "step": 15834
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.30408163265306e-05,
+      "loss": 0.0343,
+      "step": 15835
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.302551020408163e-05,
+      "loss": 0.0301,
+      "step": 15836
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.301020408163265e-05,
+      "loss": 0.0393,
+      "step": 15837
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.299489795918367e-05,
+      "loss": 0.1405,
+      "step": 15838
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.297959183673468e-05,
+      "loss": 0.0902,
+      "step": 15839
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.296428571428571e-05,
+      "loss": 0.0403,
+      "step": 15840
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.294897959183673e-05,
+      "loss": 0.0608,
+      "step": 15841
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.293367346938775e-05,
+      "loss": 0.0745,
+      "step": 15842
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.291836734693877e-05,
+      "loss": 0.0811,
+      "step": 15843
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.29030612244898e-05,
+      "loss": 0.1115,
+      "step": 15844
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.288775510204081e-05,
+      "loss": 0.1199,
+      "step": 15845
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.287244897959183e-05,
+      "loss": 0.0831,
+      "step": 15846
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.285714285714285e-05,
+      "loss": 0.0433,
+      "step": 15847
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.284183673469386e-05,
+      "loss": 0.1195,
+      "step": 15848
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.28265306122449e-05,
+      "loss": 0.0848,
+      "step": 15849
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.281122448979591e-05,
+      "loss": 0.0684,
+      "step": 15850
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.279591836734693e-05,
+      "loss": 0.1504,
+      "step": 15851
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.278061224489795e-05,
+      "loss": 0.0353,
+      "step": 15852
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.276530612244898e-05,
+      "loss": 0.1163,
+      "step": 15853
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.274999999999999e-05,
+      "loss": 0.0919,
+      "step": 15854
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.273469387755101e-05,
+      "loss": 0.125,
+      "step": 15855
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 7.271938775510203e-05,
+      "loss": 0.1224,
+      "step": 15856
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.270408163265306e-05,
+      "loss": 0.1603,
+      "step": 15857
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.268877551020407e-05,
+      "loss": 0.098,
+      "step": 15858
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.267346938775509e-05,
+      "loss": 0.1153,
+      "step": 15859
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.265816326530611e-05,
+      "loss": 0.1249,
+      "step": 15860
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.264285714285714e-05,
+      "loss": 0.11,
+      "step": 15861
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.262755102040816e-05,
+      "loss": 0.0331,
+      "step": 15862
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.261224489795917e-05,
+      "loss": 0.0767,
+      "step": 15863
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.259693877551019e-05,
+      "loss": 0.0497,
+      "step": 15864
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.258163265306122e-05,
+      "loss": 0.1153,
+      "step": 15865
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.256632653061224e-05,
+      "loss": 0.08,
+      "step": 15866
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.255102040816327e-05,
+      "loss": 0.0409,
+      "step": 15867
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.253571428571429e-05,
+      "loss": 0.0551,
+      "step": 15868
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.25204081632653e-05,
+      "loss": 0.1896,
+      "step": 15869
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.250510204081632e-05,
+      "loss": 0.0878,
+      "step": 15870
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.248979591836735e-05,
+      "loss": 0.0848,
+      "step": 15871
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.247448979591837e-05,
+      "loss": 0.0613,
+      "step": 15872
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.245918367346938e-05,
+      "loss": 0.1602,
+      "step": 15873
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.24438775510204e-05,
+      "loss": 0.0401,
+      "step": 15874
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.242857142857143e-05,
+      "loss": 0.2331,
+      "step": 15875
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.241326530612245e-05,
+      "loss": 0.0397,
+      "step": 15876
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.239795918367347e-05,
+      "loss": 0.0852,
+      "step": 15877
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.238265306122448e-05,
+      "loss": 0.1207,
+      "step": 15878
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.23673469387755e-05,
+      "loss": 0.073,
+      "step": 15879
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.235204081632653e-05,
+      "loss": 0.0739,
+      "step": 15880
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.233673469387755e-05,
+      "loss": 0.0651,
+      "step": 15881
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.232142857142856e-05,
+      "loss": 0.219,
+      "step": 15882
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.230612244897958e-05,
+      "loss": 0.0697,
+      "step": 15883
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.229081632653061e-05,
+      "loss": 0.1065,
+      "step": 15884
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.227551020408163e-05,
+      "loss": 0.1087,
+      "step": 15885
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.226020408163265e-05,
+      "loss": 0.072,
+      "step": 15886
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.224489795918366e-05,
+      "loss": 0.039,
+      "step": 15887
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.222959183673469e-05,
+      "loss": 0.1301,
+      "step": 15888
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.221428571428571e-05,
+      "loss": 0.0368,
+      "step": 15889
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.219897959183673e-05,
+      "loss": 0.1174,
+      "step": 15890
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.218367346938774e-05,
+      "loss": 0.0332,
+      "step": 15891
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.216836734693877e-05,
+      "loss": 0.1031,
+      "step": 15892
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.215306122448979e-05,
+      "loss": 0.0291,
+      "step": 15893
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.213775510204081e-05,
+      "loss": 0.2625,
+      "step": 15894
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.212244897959183e-05,
+      "loss": 0.1287,
+      "step": 15895
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.210714285714286e-05,
+      "loss": 0.0812,
+      "step": 15896
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.209183673469387e-05,
+      "loss": 0.0892,
+      "step": 15897
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.207653061224489e-05,
+      "loss": 0.0523,
+      "step": 15898
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.20612244897959e-05,
+      "loss": 0.0893,
+      "step": 15899
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.204591836734692e-05,
+      "loss": 0.0352,
+      "step": 15900
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.203061224489795e-05,
+      "loss": 0.1683,
+      "step": 15901
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.201530612244897e-05,
+      "loss": 0.127,
+      "step": 15902
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.199999999999999e-05,
+      "loss": 0.0868,
+      "step": 15903
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.1984693877551e-05,
+      "loss": 0.0396,
+      "step": 15904
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.196938775510204e-05,
+      "loss": 0.1165,
+      "step": 15905
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.195408163265305e-05,
+      "loss": 0.0853,
+      "step": 15906
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.193877551020407e-05,
+      "loss": 0.0912,
+      "step": 15907
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 7.192346938775509e-05,
+      "loss": 0.0931,
+      "step": 15908
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.190816326530612e-05,
+      "loss": 0.0343,
+      "step": 15909
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.189285714285713e-05,
+      "loss": 0.0757,
+      "step": 15910
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.187755102040815e-05,
+      "loss": 0.0808,
+      "step": 15911
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.186224489795918e-05,
+      "loss": 0.0936,
+      "step": 15912
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.18469387755102e-05,
+      "loss": 0.0903,
+      "step": 15913
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.183163265306122e-05,
+      "loss": 0.039,
+      "step": 15914
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.181632653061225e-05,
+      "loss": 0.1563,
+      "step": 15915
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.180102040816326e-05,
+      "loss": 0.0354,
+      "step": 15916
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.178571428571428e-05,
+      "loss": 0.0475,
+      "step": 15917
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.17704081632653e-05,
+      "loss": 0.0962,
+      "step": 15918
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.175510204081633e-05,
+      "loss": 0.073,
+      "step": 15919
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.173979591836734e-05,
+      "loss": 0.1182,
+      "step": 15920
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.172448979591836e-05,
+      "loss": 0.0852,
+      "step": 15921
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.170918367346938e-05,
+      "loss": 0.033,
+      "step": 15922
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.169387755102041e-05,
+      "loss": 0.0766,
+      "step": 15923
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.167857142857143e-05,
+      "loss": 0.065,
+      "step": 15924
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.166326530612244e-05,
+      "loss": 0.112,
+      "step": 15925
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.164795918367346e-05,
+      "loss": 0.0704,
+      "step": 15926
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.163265306122449e-05,
+      "loss": 0.0367,
+      "step": 15927
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.161734693877551e-05,
+      "loss": 0.0782,
+      "step": 15928
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.160204081632652e-05,
+      "loss": 0.0334,
+      "step": 15929
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.158673469387754e-05,
+      "loss": 0.1081,
+      "step": 15930
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.157142857142856e-05,
+      "loss": 0.1944,
+      "step": 15931
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.155612244897959e-05,
+      "loss": 0.1288,
+      "step": 15932
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.15408163265306e-05,
+      "loss": 0.0338,
+      "step": 15933
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.152551020408162e-05,
+      "loss": 0.149,
+      "step": 15934
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.151020408163264e-05,
+      "loss": 0.038,
+      "step": 15935
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.149489795918367e-05,
+      "loss": 0.037,
+      "step": 15936
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.147959183673469e-05,
+      "loss": 0.0542,
+      "step": 15937
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.14642857142857e-05,
+      "loss": 0.0697,
+      "step": 15938
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 7.144897959183672e-05,
+      "loss": 0.0527,
+      "step": 15939
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.143367346938775e-05,
+      "loss": 0.0699,
+      "step": 15940
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.141836734693877e-05,
+      "loss": 0.0907,
+      "step": 15941
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.140306122448979e-05,
+      "loss": 0.0271,
+      "step": 15942
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.13877551020408e-05,
+      "loss": 0.039,
+      "step": 15943
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.137244897959183e-05,
+      "loss": 0.1334,
+      "step": 15944
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.135714285714285e-05,
+      "loss": 0.1601,
+      "step": 15945
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.134183673469387e-05,
+      "loss": 0.0441,
+      "step": 15946
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.132653061224489e-05,
+      "loss": 0.0911,
+      "step": 15947
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.131122448979592e-05,
+      "loss": 0.0676,
+      "step": 15948
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.129591836734693e-05,
+      "loss": 0.0448,
+      "step": 15949
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.128061224489795e-05,
+      "loss": 0.1901,
+      "step": 15950
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.126530612244897e-05,
+      "loss": 0.1148,
+      "step": 15951
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.125e-05,
+      "loss": 0.111,
+      "step": 15952
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.123469387755101e-05,
+      "loss": 0.0413,
+      "step": 15953
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.121938775510203e-05,
+      "loss": 0.0234,
+      "step": 15954
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.120408163265305e-05,
+      "loss": 0.102,
+      "step": 15955
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.118877551020407e-05,
+      "loss": 0.0343,
+      "step": 15956
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.11734693877551e-05,
+      "loss": 0.1928,
+      "step": 15957
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.115816326530613e-05,
+      "loss": 0.0879,
+      "step": 15958
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 7.114285714285714e-05,
+      "loss": 0.0738,
+      "step": 15959
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.112755102040816e-05,
+      "loss": 0.1944,
+      "step": 15960
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.111224489795918e-05,
+      "loss": 0.0372,
+      "step": 15961
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.109693877551021e-05,
+      "loss": 0.0849,
+      "step": 15962
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.108163265306122e-05,
+      "loss": 0.0854,
+      "step": 15963
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.106632653061224e-05,
+      "loss": 0.085,
+      "step": 15964
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.105102040816326e-05,
+      "loss": 0.0395,
+      "step": 15965
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.103571428571428e-05,
+      "loss": 0.0309,
+      "step": 15966
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.10204081632653e-05,
+      "loss": 0.0232,
+      "step": 15967
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.100510204081632e-05,
+      "loss": 0.0404,
+      "step": 15968
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.098979591836734e-05,
+      "loss": 0.253,
+      "step": 15969
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 7.097448979591836e-05,
+      "loss": 0.0337,
+      "step": 15970
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.095918367346939e-05,
+      "loss": 0.2321,
+      "step": 15971
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.09438775510204e-05,
+      "loss": 0.0829,
+      "step": 15972
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.092857142857142e-05,
+      "loss": 0.0348,
+      "step": 15973
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.091326530612244e-05,
+      "loss": 0.0462,
+      "step": 15974
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.089795918367347e-05,
+      "loss": 0.1025,
+      "step": 15975
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.088265306122449e-05,
+      "loss": 0.1177,
+      "step": 15976
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.08673469387755e-05,
+      "loss": 0.077,
+      "step": 15977
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.085204081632652e-05,
+      "loss": 0.1014,
+      "step": 15978
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.083673469387755e-05,
+      "loss": 0.072,
+      "step": 15979
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 7.082142857142857e-05,
+      "loss": 0.0448,
+      "step": 15980
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.080612244897958e-05,
+      "loss": 0.1063,
+      "step": 15981
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.07908163265306e-05,
+      "loss": 0.1049,
+      "step": 15982
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.077551020408163e-05,
+      "loss": 0.1972,
+      "step": 15983
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.076020408163265e-05,
+      "loss": 0.1269,
+      "step": 15984
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.074489795918367e-05,
+      "loss": 0.0309,
+      "step": 15985
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.072959183673468e-05,
+      "loss": 0.0853,
+      "step": 15986
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.07142857142857e-05,
+      "loss": 0.077,
+      "step": 15987
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.069897959183673e-05,
+      "loss": 0.1205,
+      "step": 15988
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.068367346938775e-05,
+      "loss": 0.1692,
+      "step": 15989
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 7.066836734693876e-05,
+      "loss": 0.0333,
+      "step": 15990
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.065306122448978e-05,
+      "loss": 0.0744,
+      "step": 15991
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.063775510204081e-05,
+      "loss": 0.0795,
+      "step": 15992
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.062244897959183e-05,
+      "loss": 0.1497,
+      "step": 15993
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.060714285714285e-05,
+      "loss": 0.055,
+      "step": 15994
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.059183673469386e-05,
+      "loss": 0.0999,
+      "step": 15995
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.05765306122449e-05,
+      "loss": 0.0321,
+      "step": 15996
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.056122448979591e-05,
+      "loss": 0.087,
+      "step": 15997
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.054591836734693e-05,
+      "loss": 0.1199,
+      "step": 15998
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.053061224489794e-05,
+      "loss": 0.0267,
+      "step": 15999
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.051530612244898e-05,
+      "loss": 0.1396,
+      "step": 16000
+    },
+    {
+      "epoch": 15.53,
+      "eval_cer": 0.24182342123425674,
+      "eval_loss": 1.1099697351455688,
+      "eval_runtime": 226.7366,
+      "eval_samples_per_second": 38.873,
+      "eval_steps_per_second": 4.86,
+      "eval_wer": 0.6336661942897318,
+      "step": 16000
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.049999999999999e-05,
+      "loss": 0.1722,
+      "step": 16001
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.048469387755102e-05,
+      "loss": 0.0815,
+      "step": 16002
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.046938775510204e-05,
+      "loss": 0.0387,
+      "step": 16003
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.045408163265306e-05,
+      "loss": 0.0761,
+      "step": 16004
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.043877551020407e-05,
+      "loss": 0.0699,
+      "step": 16005
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.04234693877551e-05,
+      "loss": 0.0412,
+      "step": 16006
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.040816326530612e-05,
+      "loss": 0.2025,
+      "step": 16007
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.039285714285714e-05,
+      "loss": 0.0781,
+      "step": 16008
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.037755102040816e-05,
+      "loss": 0.0942,
+      "step": 16009
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.036224489795919e-05,
+      "loss": 0.028,
+      "step": 16010
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 7.03469387755102e-05,
+      "loss": 0.0335,
+      "step": 16011
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.033163265306122e-05,
+      "loss": 0.0738,
+      "step": 16012
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.031632653061224e-05,
+      "loss": 0.1232,
+      "step": 16013
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.030102040816327e-05,
+      "loss": 0.1589,
+      "step": 16014
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.028571428571428e-05,
+      "loss": 0.0271,
+      "step": 16015
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.02704081632653e-05,
+      "loss": 0.042,
+      "step": 16016
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.025510204081632e-05,
+      "loss": 0.052,
+      "step": 16017
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.023979591836734e-05,
+      "loss": 0.07,
+      "step": 16018
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.022448979591837e-05,
+      "loss": 0.0693,
+      "step": 16019
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.020918367346938e-05,
+      "loss": 0.0798,
+      "step": 16020
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.01938775510204e-05,
+      "loss": 0.067,
+      "step": 16021
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.017857142857142e-05,
+      "loss": 0.0329,
+      "step": 16022
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.016326530612245e-05,
+      "loss": 0.0359,
+      "step": 16023
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.014795918367346e-05,
+      "loss": 0.0415,
+      "step": 16024
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.013265306122448e-05,
+      "loss": 0.1023,
+      "step": 16025
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.01173469387755e-05,
+      "loss": 0.1268,
+      "step": 16026
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.010204081632653e-05,
+      "loss": 0.076,
+      "step": 16027
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.008673469387755e-05,
+      "loss": 0.0524,
+      "step": 16028
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.007142857142856e-05,
+      "loss": 0.0716,
+      "step": 16029
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.005612244897958e-05,
+      "loss": 0.1137,
+      "step": 16030
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 7.004081632653061e-05,
+      "loss": 0.0736,
+      "step": 16031
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 7.002551020408163e-05,
+      "loss": 0.0782,
+      "step": 16032
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 7.001020408163264e-05,
+      "loss": 0.0368,
+      "step": 16033
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.999489795918366e-05,
+      "loss": 0.0732,
+      "step": 16034
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.997959183673469e-05,
+      "loss": 0.1576,
+      "step": 16035
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.996428571428571e-05,
+      "loss": 0.1572,
+      "step": 16036
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.994897959183673e-05,
+      "loss": 0.0428,
+      "step": 16037
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.993367346938774e-05,
+      "loss": 0.081,
+      "step": 16038
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.991836734693876e-05,
+      "loss": 0.0702,
+      "step": 16039
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.990306122448979e-05,
+      "loss": 0.0865,
+      "step": 16040
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.988775510204081e-05,
+      "loss": 0.0731,
+      "step": 16041
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.987244897959182e-05,
+      "loss": 0.1546,
+      "step": 16042
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.985714285714284e-05,
+      "loss": 0.0966,
+      "step": 16043
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.984183673469387e-05,
+      "loss": 0.2042,
+      "step": 16044
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.982653061224489e-05,
+      "loss": 0.042,
+      "step": 16045
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.98112244897959e-05,
+      "loss": 0.1189,
+      "step": 16046
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.979591836734692e-05,
+      "loss": 0.0371,
+      "step": 16047
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.978061224489795e-05,
+      "loss": 0.0681,
+      "step": 16048
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.976530612244897e-05,
+      "loss": 0.0677,
+      "step": 16049
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.975e-05,
+      "loss": 0.0932,
+      "step": 16050
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.973469387755102e-05,
+      "loss": 0.039,
+      "step": 16051
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 6.971938775510204e-05,
+      "loss": 0.075,
+      "step": 16052
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.970408163265305e-05,
+      "loss": 0.0294,
+      "step": 16053
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.968877551020408e-05,
+      "loss": 0.0796,
+      "step": 16054
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.96734693877551e-05,
+      "loss": 0.0286,
+      "step": 16055
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.965816326530612e-05,
+      "loss": 0.1113,
+      "step": 16056
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.964285714285713e-05,
+      "loss": 0.1151,
+      "step": 16057
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.962755102040816e-05,
+      "loss": 0.16,
+      "step": 16058
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.961224489795918e-05,
+      "loss": 0.1077,
+      "step": 16059
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.95969387755102e-05,
+      "loss": 0.0493,
+      "step": 16060
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.958163265306122e-05,
+      "loss": 0.1546,
+      "step": 16061
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.956632653061225e-05,
+      "loss": 0.1072,
+      "step": 16062
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.955102040816326e-05,
+      "loss": 0.0939,
+      "step": 16063
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.953571428571428e-05,
+      "loss": 0.1229,
+      "step": 16064
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.95204081632653e-05,
+      "loss": 0.0764,
+      "step": 16065
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.950510204081633e-05,
+      "loss": 0.0657,
+      "step": 16066
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.948979591836734e-05,
+      "loss": 0.0734,
+      "step": 16067
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.947448979591836e-05,
+      "loss": 0.1628,
+      "step": 16068
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.945918367346938e-05,
+      "loss": 0.0997,
+      "step": 16069
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.94438775510204e-05,
+      "loss": 0.0722,
+      "step": 16070
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.942857142857143e-05,
+      "loss": 0.0591,
+      "step": 16071
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.941326530612244e-05,
+      "loss": 0.1092,
+      "step": 16072
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.939795918367346e-05,
+      "loss": 0.1128,
+      "step": 16073
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.938265306122448e-05,
+      "loss": 0.0406,
+      "step": 16074
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.936734693877551e-05,
+      "loss": 0.1262,
+      "step": 16075
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.935204081632652e-05,
+      "loss": 0.0398,
+      "step": 16076
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.933673469387754e-05,
+      "loss": 0.115,
+      "step": 16077
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.932142857142856e-05,
+      "loss": 0.0775,
+      "step": 16078
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.930612244897959e-05,
+      "loss": 0.112,
+      "step": 16079
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.92908163265306e-05,
+      "loss": 0.0635,
+      "step": 16080
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.927551020408162e-05,
+      "loss": 0.0746,
+      "step": 16081
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.926020408163264e-05,
+      "loss": 0.0381,
+      "step": 16082
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.924489795918367e-05,
+      "loss": 0.0391,
+      "step": 16083
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.922959183673469e-05,
+      "loss": 0.1249,
+      "step": 16084
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.92142857142857e-05,
+      "loss": 0.1345,
+      "step": 16085
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.919897959183672e-05,
+      "loss": 0.0505,
+      "step": 16086
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.918367346938775e-05,
+      "loss": 0.0851,
+      "step": 16087
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.916836734693877e-05,
+      "loss": 0.0636,
+      "step": 16088
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.915306122448979e-05,
+      "loss": 0.1174,
+      "step": 16089
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.91377551020408e-05,
+      "loss": 0.0357,
+      "step": 16090
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.912244897959182e-05,
+      "loss": 0.0676,
+      "step": 16091
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.910714285714285e-05,
+      "loss": 0.0444,
+      "step": 16092
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.909183673469388e-05,
+      "loss": 0.0539,
+      "step": 16093
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.90765306122449e-05,
+      "loss": 0.1059,
+      "step": 16094
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.906122448979592e-05,
+      "loss": 0.0423,
+      "step": 16095
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.904591836734693e-05,
+      "loss": 0.0724,
+      "step": 16096
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.903061224489796e-05,
+      "loss": 0.0373,
+      "step": 16097
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.901530612244898e-05,
+      "loss": 0.0386,
+      "step": 16098
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.9e-05,
+      "loss": 0.1443,
+      "step": 16099
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.898469387755101e-05,
+      "loss": 0.1235,
+      "step": 16100
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.896938775510203e-05,
+      "loss": 0.0759,
+      "step": 16101
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.895408163265306e-05,
+      "loss": 0.0858,
+      "step": 16102
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.893877551020408e-05,
+      "loss": 0.0329,
+      "step": 16103
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 6.89234693877551e-05,
+      "loss": 0.0708,
+      "step": 16104
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.890816326530611e-05,
+      "loss": 0.0562,
+      "step": 16105
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.889285714285714e-05,
+      "loss": 0.0861,
+      "step": 16106
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.887755102040816e-05,
+      "loss": 0.1218,
+      "step": 16107
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.886224489795918e-05,
+      "loss": 0.1256,
+      "step": 16108
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.88469387755102e-05,
+      "loss": 0.0367,
+      "step": 16109
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.883163265306122e-05,
+      "loss": 0.0428,
+      "step": 16110
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.881632653061224e-05,
+      "loss": 0.0775,
+      "step": 16111
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.880102040816326e-05,
+      "loss": 0.0562,
+      "step": 16112
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.878571428571428e-05,
+      "loss": 0.0754,
+      "step": 16113
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.87704081632653e-05,
+      "loss": 0.0752,
+      "step": 16114
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.875510204081632e-05,
+      "loss": 0.0676,
+      "step": 16115
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.873979591836734e-05,
+      "loss": 0.0696,
+      "step": 16116
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.872448979591836e-05,
+      "loss": 0.0757,
+      "step": 16117
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.870918367346939e-05,
+      "loss": 0.1023,
+      "step": 16118
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.86938775510204e-05,
+      "loss": 0.1004,
+      "step": 16119
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.867857142857142e-05,
+      "loss": 0.1053,
+      "step": 16120
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.866326530612244e-05,
+      "loss": 0.0453,
+      "step": 16121
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.864795918367347e-05,
+      "loss": 0.0794,
+      "step": 16122
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.863265306122449e-05,
+      "loss": 0.1327,
+      "step": 16123
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.86173469387755e-05,
+      "loss": 0.1055,
+      "step": 16124
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.860204081632652e-05,
+      "loss": 0.0697,
+      "step": 16125
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.858673469387754e-05,
+      "loss": 0.0552,
+      "step": 16126
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.857142857142857e-05,
+      "loss": 0.1239,
+      "step": 16127
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.855612244897958e-05,
+      "loss": 0.0278,
+      "step": 16128
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.85408163265306e-05,
+      "loss": 0.0247,
+      "step": 16129
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.852551020408162e-05,
+      "loss": 0.0682,
+      "step": 16130
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.851020408163265e-05,
+      "loss": 0.0315,
+      "step": 16131
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.849489795918367e-05,
+      "loss": 0.092,
+      "step": 16132
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.847959183673468e-05,
+      "loss": 0.0398,
+      "step": 16133
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.84642857142857e-05,
+      "loss": 0.0239,
+      "step": 16134
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.844897959183673e-05,
+      "loss": 0.0647,
+      "step": 16135
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.843367346938775e-05,
+      "loss": 0.0756,
+      "step": 16136
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.841836734693876e-05,
+      "loss": 0.0368,
+      "step": 16137
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.840306122448978e-05,
+      "loss": 0.2564,
+      "step": 16138
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.838775510204081e-05,
+      "loss": 0.2085,
+      "step": 16139
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.837244897959183e-05,
+      "loss": 0.0858,
+      "step": 16140
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.835714285714286e-05,
+      "loss": 0.0853,
+      "step": 16141
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.834183673469388e-05,
+      "loss": 0.0689,
+      "step": 16142
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.83265306122449e-05,
+      "loss": 0.0671,
+      "step": 16143
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.831122448979591e-05,
+      "loss": 0.1636,
+      "step": 16144
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 6.829591836734694e-05,
+      "loss": 0.1565,
+      "step": 16145
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.828061224489796e-05,
+      "loss": 0.0903,
+      "step": 16146
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.826530612244897e-05,
+      "loss": 0.0754,
+      "step": 16147
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.824999999999999e-05,
+      "loss": 0.0755,
+      "step": 16148
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.823469387755102e-05,
+      "loss": 0.119,
+      "step": 16149
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.821938775510204e-05,
+      "loss": 0.113,
+      "step": 16150
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.820408163265306e-05,
+      "loss": 0.042,
+      "step": 16151
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.818877551020407e-05,
+      "loss": 0.0789,
+      "step": 16152
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.81734693877551e-05,
+      "loss": 0.0588,
+      "step": 16153
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.815816326530612e-05,
+      "loss": 0.0277,
+      "step": 16154
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.814285714285714e-05,
+      "loss": 0.0664,
+      "step": 16155
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.812755102040816e-05,
+      "loss": 0.0839,
+      "step": 16156
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.811224489795917e-05,
+      "loss": 0.2231,
+      "step": 16157
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.80969387755102e-05,
+      "loss": 0.0764,
+      "step": 16158
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.808163265306122e-05,
+      "loss": 0.0318,
+      "step": 16159
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.806632653061224e-05,
+      "loss": 0.0803,
+      "step": 16160
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.805102040816325e-05,
+      "loss": 0.0231,
+      "step": 16161
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.803571428571428e-05,
+      "loss": 0.0183,
+      "step": 16162
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.80204081632653e-05,
+      "loss": 0.098,
+      "step": 16163
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.800510204081632e-05,
+      "loss": 0.0802,
+      "step": 16164
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.798979591836734e-05,
+      "loss": 0.0395,
+      "step": 16165
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.797448979591837e-05,
+      "loss": 0.0708,
+      "step": 16166
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.795918367346938e-05,
+      "loss": 0.0269,
+      "step": 16167
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.79438775510204e-05,
+      "loss": 0.0413,
+      "step": 16168
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.792857142857142e-05,
+      "loss": 0.0339,
+      "step": 16169
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.791326530612245e-05,
+      "loss": 0.0816,
+      "step": 16170
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.789795918367346e-05,
+      "loss": 0.0324,
+      "step": 16171
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.788265306122448e-05,
+      "loss": 0.0787,
+      "step": 16172
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.78673469387755e-05,
+      "loss": 0.0552,
+      "step": 16173
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.785204081632653e-05,
+      "loss": 0.0329,
+      "step": 16174
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.783673469387755e-05,
+      "loss": 0.0624,
+      "step": 16175
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.782142857142856e-05,
+      "loss": 0.0456,
+      "step": 16176
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.780612244897958e-05,
+      "loss": 0.1147,
+      "step": 16177
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.77908163265306e-05,
+      "loss": 0.0428,
+      "step": 16178
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.777551020408163e-05,
+      "loss": 0.0362,
+      "step": 16179
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.776020408163264e-05,
+      "loss": 0.1112,
+      "step": 16180
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.774489795918366e-05,
+      "loss": 0.1082,
+      "step": 16181
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.772959183673468e-05,
+      "loss": 0.0606,
+      "step": 16182
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.771428571428571e-05,
+      "loss": 0.0361,
+      "step": 16183
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.769897959183674e-05,
+      "loss": 0.0311,
+      "step": 16184
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.768367346938776e-05,
+      "loss": 0.0456,
+      "step": 16185
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.766836734693877e-05,
+      "loss": 0.0279,
+      "step": 16186
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.765306122448979e-05,
+      "loss": 0.0737,
+      "step": 16187
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.763775510204081e-05,
+      "loss": 0.1239,
+      "step": 16188
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.762244897959184e-05,
+      "loss": 0.1099,
+      "step": 16189
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.760714285714285e-05,
+      "loss": 0.1505,
+      "step": 16190
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.759183673469387e-05,
+      "loss": 0.054,
+      "step": 16191
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.757653061224489e-05,
+      "loss": 0.0925,
+      "step": 16192
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.756122448979592e-05,
+      "loss": 0.0555,
+      "step": 16193
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.754591836734694e-05,
+      "loss": 0.0988,
+      "step": 16194
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.753061224489795e-05,
+      "loss": 0.1262,
+      "step": 16195
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 6.751530612244897e-05,
+      "loss": 0.0744,
+      "step": 16196
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.75e-05,
+      "loss": 0.1637,
+      "step": 16197
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.748469387755102e-05,
+      "loss": 0.0818,
+      "step": 16198
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.746938775510203e-05,
+      "loss": 0.0326,
+      "step": 16199
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.745408163265305e-05,
+      "loss": 0.1895,
+      "step": 16200
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.743877551020408e-05,
+      "loss": 0.1144,
+      "step": 16201
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.74234693877551e-05,
+      "loss": 0.0777,
+      "step": 16202
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.740816326530612e-05,
+      "loss": 0.0336,
+      "step": 16203
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.739285714285713e-05,
+      "loss": 0.1131,
+      "step": 16204
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.737755102040816e-05,
+      "loss": 0.0333,
+      "step": 16205
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.736224489795918e-05,
+      "loss": 0.0382,
+      "step": 16206
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.73469387755102e-05,
+      "loss": 0.1634,
+      "step": 16207
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.733163265306121e-05,
+      "loss": 0.044,
+      "step": 16208
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.731632653061223e-05,
+      "loss": 0.0273,
+      "step": 16209
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.730102040816326e-05,
+      "loss": 0.0693,
+      "step": 16210
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.728571428571428e-05,
+      "loss": 0.0794,
+      "step": 16211
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.72704081632653e-05,
+      "loss": 0.0777,
+      "step": 16212
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.725510204081631e-05,
+      "loss": 0.1855,
+      "step": 16213
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.723979591836734e-05,
+      "loss": 0.1585,
+      "step": 16214
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.722448979591836e-05,
+      "loss": 0.0372,
+      "step": 16215
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.720918367346938e-05,
+      "loss": 0.2101,
+      "step": 16216
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.71938775510204e-05,
+      "loss": 0.0861,
+      "step": 16217
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.717857142857143e-05,
+      "loss": 0.0381,
+      "step": 16218
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.716326530612244e-05,
+      "loss": 0.147,
+      "step": 16219
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.714795918367346e-05,
+      "loss": 0.1179,
+      "step": 16220
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.713265306122448e-05,
+      "loss": 0.1296,
+      "step": 16221
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.711734693877551e-05,
+      "loss": 0.0766,
+      "step": 16222
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.710204081632652e-05,
+      "loss": 0.0354,
+      "step": 16223
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.708673469387754e-05,
+      "loss": 0.032,
+      "step": 16224
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.707142857142856e-05,
+      "loss": 0.052,
+      "step": 16225
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.705612244897959e-05,
+      "loss": 0.0907,
+      "step": 16226
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.70408163265306e-05,
+      "loss": 0.0733,
+      "step": 16227
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.702551020408162e-05,
+      "loss": 0.0812,
+      "step": 16228
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.701020408163265e-05,
+      "loss": 0.0779,
+      "step": 16229
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.699489795918367e-05,
+      "loss": 0.1033,
+      "step": 16230
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.697959183673469e-05,
+      "loss": 0.0706,
+      "step": 16231
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.696428571428572e-05,
+      "loss": 0.0804,
+      "step": 16232
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.694897959183673e-05,
+      "loss": 0.0352,
+      "step": 16233
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.693367346938775e-05,
+      "loss": 0.0331,
+      "step": 16234
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.691836734693877e-05,
+      "loss": 0.095,
+      "step": 16235
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.69030612244898e-05,
+      "loss": 0.0724,
+      "step": 16236
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.688775510204082e-05,
+      "loss": 0.0614,
+      "step": 16237
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.687244897959183e-05,
+      "loss": 0.1434,
+      "step": 16238
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.685714285714285e-05,
+      "loss": 0.0412,
+      "step": 16239
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.684183673469387e-05,
+      "loss": 0.0358,
+      "step": 16240
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.68265306122449e-05,
+      "loss": 0.0795,
+      "step": 16241
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.681122448979591e-05,
+      "loss": 0.0306,
+      "step": 16242
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.679591836734693e-05,
+      "loss": 0.0647,
+      "step": 16243
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.678061224489795e-05,
+      "loss": 0.1646,
+      "step": 16244
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.676530612244898e-05,
+      "loss": 0.0814,
+      "step": 16245
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.675e-05,
+      "loss": 0.0349,
+      "step": 16246
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.673469387755101e-05,
+      "loss": 0.0364,
+      "step": 16247
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 6.671938775510203e-05,
+      "loss": 0.0286,
+      "step": 16248
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.670408163265306e-05,
+      "loss": 0.0281,
+      "step": 16249
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.668877551020408e-05,
+      "loss": 0.1139,
+      "step": 16250
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.66734693877551e-05,
+      "loss": 0.1601,
+      "step": 16251
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.665816326530611e-05,
+      "loss": 0.0841,
+      "step": 16252
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.664285714285714e-05,
+      "loss": 0.1124,
+      "step": 16253
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.662755102040816e-05,
+      "loss": 0.0709,
+      "step": 16254
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.661224489795918e-05,
+      "loss": 0.0305,
+      "step": 16255
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.65969387755102e-05,
+      "loss": 0.0795,
+      "step": 16256
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.658163265306122e-05,
+      "loss": 0.178,
+      "step": 16257
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.656632653061224e-05,
+      "loss": 0.0846,
+      "step": 16258
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.655102040816326e-05,
+      "loss": 0.0694,
+      "step": 16259
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.653571428571427e-05,
+      "loss": 0.0345,
+      "step": 16260
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.65204081632653e-05,
+      "loss": 0.1351,
+      "step": 16261
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.650510204081632e-05,
+      "loss": 0.0715,
+      "step": 16262
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.648979591836734e-05,
+      "loss": 0.0722,
+      "step": 16263
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.647448979591836e-05,
+      "loss": 0.0519,
+      "step": 16264
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.645918367346937e-05,
+      "loss": 0.1527,
+      "step": 16265
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.64438775510204e-05,
+      "loss": 0.0312,
+      "step": 16266
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.642857142857142e-05,
+      "loss": 0.0597,
+      "step": 16267
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.641326530612244e-05,
+      "loss": 0.0656,
+      "step": 16268
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.639795918367345e-05,
+      "loss": 0.1356,
+      "step": 16269
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.638265306122449e-05,
+      "loss": 0.1739,
+      "step": 16270
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.63673469387755e-05,
+      "loss": 0.0806,
+      "step": 16271
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.635204081632652e-05,
+      "loss": 0.035,
+      "step": 16272
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.633673469387754e-05,
+      "loss": 0.0283,
+      "step": 16273
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.632142857142857e-05,
+      "loss": 0.1106,
+      "step": 16274
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.630612244897958e-05,
+      "loss": 0.1577,
+      "step": 16275
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.629081632653061e-05,
+      "loss": 0.086,
+      "step": 16276
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.627551020408163e-05,
+      "loss": 0.045,
+      "step": 16277
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.626020408163265e-05,
+      "loss": 0.1542,
+      "step": 16278
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.624489795918367e-05,
+      "loss": 0.0844,
+      "step": 16279
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.62295918367347e-05,
+      "loss": 0.1494,
+      "step": 16280
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.621428571428571e-05,
+      "loss": 0.0204,
+      "step": 16281
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.619897959183673e-05,
+      "loss": 0.1866,
+      "step": 16282
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.618367346938775e-05,
+      "loss": 0.0435,
+      "step": 16283
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.616836734693878e-05,
+      "loss": 0.0338,
+      "step": 16284
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.61530612244898e-05,
+      "loss": 0.1101,
+      "step": 16285
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.613775510204081e-05,
+      "loss": 0.0684,
+      "step": 16286
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.612244897959183e-05,
+      "loss": 0.0735,
+      "step": 16287
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.610714285714286e-05,
+      "loss": 0.1269,
+      "step": 16288
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 6.609183673469388e-05,
+      "loss": 0.2029,
+      "step": 16289
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.607653061224489e-05,
+      "loss": 0.0511,
+      "step": 16290
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.606122448979591e-05,
+      "loss": 0.053,
+      "step": 16291
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.604591836734694e-05,
+      "loss": 0.0871,
+      "step": 16292
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.603061224489796e-05,
+      "loss": 0.0649,
+      "step": 16293
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.601530612244897e-05,
+      "loss": 0.1217,
+      "step": 16294
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.599999999999999e-05,
+      "loss": 0.0307,
+      "step": 16295
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.598469387755101e-05,
+      "loss": 0.1134,
+      "step": 16296
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.596938775510204e-05,
+      "loss": 0.0282,
+      "step": 16297
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.595408163265306e-05,
+      "loss": 0.0446,
+      "step": 16298
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.593877551020407e-05,
+      "loss": 0.0572,
+      "step": 16299
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.592346938775509e-05,
+      "loss": 0.1093,
+      "step": 16300
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.590816326530612e-05,
+      "loss": 0.1198,
+      "step": 16301
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.589285714285714e-05,
+      "loss": 0.0909,
+      "step": 16302
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.587755102040815e-05,
+      "loss": 0.1274,
+      "step": 16303
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.586224489795917e-05,
+      "loss": 0.0739,
+      "step": 16304
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.58469387755102e-05,
+      "loss": 0.0608,
+      "step": 16305
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.583163265306122e-05,
+      "loss": 0.0668,
+      "step": 16306
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.581632653061224e-05,
+      "loss": 0.1021,
+      "step": 16307
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.580102040816325e-05,
+      "loss": 0.0395,
+      "step": 16308
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.578571428571428e-05,
+      "loss": 0.0324,
+      "step": 16309
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.57704081632653e-05,
+      "loss": 0.0753,
+      "step": 16310
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.575510204081632e-05,
+      "loss": 0.0629,
+      "step": 16311
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.573979591836733e-05,
+      "loss": 0.1168,
+      "step": 16312
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.572448979591837e-05,
+      "loss": 0.1153,
+      "step": 16313
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.570918367346938e-05,
+      "loss": 0.0726,
+      "step": 16314
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.56938775510204e-05,
+      "loss": 0.0279,
+      "step": 16315
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.567857142857142e-05,
+      "loss": 0.1109,
+      "step": 16316
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.566326530612243e-05,
+      "loss": 0.0829,
+      "step": 16317
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.564795918367346e-05,
+      "loss": 0.0408,
+      "step": 16318
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.563265306122448e-05,
+      "loss": 0.1763,
+      "step": 16319
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.561734693877551e-05,
+      "loss": 0.0807,
+      "step": 16320
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.560204081632653e-05,
+      "loss": 0.0693,
+      "step": 16321
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.558673469387755e-05,
+      "loss": 0.1056,
+      "step": 16322
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.557142857142858e-05,
+      "loss": 0.0442,
+      "step": 16323
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.555612244897959e-05,
+      "loss": 0.0342,
+      "step": 16324
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.554081632653061e-05,
+      "loss": 0.1721,
+      "step": 16325
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.552551020408163e-05,
+      "loss": 0.077,
+      "step": 16326
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.551020408163264e-05,
+      "loss": 0.0828,
+      "step": 16327
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.549489795918367e-05,
+      "loss": 0.0531,
+      "step": 16328
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.547959183673469e-05,
+      "loss": 0.0836,
+      "step": 16329
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.546428571428571e-05,
+      "loss": 0.0398,
+      "step": 16330
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.544897959183673e-05,
+      "loss": 0.1019,
+      "step": 16331
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.543367346938776e-05,
+      "loss": 0.106,
+      "step": 16332
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.541836734693877e-05,
+      "loss": 0.0853,
+      "step": 16333
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.540306122448979e-05,
+      "loss": 0.0704,
+      "step": 16334
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.538775510204081e-05,
+      "loss": 0.1159,
+      "step": 16335
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.537244897959184e-05,
+      "loss": 0.1153,
+      "step": 16336
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.535714285714285e-05,
+      "loss": 0.0663,
+      "step": 16337
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.535714285714285e-05,
+      "loss": 0.2234,
+      "step": 16338
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.534183673469387e-05,
+      "loss": 0.0369,
+      "step": 16339
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 6.532653061224489e-05,
+      "loss": 0.0359,
+      "step": 16340
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.531122448979592e-05,
+      "loss": 0.0976,
+      "step": 16341
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.529591836734694e-05,
+      "loss": 0.0749,
+      "step": 16342
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.528061224489795e-05,
+      "loss": 0.0799,
+      "step": 16343
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.526530612244897e-05,
+      "loss": 0.1508,
+      "step": 16344
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.525e-05,
+      "loss": 0.1269,
+      "step": 16345
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.523469387755102e-05,
+      "loss": 0.0684,
+      "step": 16346
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.521938775510203e-05,
+      "loss": 0.1064,
+      "step": 16347
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.520408163265305e-05,
+      "loss": 0.0754,
+      "step": 16348
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.518877551020407e-05,
+      "loss": 0.031,
+      "step": 16349
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.51734693877551e-05,
+      "loss": 0.1634,
+      "step": 16350
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.515816326530612e-05,
+      "loss": 0.0452,
+      "step": 16351
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.514285714285713e-05,
+      "loss": 0.0682,
+      "step": 16352
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.512755102040815e-05,
+      "loss": 0.0453,
+      "step": 16353
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.511224489795918e-05,
+      "loss": 0.0298,
+      "step": 16354
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.50969387755102e-05,
+      "loss": 0.1597,
+      "step": 16355
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.508163265306121e-05,
+      "loss": 0.0554,
+      "step": 16356
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.506632653061223e-05,
+      "loss": 0.1956,
+      "step": 16357
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.505102040816326e-05,
+      "loss": 0.0846,
+      "step": 16358
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.503571428571428e-05,
+      "loss": 0.0732,
+      "step": 16359
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.50204081632653e-05,
+      "loss": 0.073,
+      "step": 16360
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.500510204081631e-05,
+      "loss": 0.0517,
+      "step": 16361
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.498979591836734e-05,
+      "loss": 0.084,
+      "step": 16362
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.497448979591836e-05,
+      "loss": 0.1138,
+      "step": 16363
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.495918367346938e-05,
+      "loss": 0.0876,
+      "step": 16364
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.49438775510204e-05,
+      "loss": 0.0661,
+      "step": 16365
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.492857142857143e-05,
+      "loss": 0.0343,
+      "step": 16366
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.491326530612244e-05,
+      "loss": 0.0682,
+      "step": 16367
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.489795918367347e-05,
+      "loss": 0.0585,
+      "step": 16368
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.488265306122449e-05,
+      "loss": 0.068,
+      "step": 16369
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.48673469387755e-05,
+      "loss": 0.2283,
+      "step": 16370
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.485204081632652e-05,
+      "loss": 0.0334,
+      "step": 16371
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.483673469387755e-05,
+      "loss": 0.111,
+      "step": 16372
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.482142857142857e-05,
+      "loss": 0.0836,
+      "step": 16373
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.480612244897959e-05,
+      "loss": 0.0327,
+      "step": 16374
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.47908163265306e-05,
+      "loss": 0.1584,
+      "step": 16375
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.477551020408164e-05,
+      "loss": 0.1195,
+      "step": 16376
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.476020408163265e-05,
+      "loss": 0.1062,
+      "step": 16377
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.474489795918367e-05,
+      "loss": 0.1082,
+      "step": 16378
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.472959183673469e-05,
+      "loss": 0.0308,
+      "step": 16379
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.47142857142857e-05,
+      "loss": 0.0743,
+      "step": 16380
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.469897959183673e-05,
+      "loss": 0.0569,
+      "step": 16381
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.468367346938775e-05,
+      "loss": 0.1706,
+      "step": 16382
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.466836734693877e-05,
+      "loss": 0.142,
+      "step": 16383
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.465306122448979e-05,
+      "loss": 0.1472,
+      "step": 16384
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.463775510204082e-05,
+      "loss": 0.0277,
+      "step": 16385
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.462244897959183e-05,
+      "loss": 0.0732,
+      "step": 16386
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.460714285714285e-05,
+      "loss": 0.0554,
+      "step": 16387
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.459183673469387e-05,
+      "loss": 0.2666,
+      "step": 16388
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.45765306122449e-05,
+      "loss": 0.0458,
+      "step": 16389
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.456122448979591e-05,
+      "loss": 0.0353,
+      "step": 16390
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.454591836734693e-05,
+      "loss": 0.0364,
+      "step": 16391
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 6.453061224489795e-05,
+      "loss": 0.0769,
+      "step": 16392
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.451530612244898e-05,
+      "loss": 0.0315,
+      "step": 16393
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.45e-05,
+      "loss": 0.1019,
+      "step": 16394
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.448469387755101e-05,
+      "loss": 0.0896,
+      "step": 16395
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.446938775510203e-05,
+      "loss": 0.1128,
+      "step": 16396
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.445408163265306e-05,
+      "loss": 0.1586,
+      "step": 16397
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.443877551020408e-05,
+      "loss": 0.06,
+      "step": 16398
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.44234693877551e-05,
+      "loss": 0.0352,
+      "step": 16399
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.440816326530611e-05,
+      "loss": 0.1199,
+      "step": 16400
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.439285714285713e-05,
+      "loss": 0.0909,
+      "step": 16401
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.437755102040816e-05,
+      "loss": 0.0367,
+      "step": 16402
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.436224489795918e-05,
+      "loss": 0.134,
+      "step": 16403
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.434693877551019e-05,
+      "loss": 0.1646,
+      "step": 16404
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.433163265306121e-05,
+      "loss": 0.0734,
+      "step": 16405
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.431632653061224e-05,
+      "loss": 0.0312,
+      "step": 16406
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.430102040816326e-05,
+      "loss": 0.1043,
+      "step": 16407
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.428571428571427e-05,
+      "loss": 0.1626,
+      "step": 16408
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.427040816326529e-05,
+      "loss": 0.0873,
+      "step": 16409
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.425510204081632e-05,
+      "loss": 0.0442,
+      "step": 16410
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.423979591836734e-05,
+      "loss": 0.0708,
+      "step": 16411
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.422448979591837e-05,
+      "loss": 0.0353,
+      "step": 16412
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.420918367346939e-05,
+      "loss": 0.129,
+      "step": 16413
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.41938775510204e-05,
+      "loss": 0.1117,
+      "step": 16414
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.417857142857142e-05,
+      "loss": 0.038,
+      "step": 16415
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.416326530612245e-05,
+      "loss": 0.1494,
+      "step": 16416
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.414795918367347e-05,
+      "loss": 0.0678,
+      "step": 16417
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.413265306122448e-05,
+      "loss": 0.0252,
+      "step": 16418
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.41173469387755e-05,
+      "loss": 0.1956,
+      "step": 16419
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.410204081632653e-05,
+      "loss": 0.0741,
+      "step": 16420
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.408673469387755e-05,
+      "loss": 0.0368,
+      "step": 16421
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.407142857142857e-05,
+      "loss": 0.1191,
+      "step": 16422
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.405612244897958e-05,
+      "loss": 0.0523,
+      "step": 16423
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.404081632653061e-05,
+      "loss": 0.0706,
+      "step": 16424
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.402551020408163e-05,
+      "loss": 0.085,
+      "step": 16425
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.401020408163265e-05,
+      "loss": 0.131,
+      "step": 16426
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.399489795918366e-05,
+      "loss": 0.0714,
+      "step": 16427
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.39795918367347e-05,
+      "loss": 0.0747,
+      "step": 16428
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.396428571428571e-05,
+      "loss": 0.0341,
+      "step": 16429
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.394897959183673e-05,
+      "loss": 0.0441,
+      "step": 16430
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.393367346938775e-05,
+      "loss": 0.028,
+      "step": 16431
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.391836734693878e-05,
+      "loss": 0.1148,
+      "step": 16432
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 6.39030612244898e-05,
+      "loss": 0.0856,
+      "step": 16433
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.388775510204081e-05,
+      "loss": 0.0714,
+      "step": 16434
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.387244897959183e-05,
+      "loss": 0.1427,
+      "step": 16435
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.385714285714284e-05,
+      "loss": 0.0505,
+      "step": 16436
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.384183673469388e-05,
+      "loss": 0.044,
+      "step": 16437
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.382653061224489e-05,
+      "loss": 0.2648,
+      "step": 16438
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.381122448979591e-05,
+      "loss": 0.0403,
+      "step": 16439
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.379591836734693e-05,
+      "loss": 0.0757,
+      "step": 16440
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.378061224489796e-05,
+      "loss": 0.0419,
+      "step": 16441
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.376530612244897e-05,
+      "loss": 0.0846,
+      "step": 16442
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.374999999999999e-05,
+      "loss": 0.0666,
+      "step": 16443
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.373469387755101e-05,
+      "loss": 0.0905,
+      "step": 16444
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.371938775510204e-05,
+      "loss": 0.0711,
+      "step": 16445
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.370408163265306e-05,
+      "loss": 0.0411,
+      "step": 16446
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.368877551020407e-05,
+      "loss": 0.0712,
+      "step": 16447
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.367346938775509e-05,
+      "loss": 0.118,
+      "step": 16448
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.365816326530612e-05,
+      "loss": 0.0652,
+      "step": 16449
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.364285714285714e-05,
+      "loss": 0.1376,
+      "step": 16450
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.362755102040815e-05,
+      "loss": 0.1267,
+      "step": 16451
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.361224489795917e-05,
+      "loss": 0.1436,
+      "step": 16452
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.35969387755102e-05,
+      "loss": 0.0669,
+      "step": 16453
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.358163265306122e-05,
+      "loss": 0.0334,
+      "step": 16454
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.356632653061224e-05,
+      "loss": 0.1074,
+      "step": 16455
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.355102040816325e-05,
+      "loss": 0.1251,
+      "step": 16456
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.353571428571428e-05,
+      "loss": 0.1192,
+      "step": 16457
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.35204081632653e-05,
+      "loss": 0.0744,
+      "step": 16458
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.350510204081633e-05,
+      "loss": 0.1252,
+      "step": 16459
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.348979591836735e-05,
+      "loss": 0.0784,
+      "step": 16460
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.347448979591836e-05,
+      "loss": 0.062,
+      "step": 16461
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.345918367346938e-05,
+      "loss": 0.0719,
+      "step": 16462
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.344387755102041e-05,
+      "loss": 0.1845,
+      "step": 16463
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.342857142857143e-05,
+      "loss": 0.0804,
+      "step": 16464
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.341326530612245e-05,
+      "loss": 0.0751,
+      "step": 16465
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.339795918367346e-05,
+      "loss": 0.0476,
+      "step": 16466
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.338265306122448e-05,
+      "loss": 0.0701,
+      "step": 16467
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.336734693877551e-05,
+      "loss": 0.0305,
+      "step": 16468
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.335204081632653e-05,
+      "loss": 0.074,
+      "step": 16469
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.333673469387754e-05,
+      "loss": 0.1554,
+      "step": 16470
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.332142857142856e-05,
+      "loss": 0.0383,
+      "step": 16471
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.330612244897959e-05,
+      "loss": 0.0942,
+      "step": 16472
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.329081632653061e-05,
+      "loss": 0.1953,
+      "step": 16473
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.327551020408163e-05,
+      "loss": 0.0324,
+      "step": 16474
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.326020408163264e-05,
+      "loss": 0.1121,
+      "step": 16475
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.324489795918367e-05,
+      "loss": 0.157,
+      "step": 16476
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.322959183673469e-05,
+      "loss": 0.0731,
+      "step": 16477
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.321428571428571e-05,
+      "loss": 0.0753,
+      "step": 16478
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.319897959183672e-05,
+      "loss": 0.0421,
+      "step": 16479
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.318367346938776e-05,
+      "loss": 0.1571,
+      "step": 16480
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.316836734693877e-05,
+      "loss": 0.0791,
+      "step": 16481
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.315306122448979e-05,
+      "loss": 0.0418,
+      "step": 16482
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.31377551020408e-05,
+      "loss": 0.0683,
+      "step": 16483
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.312244897959184e-05,
+      "loss": 0.0734,
+      "step": 16484
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 6.310714285714285e-05,
+      "loss": 0.0393,
+      "step": 16485
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.309183673469387e-05,
+      "loss": 0.0834,
+      "step": 16486
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.307653061224489e-05,
+      "loss": 0.0717,
+      "step": 16487
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.30612244897959e-05,
+      "loss": 0.0313,
+      "step": 16488
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.304591836734694e-05,
+      "loss": 0.0729,
+      "step": 16489
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.303061224489795e-05,
+      "loss": 0.1329,
+      "step": 16490
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.301530612244897e-05,
+      "loss": 0.0337,
+      "step": 16491
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.299999999999999e-05,
+      "loss": 0.0544,
+      "step": 16492
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.298469387755102e-05,
+      "loss": 0.1471,
+      "step": 16493
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.296938775510203e-05,
+      "loss": 0.0372,
+      "step": 16494
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.295408163265305e-05,
+      "loss": 0.0353,
+      "step": 16495
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.293877551020407e-05,
+      "loss": 0.0411,
+      "step": 16496
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.29234693877551e-05,
+      "loss": 0.102,
+      "step": 16497
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.290816326530612e-05,
+      "loss": 0.0686,
+      "step": 16498
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.289285714285713e-05,
+      "loss": 0.0757,
+      "step": 16499
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.287755102040815e-05,
+      "loss": 0.0701,
+      "step": 16500
+    },
+    {
+      "epoch": 16.02,
+      "eval_cer": 0.24147702794558318,
+      "eval_loss": 1.173096776008606,
+      "eval_runtime": 229.5682,
+      "eval_samples_per_second": 38.394,
+      "eval_steps_per_second": 4.8,
+      "eval_wer": 0.630897845741467,
+      "step": 16500
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.286224489795918e-05,
+      "loss": 0.1179,
+      "step": 16501
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.28469387755102e-05,
+      "loss": 0.0342,
+      "step": 16502
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.283163265306123e-05,
+      "loss": 0.0979,
+      "step": 16503
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.281632653061224e-05,
+      "loss": 0.0319,
+      "step": 16504
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.280102040816326e-05,
+      "loss": 0.0562,
+      "step": 16505
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.278571428571428e-05,
+      "loss": 0.0346,
+      "step": 16506
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.277040816326531e-05,
+      "loss": 0.0697,
+      "step": 16507
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.275510204081633e-05,
+      "loss": 0.1323,
+      "step": 16508
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.273979591836734e-05,
+      "loss": 0.0252,
+      "step": 16509
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.272448979591836e-05,
+      "loss": 0.0324,
+      "step": 16510
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.270918367346939e-05,
+      "loss": 0.1033,
+      "step": 16511
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.269387755102041e-05,
+      "loss": 0.199,
+      "step": 16512
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.267857142857142e-05,
+      "loss": 0.1271,
+      "step": 16513
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.266326530612244e-05,
+      "loss": 0.0675,
+      "step": 16514
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.264795918367347e-05,
+      "loss": 0.0343,
+      "step": 16515
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.263265306122449e-05,
+      "loss": 0.0325,
+      "step": 16516
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.26173469387755e-05,
+      "loss": 0.0314,
+      "step": 16517
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.260204081632652e-05,
+      "loss": 0.174,
+      "step": 16518
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.258673469387754e-05,
+      "loss": 0.0467,
+      "step": 16519
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.257142857142857e-05,
+      "loss": 0.069,
+      "step": 16520
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.255612244897959e-05,
+      "loss": 0.1455,
+      "step": 16521
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.25408163265306e-05,
+      "loss": 0.0875,
+      "step": 16522
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.252551020408162e-05,
+      "loss": 0.0433,
+      "step": 16523
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.251020408163265e-05,
+      "loss": 0.1039,
+      "step": 16524
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.249489795918367e-05,
+      "loss": 0.0367,
+      "step": 16525
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 6.247959183673469e-05,
+      "loss": 0.1213,
+      "step": 16526
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.24642857142857e-05,
+      "loss": 0.1149,
+      "step": 16527
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.244897959183673e-05,
+      "loss": 0.0429,
+      "step": 16528
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.243367346938775e-05,
+      "loss": 0.0683,
+      "step": 16529
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.241836734693877e-05,
+      "loss": 0.0692,
+      "step": 16530
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.240306122448978e-05,
+      "loss": 0.1162,
+      "step": 16531
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.238775510204082e-05,
+      "loss": 0.1163,
+      "step": 16532
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.237244897959183e-05,
+      "loss": 0.0358,
+      "step": 16533
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.235714285714285e-05,
+      "loss": 0.0822,
+      "step": 16534
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.234183673469387e-05,
+      "loss": 0.0353,
+      "step": 16535
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.23265306122449e-05,
+      "loss": 0.0786,
+      "step": 16536
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.231122448979591e-05,
+      "loss": 0.1596,
+      "step": 16537
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.229591836734693e-05,
+      "loss": 0.1125,
+      "step": 16538
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.228061224489795e-05,
+      "loss": 0.0451,
+      "step": 16539
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.226530612244896e-05,
+      "loss": 0.0664,
+      "step": 16540
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.225e-05,
+      "loss": 0.0283,
+      "step": 16541
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.223469387755101e-05,
+      "loss": 0.077,
+      "step": 16542
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.221938775510203e-05,
+      "loss": 0.1254,
+      "step": 16543
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.220408163265305e-05,
+      "loss": 0.085,
+      "step": 16544
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.218877551020408e-05,
+      "loss": 0.0692,
+      "step": 16545
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.21734693877551e-05,
+      "loss": 0.0447,
+      "step": 16546
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.215816326530611e-05,
+      "loss": 0.0723,
+      "step": 16547
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.214285714285714e-05,
+      "loss": 0.0542,
+      "step": 16548
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.212755102040816e-05,
+      "loss": 0.1431,
+      "step": 16549
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.211224489795918e-05,
+      "loss": 0.0888,
+      "step": 16550
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.20969387755102e-05,
+      "loss": 0.0693,
+      "step": 16551
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.208163265306122e-05,
+      "loss": 0.136,
+      "step": 16552
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.206632653061224e-05,
+      "loss": 0.0979,
+      "step": 16553
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.205102040816326e-05,
+      "loss": 0.1101,
+      "step": 16554
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.203571428571429e-05,
+      "loss": 0.0678,
+      "step": 16555
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.20204081632653e-05,
+      "loss": 0.1572,
+      "step": 16556
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.200510204081632e-05,
+      "loss": 0.0718,
+      "step": 16557
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.198979591836734e-05,
+      "loss": 0.0347,
+      "step": 16558
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.197448979591837e-05,
+      "loss": 0.1489,
+      "step": 16559
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.195918367346939e-05,
+      "loss": 0.0604,
+      "step": 16560
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.19438775510204e-05,
+      "loss": 0.04,
+      "step": 16561
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.192857142857142e-05,
+      "loss": 0.0956,
+      "step": 16562
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.191326530612245e-05,
+      "loss": 0.0363,
+      "step": 16563
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.189795918367347e-05,
+      "loss": 0.0858,
+      "step": 16564
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.188265306122448e-05,
+      "loss": 0.0678,
+      "step": 16565
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.18673469387755e-05,
+      "loss": 0.1011,
+      "step": 16566
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.185204081632653e-05,
+      "loss": 0.0277,
+      "step": 16567
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.183673469387755e-05,
+      "loss": 0.0336,
+      "step": 16568
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.182142857142857e-05,
+      "loss": 0.049,
+      "step": 16569
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.180612244897958e-05,
+      "loss": 0.0715,
+      "step": 16570
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.17908163265306e-05,
+      "loss": 0.0264,
+      "step": 16571
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.177551020408163e-05,
+      "loss": 0.0763,
+      "step": 16572
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.176020408163265e-05,
+      "loss": 0.0654,
+      "step": 16573
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.174489795918366e-05,
+      "loss": 0.1675,
+      "step": 16574
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.172959183673468e-05,
+      "loss": 0.072,
+      "step": 16575
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.171428571428571e-05,
+      "loss": 0.0836,
+      "step": 16576
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.169897959183673e-05,
+      "loss": 0.09,
+      "step": 16577
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.168367346938775e-05,
+      "loss": 0.0906,
+      "step": 16578
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.166836734693876e-05,
+      "loss": 0.061,
+      "step": 16579
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.16530612244898e-05,
+      "loss": 0.0479,
+      "step": 16580
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.163775510204081e-05,
+      "loss": 0.1589,
+      "step": 16581
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.162244897959183e-05,
+      "loss": 0.1609,
+      "step": 16582
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.160714285714284e-05,
+      "loss": 0.104,
+      "step": 16583
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.159183673469388e-05,
+      "loss": 0.0242,
+      "step": 16584
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.157653061224489e-05,
+      "loss": 0.0673,
+      "step": 16585
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.156122448979591e-05,
+      "loss": 0.1128,
+      "step": 16586
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.154591836734693e-05,
+      "loss": 0.2136,
+      "step": 16587
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.153061224489796e-05,
+      "loss": 0.0812,
+      "step": 16588
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.151530612244897e-05,
+      "loss": 0.0704,
+      "step": 16589
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.149999999999999e-05,
+      "loss": 0.0728,
+      "step": 16590
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.148469387755101e-05,
+      "loss": 0.0309,
+      "step": 16591
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.146938775510204e-05,
+      "loss": 0.0316,
+      "step": 16592
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.145408163265306e-05,
+      "loss": 0.0913,
+      "step": 16593
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.143877551020409e-05,
+      "loss": 0.1986,
+      "step": 16594
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.14234693877551e-05,
+      "loss": 0.0275,
+      "step": 16595
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.140816326530612e-05,
+      "loss": 0.034,
+      "step": 16596
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.139285714285714e-05,
+      "loss": 0.0328,
+      "step": 16597
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.137755102040817e-05,
+      "loss": 0.0679,
+      "step": 16598
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.136224489795918e-05,
+      "loss": 0.1339,
+      "step": 16599
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.13469387755102e-05,
+      "loss": 0.0794,
+      "step": 16600
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.133163265306122e-05,
+      "loss": 0.1121,
+      "step": 16601
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.131632653061225e-05,
+      "loss": 0.0418,
+      "step": 16602
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.130102040816327e-05,
+      "loss": 0.0651,
+      "step": 16603
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.128571428571428e-05,
+      "loss": 0.0586,
+      "step": 16604
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.12704081632653e-05,
+      "loss": 0.0753,
+      "step": 16605
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.125510204081632e-05,
+      "loss": 0.1152,
+      "step": 16606
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.123979591836735e-05,
+      "loss": 0.1174,
+      "step": 16607
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.122448979591836e-05,
+      "loss": 0.0458,
+      "step": 16608
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.120918367346938e-05,
+      "loss": 0.0751,
+      "step": 16609
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.11938775510204e-05,
+      "loss": 0.1038,
+      "step": 16610
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.117857142857143e-05,
+      "loss": 0.0493,
+      "step": 16611
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.116326530612245e-05,
+      "loss": 0.0364,
+      "step": 16612
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.114795918367346e-05,
+      "loss": 0.1303,
+      "step": 16613
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.113265306122448e-05,
+      "loss": 0.0355,
+      "step": 16614
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.111734693877551e-05,
+      "loss": 0.0334,
+      "step": 16615
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.110204081632653e-05,
+      "loss": 0.112,
+      "step": 16616
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.108673469387754e-05,
+      "loss": 0.0877,
+      "step": 16617
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.107142857142856e-05,
+      "loss": 0.0387,
+      "step": 16618
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.105612244897959e-05,
+      "loss": 0.0717,
+      "step": 16619
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.104081632653061e-05,
+      "loss": 0.0238,
+      "step": 16620
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.1025510204081626e-05,
+      "loss": 0.1813,
+      "step": 16621
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.101020408163265e-05,
+      "loss": 0.082,
+      "step": 16622
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.0994897959183667e-05,
+      "loss": 0.0361,
+      "step": 16623
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.097959183673469e-05,
+      "loss": 0.1158,
+      "step": 16624
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.096428571428571e-05,
+      "loss": 0.0833,
+      "step": 16625
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.0948979591836724e-05,
+      "loss": 0.113,
+      "step": 16626
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.093367346938775e-05,
+      "loss": 0.0431,
+      "step": 16627
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.0918367346938765e-05,
+      "loss": 0.1118,
+      "step": 16628
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.090306122448979e-05,
+      "loss": 0.0286,
+      "step": 16629
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.0887755102040806e-05,
+      "loss": 0.0337,
+      "step": 16630
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.087244897959183e-05,
+      "loss": 0.1221,
+      "step": 16631
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.0857142857142847e-05,
+      "loss": 0.0336,
+      "step": 16632
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.084183673469387e-05,
+      "loss": 0.0331,
+      "step": 16633
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.082653061224489e-05,
+      "loss": 0.085,
+      "step": 16634
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.081122448979591e-05,
+      "loss": 0.0265,
+      "step": 16635
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.079591836734693e-05,
+      "loss": 0.1159,
+      "step": 16636
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.078061224489795e-05,
+      "loss": 0.1135,
+      "step": 16637
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.076530612244897e-05,
+      "loss": 0.0394,
+      "step": 16638
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.075e-05,
+      "loss": 0.0749,
+      "step": 16639
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.0734693877551016e-05,
+      "loss": 0.0703,
+      "step": 16640
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.071938775510204e-05,
+      "loss": 0.0203,
+      "step": 16641
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.070408163265306e-05,
+      "loss": 0.0633,
+      "step": 16642
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.068877551020408e-05,
+      "loss": 0.2131,
+      "step": 16643
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.06734693877551e-05,
+      "loss": 0.0668,
+      "step": 16644
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.065816326530612e-05,
+      "loss": 0.0664,
+      "step": 16645
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.064285714285714e-05,
+      "loss": 0.0329,
+      "step": 16646
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.062755102040816e-05,
+      "loss": 0.0708,
+      "step": 16647
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.061224489795918e-05,
+      "loss": 0.0859,
+      "step": 16648
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.05969387755102e-05,
+      "loss": 0.169,
+      "step": 16649
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.058163265306122e-05,
+      "loss": 0.0803,
+      "step": 16650
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.0566326530612244e-05,
+      "loss": 0.0368,
+      "step": 16651
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.055102040816326e-05,
+      "loss": 0.035,
+      "step": 16652
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.0535714285714285e-05,
+      "loss": 0.0708,
+      "step": 16653
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.05204081632653e-05,
+      "loss": 0.0297,
+      "step": 16654
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.0505102040816326e-05,
+      "loss": 0.0761,
+      "step": 16655
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.048979591836734e-05,
+      "loss": 0.1404,
+      "step": 16656
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.047448979591836e-05,
+      "loss": 0.0361,
+      "step": 16657
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.045918367346938e-05,
+      "loss": 0.0455,
+      "step": 16658
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.04438775510204e-05,
+      "loss": 0.1169,
+      "step": 16659
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.0428571428571424e-05,
+      "loss": 0.1703,
+      "step": 16660
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.041326530612244e-05,
+      "loss": 0.0648,
+      "step": 16661
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.0397959183673465e-05,
+      "loss": 0.1215,
+      "step": 16662
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.038265306122448e-05,
+      "loss": 0.0754,
+      "step": 16663
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.0367346938775506e-05,
+      "loss": 0.045,
+      "step": 16664
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.035204081632652e-05,
+      "loss": 0.057,
+      "step": 16665
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.0336734693877546e-05,
+      "loss": 0.0278,
+      "step": 16666
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.032142857142856e-05,
+      "loss": 0.074,
+      "step": 16667
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.030612244897959e-05,
+      "loss": 0.1992,
+      "step": 16668
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.0290816326530604e-05,
+      "loss": 0.0451,
+      "step": 16669
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.027551020408163e-05,
+      "loss": 0.1088,
+      "step": 16670
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.0260204081632645e-05,
+      "loss": 0.1129,
+      "step": 16671
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.024489795918367e-05,
+      "loss": 0.0701,
+      "step": 16672
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.0229591836734686e-05,
+      "loss": 0.0331,
+      "step": 16673
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.021428571428571e-05,
+      "loss": 0.0506,
+      "step": 16674
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.0198979591836726e-05,
+      "loss": 0.1558,
+      "step": 16675
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.018367346938775e-05,
+      "loss": 0.1467,
+      "step": 16676
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.016836734693877e-05,
+      "loss": 0.0584,
+      "step": 16677
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.0153061224489784e-05,
+      "loss": 0.074,
+      "step": 16678
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.013775510204081e-05,
+      "loss": 0.058,
+      "step": 16679
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.0122448979591825e-05,
+      "loss": 0.1182,
+      "step": 16680
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.010714285714285e-05,
+      "loss": 0.1261,
+      "step": 16681
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.0091836734693866e-05,
+      "loss": 0.0329,
+      "step": 16682
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.007653061224489e-05,
+      "loss": 0.1181,
+      "step": 16683
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.006122448979592e-05,
+      "loss": 0.025,
+      "step": 16684
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.004591836734694e-05,
+      "loss": 0.0467,
+      "step": 16685
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.003061224489796e-05,
+      "loss": 0.0773,
+      "step": 16686
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.001530612244898e-05,
+      "loss": 0.1652,
+      "step": 16687
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 0.0725,
+      "step": 16688
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 5.998469387755102e-05,
+      "loss": 0.0391,
+      "step": 16689
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 5.9969387755102036e-05,
+      "loss": 0.1457,
+      "step": 16690
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 5.995408163265306e-05,
+      "loss": 0.0574,
+      "step": 16691
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.9938775510204076e-05,
+      "loss": 0.0892,
+      "step": 16692
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.99234693877551e-05,
+      "loss": 0.1832,
+      "step": 16693
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.990816326530612e-05,
+      "loss": 0.0445,
+      "step": 16694
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.989285714285714e-05,
+      "loss": 0.0294,
+      "step": 16695
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.987755102040816e-05,
+      "loss": 0.0303,
+      "step": 16696
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.986224489795918e-05,
+      "loss": 0.0343,
+      "step": 16697
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.98469387755102e-05,
+      "loss": 0.0305,
+      "step": 16698
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.983163265306122e-05,
+      "loss": 0.1096,
+      "step": 16699
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.981632653061224e-05,
+      "loss": 0.0703,
+      "step": 16700
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.980102040816326e-05,
+      "loss": 0.0608,
+      "step": 16701
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.978571428571428e-05,
+      "loss": 0.0549,
+      "step": 16702
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.9770408163265304e-05,
+      "loss": 0.0855,
+      "step": 16703
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.975510204081632e-05,
+      "loss": 0.0653,
+      "step": 16704
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.9739795918367345e-05,
+      "loss": 0.0739,
+      "step": 16705
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.972448979591836e-05,
+      "loss": 0.1918,
+      "step": 16706
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.9709183673469385e-05,
+      "loss": 0.1129,
+      "step": 16707
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.96938775510204e-05,
+      "loss": 0.0654,
+      "step": 16708
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.9678571428571426e-05,
+      "loss": 0.0614,
+      "step": 16709
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.966326530612244e-05,
+      "loss": 0.1102,
+      "step": 16710
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.964795918367346e-05,
+      "loss": 0.0566,
+      "step": 16711
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.9632653061224484e-05,
+      "loss": 0.1176,
+      "step": 16712
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.96173469387755e-05,
+      "loss": 0.1063,
+      "step": 16713
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.9602040816326525e-05,
+      "loss": 0.0844,
+      "step": 16714
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.958673469387754e-05,
+      "loss": 0.0591,
+      "step": 16715
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.9571428571428565e-05,
+      "loss": 0.032,
+      "step": 16716
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.955612244897958e-05,
+      "loss": 0.0399,
+      "step": 16717
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.9540816326530606e-05,
+      "loss": 0.184,
+      "step": 16718
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.952551020408162e-05,
+      "loss": 0.1141,
+      "step": 16719
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.951020408163265e-05,
+      "loss": 0.0661,
+      "step": 16720
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.9494897959183664e-05,
+      "loss": 0.076,
+      "step": 16721
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.947959183673469e-05,
+      "loss": 0.0326,
+      "step": 16722
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.9464285714285705e-05,
+      "loss": 0.0778,
+      "step": 16723
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.944897959183673e-05,
+      "loss": 0.1849,
+      "step": 16724
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.9433673469387745e-05,
+      "loss": 0.0684,
+      "step": 16725
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.941836734693877e-05,
+      "loss": 0.0648,
+      "step": 16726
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.9403061224489786e-05,
+      "loss": 0.0652,
+      "step": 16727
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.938775510204081e-05,
+      "loss": 0.0809,
+      "step": 16728
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.9372448979591834e-05,
+      "loss": 0.0622,
+      "step": 16729
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.935714285714286e-05,
+      "loss": 0.0398,
+      "step": 16730
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.9341836734693875e-05,
+      "loss": 0.0846,
+      "step": 16731
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.93265306122449e-05,
+      "loss": 0.0269,
+      "step": 16732
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.9311224489795915e-05,
+      "loss": 0.0757,
+      "step": 16733
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.929591836734694e-05,
+      "loss": 0.0297,
+      "step": 16734
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.9280612244897956e-05,
+      "loss": 0.0312,
+      "step": 16735
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.926530612244898e-05,
+      "loss": 0.1264,
+      "step": 16736
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.925e-05,
+      "loss": 0.1178,
+      "step": 16737
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.923469387755102e-05,
+      "loss": 0.0364,
+      "step": 16738
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.921938775510204e-05,
+      "loss": 0.0363,
+      "step": 16739
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.920408163265306e-05,
+      "loss": 0.1603,
+      "step": 16740
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.918877551020408e-05,
+      "loss": 0.0838,
+      "step": 16741
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.9173469387755095e-05,
+      "loss": 0.1458,
+      "step": 16742
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.915816326530612e-05,
+      "loss": 0.1234,
+      "step": 16743
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.9142857142857136e-05,
+      "loss": 0.0794,
+      "step": 16744
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.912755102040816e-05,
+      "loss": 0.0362,
+      "step": 16745
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.911224489795918e-05,
+      "loss": 0.1287,
+      "step": 16746
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.90969387755102e-05,
+      "loss": 0.0675,
+      "step": 16747
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.908163265306122e-05,
+      "loss": 0.0341,
+      "step": 16748
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.906632653061224e-05,
+      "loss": 0.1143,
+      "step": 16749
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.905102040816326e-05,
+      "loss": 0.0609,
+      "step": 16750
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.903571428571428e-05,
+      "loss": 0.0381,
+      "step": 16751
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.90204081632653e-05,
+      "loss": 0.043,
+      "step": 16752
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.900510204081632e-05,
+      "loss": 0.0337,
+      "step": 16753
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.898979591836734e-05,
+      "loss": 0.0585,
+      "step": 16754
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.8974489795918364e-05,
+      "loss": 0.1292,
+      "step": 16755
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.895918367346938e-05,
+      "loss": 0.1145,
+      "step": 16756
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.8943877551020404e-05,
+      "loss": 0.0349,
+      "step": 16757
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.892857142857142e-05,
+      "loss": 0.0301,
+      "step": 16758
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.8913265306122445e-05,
+      "loss": 0.0844,
+      "step": 16759
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.889795918367346e-05,
+      "loss": 0.0268,
+      "step": 16760
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.8882653061224486e-05,
+      "loss": 0.1132,
+      "step": 16761
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.88673469387755e-05,
+      "loss": 0.0742,
+      "step": 16762
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.885204081632652e-05,
+      "loss": 0.0711,
+      "step": 16763
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.8836734693877544e-05,
+      "loss": 0.0297,
+      "step": 16764
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.882142857142856e-05,
+      "loss": 0.0325,
+      "step": 16765
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.8806122448979585e-05,
+      "loss": 0.1048,
+      "step": 16766
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.87908163265306e-05,
+      "loss": 0.0351,
+      "step": 16767
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.8775510204081625e-05,
+      "loss": 0.0739,
+      "step": 16768
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.876020408163264e-05,
+      "loss": 0.1498,
+      "step": 16769
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.8744897959183666e-05,
+      "loss": 0.0284,
+      "step": 16770
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.872959183673468e-05,
+      "loss": 0.0262,
+      "step": 16771
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.871428571428571e-05,
+      "loss": 0.0247,
+      "step": 16772
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.8698979591836724e-05,
+      "loss": 0.05,
+      "step": 16773
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.868367346938775e-05,
+      "loss": 0.1008,
+      "step": 16774
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.866836734693877e-05,
+      "loss": 0.0891,
+      "step": 16775
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.8653061224489795e-05,
+      "loss": 0.0229,
+      "step": 16776
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.863775510204081e-05,
+      "loss": 0.1242,
+      "step": 16777
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.8622448979591836e-05,
+      "loss": 0.1308,
+      "step": 16778
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.860714285714285e-05,
+      "loss": 0.0658,
+      "step": 16779
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.859183673469388e-05,
+      "loss": 0.2538,
+      "step": 16780
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.8576530612244894e-05,
+      "loss": 0.1522,
+      "step": 16781
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.856122448979592e-05,
+      "loss": 0.033,
+      "step": 16782
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.8545918367346934e-05,
+      "loss": 0.2416,
+      "step": 16783
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.853061224489796e-05,
+      "loss": 0.041,
+      "step": 16784
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.8515306122448975e-05,
+      "loss": 0.1132,
+      "step": 16785
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.85e-05,
+      "loss": 0.0419,
+      "step": 16786
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.8484693877551016e-05,
+      "loss": 0.1571,
+      "step": 16787
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.846938775510204e-05,
+      "loss": 0.1104,
+      "step": 16788
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.845408163265306e-05,
+      "loss": 0.1111,
+      "step": 16789
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.843877551020408e-05,
+      "loss": 0.0275,
+      "step": 16790
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.84234693877551e-05,
+      "loss": 0.0778,
+      "step": 16791
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.840816326530612e-05,
+      "loss": 0.1409,
+      "step": 16792
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.839285714285714e-05,
+      "loss": 0.1764,
+      "step": 16793
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.837755102040816e-05,
+      "loss": 0.1067,
+      "step": 16794
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.836224489795918e-05,
+      "loss": 0.0756,
+      "step": 16795
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.8346938775510196e-05,
+      "loss": 0.1497,
+      "step": 16796
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.833163265306122e-05,
+      "loss": 0.0209,
+      "step": 16797
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.831632653061224e-05,
+      "loss": 0.0373,
+      "step": 16798
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.830102040816326e-05,
+      "loss": 0.0491,
+      "step": 16799
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.828571428571428e-05,
+      "loss": 0.1094,
+      "step": 16800
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.82704081632653e-05,
+      "loss": 0.0696,
+      "step": 16801
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.825510204081632e-05,
+      "loss": 0.0237,
+      "step": 16802
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.823979591836734e-05,
+      "loss": 0.0372,
+      "step": 16803
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.822448979591836e-05,
+      "loss": 0.0683,
+      "step": 16804
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.820918367346938e-05,
+      "loss": 0.1175,
+      "step": 16805
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.81938775510204e-05,
+      "loss": 0.0307,
+      "step": 16806
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.8178571428571424e-05,
+      "loss": 0.0331,
+      "step": 16807
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.816326530612244e-05,
+      "loss": 0.0737,
+      "step": 16808
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.8147959183673464e-05,
+      "loss": 0.0294,
+      "step": 16809
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.813265306122448e-05,
+      "loss": 0.0334,
+      "step": 16810
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.8117346938775505e-05,
+      "loss": 0.028,
+      "step": 16811
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.810204081632652e-05,
+      "loss": 0.0814,
+      "step": 16812
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.8086734693877546e-05,
+      "loss": 0.062,
+      "step": 16813
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.807142857142856e-05,
+      "loss": 0.037,
+      "step": 16814
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.805612244897959e-05,
+      "loss": 0.0842,
+      "step": 16815
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.8040816326530604e-05,
+      "loss": 0.0926,
+      "step": 16816
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.802551020408162e-05,
+      "loss": 0.0371,
+      "step": 16817
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.8010204081632644e-05,
+      "loss": 0.1587,
+      "step": 16818
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.799489795918366e-05,
+      "loss": 0.1097,
+      "step": 16819
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.797959183673469e-05,
+      "loss": 0.1078,
+      "step": 16820
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.7964285714285716e-05,
+      "loss": 0.0743,
+      "step": 16821
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.794897959183673e-05,
+      "loss": 0.1048,
+      "step": 16822
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.7933673469387756e-05,
+      "loss": 0.0231,
+      "step": 16823
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.7918367346938773e-05,
+      "loss": 0.086,
+      "step": 16824
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.79030612244898e-05,
+      "loss": 0.074,
+      "step": 16825
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.7887755102040814e-05,
+      "loss": 0.073,
+      "step": 16826
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.787244897959183e-05,
+      "loss": 0.0225,
+      "step": 16827
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.7857142857142855e-05,
+      "loss": 0.0712,
+      "step": 16828
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.784183673469387e-05,
+      "loss": 0.0355,
+      "step": 16829
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.7826530612244896e-05,
+      "loss": 0.1425,
+      "step": 16830
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.781122448979591e-05,
+      "loss": 0.0344,
+      "step": 16831
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.7795918367346937e-05,
+      "loss": 0.1076,
+      "step": 16832
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.7780612244897953e-05,
+      "loss": 0.0727,
+      "step": 16833
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.776530612244898e-05,
+      "loss": 0.0569,
+      "step": 16834
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.7749999999999994e-05,
+      "loss": 0.065,
+      "step": 16835
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.773469387755102e-05,
+      "loss": 0.0863,
+      "step": 16836
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.7719387755102035e-05,
+      "loss": 0.0806,
+      "step": 16837
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.770408163265306e-05,
+      "loss": 0.119,
+      "step": 16838
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.7688775510204076e-05,
+      "loss": 0.0778,
+      "step": 16839
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.76734693877551e-05,
+      "loss": 0.0294,
+      "step": 16840
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.7658163265306117e-05,
+      "loss": 0.0394,
+      "step": 16841
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.764285714285714e-05,
+      "loss": 0.0627,
+      "step": 16842
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.762755102040816e-05,
+      "loss": 0.0772,
+      "step": 16843
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.761224489795918e-05,
+      "loss": 0.0288,
+      "step": 16844
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.75969387755102e-05,
+      "loss": 0.0891,
+      "step": 16845
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.758163265306122e-05,
+      "loss": 0.0754,
+      "step": 16846
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.756632653061224e-05,
+      "loss": 0.0693,
+      "step": 16847
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.7551020408163256e-05,
+      "loss": 0.0507,
+      "step": 16848
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.753571428571428e-05,
+      "loss": 0.1017,
+      "step": 16849
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.7520408163265297e-05,
+      "loss": 0.1542,
+      "step": 16850
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.750510204081632e-05,
+      "loss": 0.0263,
+      "step": 16851
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.748979591836734e-05,
+      "loss": 0.046,
+      "step": 16852
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.747448979591836e-05,
+      "loss": 0.1491,
+      "step": 16853
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.745918367346938e-05,
+      "loss": 0.0301,
+      "step": 16854
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.74438775510204e-05,
+      "loss": 0.106,
+      "step": 16855
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.742857142857142e-05,
+      "loss": 0.0736,
+      "step": 16856
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.741326530612244e-05,
+      "loss": 0.0753,
+      "step": 16857
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.739795918367346e-05,
+      "loss": 0.0235,
+      "step": 16858
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.7382653061224483e-05,
+      "loss": 0.1422,
+      "step": 16859
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.73673469387755e-05,
+      "loss": 0.1451,
+      "step": 16860
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.7352040816326524e-05,
+      "loss": 0.0351,
+      "step": 16861
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.733673469387754e-05,
+      "loss": 0.0937,
+      "step": 16862
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.7321428571428565e-05,
+      "loss": 0.0288,
+      "step": 16863
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.730612244897958e-05,
+      "loss": 0.0379,
+      "step": 16864
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.729081632653061e-05,
+      "loss": 0.0254,
+      "step": 16865
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.727551020408163e-05,
+      "loss": 0.1884,
+      "step": 16866
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.726020408163265e-05,
+      "loss": 0.0376,
+      "step": 16867
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.724489795918367e-05,
+      "loss": 0.1098,
+      "step": 16868
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.7229591836734694e-05,
+      "loss": 0.1205,
+      "step": 16869
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.721428571428571e-05,
+      "loss": 0.0762,
+      "step": 16870
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.7198979591836735e-05,
+      "loss": 0.0239,
+      "step": 16871
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.718367346938775e-05,
+      "loss": 0.0277,
+      "step": 16872
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.7168367346938776e-05,
+      "loss": 0.0425,
+      "step": 16873
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.715306122448979e-05,
+      "loss": 0.0712,
+      "step": 16874
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.7137755102040816e-05,
+      "loss": 0.0392,
+      "step": 16875
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.712244897959183e-05,
+      "loss": 0.0786,
+      "step": 16876
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.710714285714286e-05,
+      "loss": 0.0329,
+      "step": 16877
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.7091836734693874e-05,
+      "loss": 0.0237,
+      "step": 16878
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.70765306122449e-05,
+      "loss": 0.0517,
+      "step": 16879
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.7061224489795915e-05,
+      "loss": 0.1568,
+      "step": 16880
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.704591836734693e-05,
+      "loss": 0.1194,
+      "step": 16881
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.7030612244897956e-05,
+      "loss": 0.0673,
+      "step": 16882
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.701530612244897e-05,
+      "loss": 0.1039,
+      "step": 16883
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 0.1078,
+      "step": 16884
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.698469387755101e-05,
+      "loss": 0.0715,
+      "step": 16885
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.696938775510204e-05,
+      "loss": 0.051,
+      "step": 16886
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.6954081632653054e-05,
+      "loss": 0.0675,
+      "step": 16887
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.693877551020408e-05,
+      "loss": 0.0367,
+      "step": 16888
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.6923469387755095e-05,
+      "loss": 0.0243,
+      "step": 16889
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.690816326530612e-05,
+      "loss": 0.0298,
+      "step": 16890
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.6892857142857136e-05,
+      "loss": 0.0351,
+      "step": 16891
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.687755102040816e-05,
+      "loss": 0.0862,
+      "step": 16892
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.6862244897959176e-05,
+      "loss": 0.1332,
+      "step": 16893
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.68469387755102e-05,
+      "loss": 0.0363,
+      "step": 16894
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.683163265306122e-05,
+      "loss": 0.073,
+      "step": 16895
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.681632653061224e-05,
+      "loss": 0.1098,
+      "step": 16896
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.680102040816326e-05,
+      "loss": 0.0446,
+      "step": 16897
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.678571428571428e-05,
+      "loss": 0.0545,
+      "step": 16898
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.67704081632653e-05,
+      "loss": 0.1834,
+      "step": 16899
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.675510204081632e-05,
+      "loss": 0.0726,
+      "step": 16900
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.673979591836734e-05,
+      "loss": 0.0375,
+      "step": 16901
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.6724489795918356e-05,
+      "loss": 0.041,
+      "step": 16902
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.670918367346938e-05,
+      "loss": 0.0679,
+      "step": 16903
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.66938775510204e-05,
+      "loss": 0.0664,
+      "step": 16904
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.667857142857142e-05,
+      "loss": 0.0518,
+      "step": 16905
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.666326530612244e-05,
+      "loss": 0.1112,
+      "step": 16906
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.664795918367346e-05,
+      "loss": 0.0342,
+      "step": 16907
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.663265306122448e-05,
+      "loss": 0.1474,
+      "step": 16908
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.66173469387755e-05,
+      "loss": 0.1149,
+      "step": 16909
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.660204081632652e-05,
+      "loss": 0.0246,
+      "step": 16910
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.658673469387755e-05,
+      "loss": 0.0291,
+      "step": 16911
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.657142857142857e-05,
+      "loss": 0.2068,
+      "step": 16912
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.655612244897959e-05,
+      "loss": 0.0665,
+      "step": 16913
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.654081632653061e-05,
+      "loss": 0.1448,
+      "step": 16914
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.652551020408163e-05,
+      "loss": 0.0709,
+      "step": 16915
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.651020408163265e-05,
+      "loss": 0.061,
+      "step": 16916
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.649489795918367e-05,
+      "loss": 0.0393,
+      "step": 16917
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.647959183673469e-05,
+      "loss": 0.0706,
+      "step": 16918
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.646428571428571e-05,
+      "loss": 0.0712,
+      "step": 16919
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.644897959183673e-05,
+      "loss": 0.0365,
+      "step": 16920
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.6433673469387754e-05,
+      "loss": 0.0858,
+      "step": 16921
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.641836734693877e-05,
+      "loss": 0.0229,
+      "step": 16922
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.6403061224489795e-05,
+      "loss": 0.0344,
+      "step": 16923
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.638775510204081e-05,
+      "loss": 0.1331,
+      "step": 16924
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.6372448979591835e-05,
+      "loss": 0.1154,
+      "step": 16925
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.635714285714285e-05,
+      "loss": 0.037,
+      "step": 16926
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.6341836734693876e-05,
+      "loss": 0.0299,
+      "step": 16927
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.632653061224489e-05,
+      "loss": 0.0305,
+      "step": 16928
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.631122448979592e-05,
+      "loss": 0.0339,
+      "step": 16929
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.6295918367346934e-05,
+      "loss": 0.1187,
+      "step": 16930
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.628061224489796e-05,
+      "loss": 0.0287,
+      "step": 16931
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.6265306122448975e-05,
+      "loss": 0.0741,
+      "step": 16932
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.625e-05,
+      "loss": 0.0999,
+      "step": 16933
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.6234693877551015e-05,
+      "loss": 0.0636,
+      "step": 16934
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.621938775510203e-05,
+      "loss": 0.0537,
+      "step": 16935
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.6204081632653056e-05,
+      "loss": 0.0407,
+      "step": 16936
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.618877551020407e-05,
+      "loss": 0.112,
+      "step": 16937
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.61734693877551e-05,
+      "loss": 0.0238,
+      "step": 16938
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.6158163265306114e-05,
+      "loss": 0.0362,
+      "step": 16939
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.614285714285714e-05,
+      "loss": 0.0492,
+      "step": 16940
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.6127551020408155e-05,
+      "loss": 0.0721,
+      "step": 16941
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.611224489795918e-05,
+      "loss": 0.0396,
+      "step": 16942
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.6096938775510196e-05,
+      "loss": 0.0451,
+      "step": 16943
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.608163265306122e-05,
+      "loss": 0.0739,
+      "step": 16944
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.6066326530612236e-05,
+      "loss": 0.105,
+      "step": 16945
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.605102040816326e-05,
+      "loss": 0.1488,
+      "step": 16946
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.603571428571428e-05,
+      "loss": 0.0406,
+      "step": 16947
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.60204081632653e-05,
+      "loss": 0.0355,
+      "step": 16948
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.600510204081632e-05,
+      "loss": 0.105,
+      "step": 16949
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.598979591836734e-05,
+      "loss": 0.0674,
+      "step": 16950
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.597448979591836e-05,
+      "loss": 0.0834,
+      "step": 16951
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.595918367346938e-05,
+      "loss": 0.043,
+      "step": 16952
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.59438775510204e-05,
+      "loss": 0.0229,
+      "step": 16953
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.592857142857142e-05,
+      "loss": 0.0353,
+      "step": 16954
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.591326530612244e-05,
+      "loss": 0.1952,
+      "step": 16955
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.589795918367347e-05,
+      "loss": 0.0925,
+      "step": 16956
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.588265306122449e-05,
+      "loss": 0.1139,
+      "step": 16957
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.586734693877551e-05,
+      "loss": 0.0675,
+      "step": 16958
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.585204081632653e-05,
+      "loss": 0.0772,
+      "step": 16959
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.583673469387755e-05,
+      "loss": 0.0762,
+      "step": 16960
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.582142857142857e-05,
+      "loss": 0.1012,
+      "step": 16961
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.580612244897959e-05,
+      "loss": 0.1631,
+      "step": 16962
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.579081632653061e-05,
+      "loss": 0.0763,
+      "step": 16963
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.5775510204081634e-05,
+      "loss": 0.0322,
+      "step": 16964
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.576020408163265e-05,
+      "loss": 0.0672,
+      "step": 16965
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.574489795918367e-05,
+      "loss": 0.0421,
+      "step": 16966
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.572959183673469e-05,
+      "loss": 0.0291,
+      "step": 16967
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.571428571428571e-05,
+      "loss": 0.0459,
+      "step": 16968
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.569897959183673e-05,
+      "loss": 0.03,
+      "step": 16969
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.568367346938775e-05,
+      "loss": 0.1497,
+      "step": 16970
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.566836734693877e-05,
+      "loss": 0.0523,
+      "step": 16971
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.565306122448979e-05,
+      "loss": 0.0823,
+      "step": 16972
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.5637755102040814e-05,
+      "loss": 0.1106,
+      "step": 16973
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.562244897959183e-05,
+      "loss": 0.1414,
+      "step": 16974
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.5607142857142854e-05,
+      "loss": 0.0728,
+      "step": 16975
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.559183673469387e-05,
+      "loss": 0.0758,
+      "step": 16976
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.5576530612244895e-05,
+      "loss": 0.1243,
+      "step": 16977
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.556122448979591e-05,
+      "loss": 0.0475,
+      "step": 16978
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.5545918367346936e-05,
+      "loss": 0.0378,
+      "step": 16979
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.553061224489795e-05,
+      "loss": 0.0536,
+      "step": 16980
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.551530612244898e-05,
+      "loss": 0.0395,
+      "step": 16981
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.5499999999999994e-05,
+      "loss": 0.0666,
+      "step": 16982
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.548469387755102e-05,
+      "loss": 0.035,
+      "step": 16983
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.5469387755102035e-05,
+      "loss": 0.1111,
+      "step": 16984
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.545408163265306e-05,
+      "loss": 0.0349,
+      "step": 16985
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.5438775510204075e-05,
+      "loss": 0.0722,
+      "step": 16986
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.542346938775509e-05,
+      "loss": 0.2401,
+      "step": 16987
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.5408163265306116e-05,
+      "loss": 0.0761,
+      "step": 16988
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.539285714285713e-05,
+      "loss": 0.0755,
+      "step": 16989
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.537755102040816e-05,
+      "loss": 0.079,
+      "step": 16990
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.5362244897959174e-05,
+      "loss": 0.1081,
+      "step": 16991
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.53469387755102e-05,
+      "loss": 0.0315,
+      "step": 16992
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.5331632653061215e-05,
+      "loss": 0.1395,
+      "step": 16993
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.531632653061224e-05,
+      "loss": 0.156,
+      "step": 16994
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.5301020408163255e-05,
+      "loss": 0.1466,
+      "step": 16995
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.528571428571428e-05,
+      "loss": 0.037,
+      "step": 16996
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.5270408163265296e-05,
+      "loss": 0.11,
+      "step": 16997
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.525510204081632e-05,
+      "loss": 0.09,
+      "step": 16998
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.523979591836734e-05,
+      "loss": 0.1339,
+      "step": 16999
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.522448979591836e-05,
+      "loss": 0.1171,
+      "step": 17000
+    },
+    {
+      "epoch": 16.5,
+      "eval_cer": 0.23955184715526687,
+      "eval_loss": 1.1302112340927124,
+      "eval_runtime": 227.6364,
+      "eval_samples_per_second": 38.72,
+      "eval_steps_per_second": 4.841,
+      "eval_wer": 0.6314910632875238,
+      "step": 17000
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.520918367346938e-05,
+      "loss": 0.0338,
+      "step": 17001
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.519387755102041e-05,
+      "loss": 0.0238,
+      "step": 17002
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.5178571428571425e-05,
+      "loss": 0.0285,
+      "step": 17003
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.516326530612245e-05,
+      "loss": 0.0289,
+      "step": 17004
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.5147959183673466e-05,
+      "loss": 0.0783,
+      "step": 17005
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.513265306122449e-05,
+      "loss": 0.0833,
+      "step": 17006
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.511734693877551e-05,
+      "loss": 0.1058,
+      "step": 17007
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.510204081632653e-05,
+      "loss": 0.1132,
+      "step": 17008
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.508673469387755e-05,
+      "loss": 0.107,
+      "step": 17009
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.507142857142857e-05,
+      "loss": 0.0667,
+      "step": 17010
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.505612244897959e-05,
+      "loss": 0.1235,
+      "step": 17011
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.504081632653061e-05,
+      "loss": 0.221,
+      "step": 17012
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.502551020408163e-05,
+      "loss": 0.2759,
+      "step": 17013
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.501020408163265e-05,
+      "loss": 0.0329,
+      "step": 17014
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.499489795918367e-05,
+      "loss": 0.033,
+      "step": 17015
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.4979591836734694e-05,
+      "loss": 0.048,
+      "step": 17016
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.496428571428571e-05,
+      "loss": 0.0917,
+      "step": 17017
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.4948979591836734e-05,
+      "loss": 0.1475,
+      "step": 17018
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.493367346938775e-05,
+      "loss": 0.0705,
+      "step": 17019
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.491836734693877e-05,
+      "loss": 0.0736,
+      "step": 17020
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.490306122448979e-05,
+      "loss": 0.037,
+      "step": 17021
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.488775510204081e-05,
+      "loss": 0.1043,
+      "step": 17022
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.487244897959183e-05,
+      "loss": 0.0687,
+      "step": 17023
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.485714285714285e-05,
+      "loss": 0.1077,
+      "step": 17024
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.4841836734693874e-05,
+      "loss": 0.0324,
+      "step": 17025
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.482653061224489e-05,
+      "loss": 0.1142,
+      "step": 17026
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.4811224489795914e-05,
+      "loss": 0.0489,
+      "step": 17027
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.479591836734693e-05,
+      "loss": 0.0895,
+      "step": 17028
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.4780612244897955e-05,
+      "loss": 0.0269,
+      "step": 17029
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.476530612244897e-05,
+      "loss": 0.162,
+      "step": 17030
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.4749999999999996e-05,
+      "loss": 0.0781,
+      "step": 17031
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.473469387755101e-05,
+      "loss": 0.1017,
+      "step": 17032
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.471938775510204e-05,
+      "loss": 0.1138,
+      "step": 17033
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.4704081632653054e-05,
+      "loss": 0.0288,
+      "step": 17034
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.468877551020408e-05,
+      "loss": 0.0427,
+      "step": 17035
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.4673469387755094e-05,
+      "loss": 0.0971,
+      "step": 17036
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.465816326530612e-05,
+      "loss": 0.1561,
+      "step": 17037
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.4642857142857135e-05,
+      "loss": 0.0586,
+      "step": 17038
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.462755102040816e-05,
+      "loss": 0.035,
+      "step": 17039
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.4612244897959176e-05,
+      "loss": 0.1095,
+      "step": 17040
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.459693877551019e-05,
+      "loss": 0.1155,
+      "step": 17041
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.458163265306122e-05,
+      "loss": 0.0716,
+      "step": 17042
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.4566326530612234e-05,
+      "loss": 0.1076,
+      "step": 17043
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.455102040816326e-05,
+      "loss": 0.1271,
+      "step": 17044
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.4535714285714274e-05,
+      "loss": 0.0996,
+      "step": 17045
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.45204081632653e-05,
+      "loss": 0.1039,
+      "step": 17046
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.450510204081633e-05,
+      "loss": 0.0819,
+      "step": 17047
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.4489795918367346e-05,
+      "loss": 0.0276,
+      "step": 17048
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.447448979591837e-05,
+      "loss": 0.1511,
+      "step": 17049
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.4459183673469387e-05,
+      "loss": 0.0699,
+      "step": 17050
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.4443877551020404e-05,
+      "loss": 0.0672,
+      "step": 17051
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.442857142857143e-05,
+      "loss": 0.0762,
+      "step": 17052
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.4413265306122444e-05,
+      "loss": 0.0331,
+      "step": 17053
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.439795918367347e-05,
+      "loss": 0.0272,
+      "step": 17054
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.4382653061224485e-05,
+      "loss": 0.0976,
+      "step": 17055
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.436734693877551e-05,
+      "loss": 0.105,
+      "step": 17056
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.4352040816326526e-05,
+      "loss": 0.1155,
+      "step": 17057
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.433673469387755e-05,
+      "loss": 0.068,
+      "step": 17058
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.4321428571428567e-05,
+      "loss": 0.0635,
+      "step": 17059
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.430612244897959e-05,
+      "loss": 0.0368,
+      "step": 17060
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.429081632653061e-05,
+      "loss": 0.0697,
+      "step": 17061
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.427551020408163e-05,
+      "loss": 0.2845,
+      "step": 17062
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.426020408163265e-05,
+      "loss": 0.0334,
+      "step": 17063
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.424489795918367e-05,
+      "loss": 0.0227,
+      "step": 17064
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.422959183673469e-05,
+      "loss": 0.1075,
+      "step": 17065
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.421428571428571e-05,
+      "loss": 0.148,
+      "step": 17066
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.419897959183673e-05,
+      "loss": 0.1342,
+      "step": 17067
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.4183673469387753e-05,
+      "loss": 0.0694,
+      "step": 17068
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.416836734693877e-05,
+      "loss": 0.0852,
+      "step": 17069
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.4153061224489794e-05,
+      "loss": 0.1158,
+      "step": 17070
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.413775510204081e-05,
+      "loss": 0.081,
+      "step": 17071
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.412244897959183e-05,
+      "loss": 0.0365,
+      "step": 17072
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.410714285714285e-05,
+      "loss": 0.1007,
+      "step": 17073
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.409183673469387e-05,
+      "loss": 0.0937,
+      "step": 17074
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.407653061224489e-05,
+      "loss": 0.1848,
+      "step": 17075
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.406122448979591e-05,
+      "loss": 0.0691,
+      "step": 17076
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.4045918367346933e-05,
+      "loss": 0.0666,
+      "step": 17077
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.403061224489795e-05,
+      "loss": 0.0596,
+      "step": 17078
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.4015306122448974e-05,
+      "loss": 0.0685,
+      "step": 17079
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.399999999999999e-05,
+      "loss": 0.0661,
+      "step": 17080
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.3984693877551015e-05,
+      "loss": 0.0318,
+      "step": 17081
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.396938775510203e-05,
+      "loss": 0.1103,
+      "step": 17082
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.3954081632653056e-05,
+      "loss": 0.0748,
+      "step": 17083
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.393877551020407e-05,
+      "loss": 0.1415,
+      "step": 17084
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.3923469387755097e-05,
+      "loss": 0.0795,
+      "step": 17085
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.3908163265306113e-05,
+      "loss": 0.0667,
+      "step": 17086
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.389285714285714e-05,
+      "loss": 0.1511,
+      "step": 17087
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.3877551020408154e-05,
+      "loss": 0.0932,
+      "step": 17088
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.386224489795918e-05,
+      "loss": 0.073,
+      "step": 17089
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.3846938775510195e-05,
+      "loss": 0.1143,
+      "step": 17090
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.383163265306122e-05,
+      "loss": 0.0291,
+      "step": 17091
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.381632653061224e-05,
+      "loss": 0.0712,
+      "step": 17092
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.3801020408163266e-05,
+      "loss": 0.0723,
+      "step": 17093
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.378571428571428e-05,
+      "loss": 0.0804,
+      "step": 17094
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.377040816326531e-05,
+      "loss": 0.1092,
+      "step": 17095
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.3755102040816324e-05,
+      "loss": 0.0752,
+      "step": 17096
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.373979591836735e-05,
+      "loss": 0.0314,
+      "step": 17097
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.3724489795918365e-05,
+      "loss": 0.0619,
+      "step": 17098
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.370918367346939e-05,
+      "loss": 0.0719,
+      "step": 17099
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.3693877551020406e-05,
+      "loss": 0.0314,
+      "step": 17100
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.367857142857143e-05,
+      "loss": 0.0368,
+      "step": 17101
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.3663265306122446e-05,
+      "loss": 0.0824,
+      "step": 17102
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.364795918367347e-05,
+      "loss": 0.068,
+      "step": 17103
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.363265306122449e-05,
+      "loss": 0.0993,
+      "step": 17104
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.3617346938775504e-05,
+      "loss": 0.1496,
+      "step": 17105
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.360204081632653e-05,
+      "loss": 0.1638,
+      "step": 17106
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.3586734693877545e-05,
+      "loss": 0.0349,
+      "step": 17107
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.357142857142857e-05,
+      "loss": 0.0422,
+      "step": 17108
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.3556122448979586e-05,
+      "loss": 0.0289,
+      "step": 17109
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.354081632653061e-05,
+      "loss": 0.0395,
+      "step": 17110
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.3525510204081626e-05,
+      "loss": 0.0853,
+      "step": 17111
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.351020408163265e-05,
+      "loss": 0.0725,
+      "step": 17112
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.349489795918367e-05,
+      "loss": 0.0791,
+      "step": 17113
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.347959183673469e-05,
+      "loss": 0.0332,
+      "step": 17114
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.346428571428571e-05,
+      "loss": 0.1073,
+      "step": 17115
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.344897959183673e-05,
+      "loss": 0.0979,
+      "step": 17116
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.343367346938775e-05,
+      "loss": 0.0318,
+      "step": 17117
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.341836734693877e-05,
+      "loss": 0.0675,
+      "step": 17118
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.340306122448979e-05,
+      "loss": 0.1954,
+      "step": 17119
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.338775510204081e-05,
+      "loss": 0.0216,
+      "step": 17120
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.337244897959183e-05,
+      "loss": 0.0836,
+      "step": 17121
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.3357142857142854e-05,
+      "loss": 0.1527,
+      "step": 17122
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.334183673469387e-05,
+      "loss": 0.0379,
+      "step": 17123
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.3326530612244895e-05,
+      "loss": 0.1265,
+      "step": 17124
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.331122448979591e-05,
+      "loss": 0.0648,
+      "step": 17125
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.329591836734693e-05,
+      "loss": 0.0403,
+      "step": 17126
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.328061224489795e-05,
+      "loss": 0.0582,
+      "step": 17127
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.326530612244897e-05,
+      "loss": 0.0284,
+      "step": 17128
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.324999999999999e-05,
+      "loss": 0.0777,
+      "step": 17129
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.323469387755101e-05,
+      "loss": 0.0388,
+      "step": 17130
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.3219387755102034e-05,
+      "loss": 0.0399,
+      "step": 17131
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.320408163265305e-05,
+      "loss": 0.0348,
+      "step": 17132
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.3188775510204075e-05,
+      "loss": 0.1063,
+      "step": 17133
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.317346938775509e-05,
+      "loss": 0.0793,
+      "step": 17134
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.3158163265306116e-05,
+      "loss": 0.0733,
+      "step": 17135
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.314285714285713e-05,
+      "loss": 0.0305,
+      "step": 17136
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.3127551020408156e-05,
+      "loss": 0.0958,
+      "step": 17137
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.311224489795918e-05,
+      "loss": 0.0919,
+      "step": 17138
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.3096938775510204e-05,
+      "loss": 0.085,
+      "step": 17139
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.308163265306122e-05,
+      "loss": 0.0295,
+      "step": 17140
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.3066326530612245e-05,
+      "loss": 0.0667,
+      "step": 17141
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.305102040816326e-05,
+      "loss": 0.0661,
+      "step": 17142
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.3035714285714285e-05,
+      "loss": 0.172,
+      "step": 17143
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.30204081632653e-05,
+      "loss": 0.0689,
+      "step": 17144
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.3005102040816326e-05,
+      "loss": 0.0461,
+      "step": 17145
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.298979591836734e-05,
+      "loss": 0.0371,
+      "step": 17146
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.297448979591837e-05,
+      "loss": 0.0292,
+      "step": 17147
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.2959183673469384e-05,
+      "loss": 0.0441,
+      "step": 17148
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.294387755102041e-05,
+      "loss": 0.1875,
+      "step": 17149
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.2928571428571425e-05,
+      "loss": 0.0711,
+      "step": 17150
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.291326530612245e-05,
+      "loss": 0.0321,
+      "step": 17151
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.2897959183673465e-05,
+      "loss": 0.0304,
+      "step": 17152
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.288265306122449e-05,
+      "loss": 0.1049,
+      "step": 17153
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.2867346938775506e-05,
+      "loss": 0.0144,
+      "step": 17154
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.285204081632653e-05,
+      "loss": 0.139,
+      "step": 17155
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.283673469387755e-05,
+      "loss": 0.0822,
+      "step": 17156
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.2821428571428564e-05,
+      "loss": 0.0263,
+      "step": 17157
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.280612244897959e-05,
+      "loss": 0.0289,
+      "step": 17158
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.2790816326530605e-05,
+      "loss": 0.0687,
+      "step": 17159
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.277551020408163e-05,
+      "loss": 0.0553,
+      "step": 17160
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.2760204081632646e-05,
+      "loss": 0.0636,
+      "step": 17161
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.274489795918367e-05,
+      "loss": 0.1952,
+      "step": 17162
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.2729591836734686e-05,
+      "loss": 0.0288,
+      "step": 17163
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.271428571428571e-05,
+      "loss": 0.1198,
+      "step": 17164
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.269897959183673e-05,
+      "loss": 0.115,
+      "step": 17165
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.268367346938775e-05,
+      "loss": 0.116,
+      "step": 17166
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.266836734693877e-05,
+      "loss": 0.095,
+      "step": 17167
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.265306122448979e-05,
+      "loss": 0.1442,
+      "step": 17168
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.263775510204081e-05,
+      "loss": 0.0778,
+      "step": 17169
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.262244897959183e-05,
+      "loss": 0.0881,
+      "step": 17170
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.260714285714285e-05,
+      "loss": 0.0702,
+      "step": 17171
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.259183673469387e-05,
+      "loss": 0.1193,
+      "step": 17172
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.257653061224489e-05,
+      "loss": 0.1045,
+      "step": 17173
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.2561224489795914e-05,
+      "loss": 0.1513,
+      "step": 17174
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.254591836734693e-05,
+      "loss": 0.0716,
+      "step": 17175
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.2530612244897955e-05,
+      "loss": 0.0348,
+      "step": 17176
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.251530612244897e-05,
+      "loss": 0.1857,
+      "step": 17177
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.2499999999999995e-05,
+      "loss": 0.0413,
+      "step": 17178
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.248469387755101e-05,
+      "loss": 0.0917,
+      "step": 17179
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.246938775510203e-05,
+      "loss": 0.1187,
+      "step": 17180
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.245408163265305e-05,
+      "loss": 0.0781,
+      "step": 17181
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.243877551020407e-05,
+      "loss": 0.1775,
+      "step": 17182
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.24234693877551e-05,
+      "loss": 0.1209,
+      "step": 17183
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.2408163265306124e-05,
+      "loss": 0.0719,
+      "step": 17184
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.239285714285714e-05,
+      "loss": 0.0638,
+      "step": 17185
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.2377551020408165e-05,
+      "loss": 0.0604,
+      "step": 17186
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.236224489795918e-05,
+      "loss": 0.1213,
+      "step": 17187
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.2346938775510206e-05,
+      "loss": 0.1109,
+      "step": 17188
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.233163265306122e-05,
+      "loss": 0.0791,
+      "step": 17189
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.231632653061224e-05,
+      "loss": 0.0247,
+      "step": 17190
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.2301020408163264e-05,
+      "loss": 0.1549,
+      "step": 17191
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.228571428571428e-05,
+      "loss": 0.0316,
+      "step": 17192
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.2270408163265305e-05,
+      "loss": 0.1352,
+      "step": 17193
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.225510204081632e-05,
+      "loss": 0.1162,
+      "step": 17194
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.2239795918367345e-05,
+      "loss": 0.1053,
+      "step": 17195
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.222448979591836e-05,
+      "loss": 0.0844,
+      "step": 17196
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.2209183673469386e-05,
+      "loss": 0.0379,
+      "step": 17197
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.21938775510204e-05,
+      "loss": 0.1718,
+      "step": 17198
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.217857142857143e-05,
+      "loss": 0.1469,
+      "step": 17199
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.2163265306122444e-05,
+      "loss": 0.1175,
+      "step": 17200
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.214795918367347e-05,
+      "loss": 0.0674,
+      "step": 17201
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.2132653061224485e-05,
+      "loss": 0.0887,
+      "step": 17202
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.211734693877551e-05,
+      "loss": 0.026,
+      "step": 17203
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.2102040816326525e-05,
+      "loss": 0.0747,
+      "step": 17204
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.208673469387755e-05,
+      "loss": 0.1228,
+      "step": 17205
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.2071428571428566e-05,
+      "loss": 0.0843,
+      "step": 17206
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.205612244897959e-05,
+      "loss": 0.0753,
+      "step": 17207
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.204081632653061e-05,
+      "loss": 0.0713,
+      "step": 17208
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.202551020408163e-05,
+      "loss": 0.064,
+      "step": 17209
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.201020408163265e-05,
+      "loss": 0.0328,
+      "step": 17210
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.1994897959183665e-05,
+      "loss": 0.0482,
+      "step": 17211
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.197959183673469e-05,
+      "loss": 0.0791,
+      "step": 17212
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.1964285714285705e-05,
+      "loss": 0.0735,
+      "step": 17213
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.194897959183673e-05,
+      "loss": 0.1559,
+      "step": 17214
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.1933673469387746e-05,
+      "loss": 0.046,
+      "step": 17215
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.191836734693877e-05,
+      "loss": 0.0915,
+      "step": 17216
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.190306122448979e-05,
+      "loss": 0.1096,
+      "step": 17217
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.188775510204081e-05,
+      "loss": 0.1541,
+      "step": 17218
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.187244897959183e-05,
+      "loss": 0.0432,
+      "step": 17219
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.185714285714285e-05,
+      "loss": 0.0322,
+      "step": 17220
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.184183673469387e-05,
+      "loss": 0.0266,
+      "step": 17221
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.182653061224489e-05,
+      "loss": 0.0637,
+      "step": 17222
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.181122448979591e-05,
+      "loss": 0.0338,
+      "step": 17223
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.179591836734693e-05,
+      "loss": 0.0425,
+      "step": 17224
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.178061224489795e-05,
+      "loss": 0.0443,
+      "step": 17225
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.1765306122448974e-05,
+      "loss": 0.1038,
+      "step": 17226
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.174999999999999e-05,
+      "loss": 0.0656,
+      "step": 17227
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.173469387755102e-05,
+      "loss": 0.0509,
+      "step": 17228
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.171938775510204e-05,
+      "loss": 0.0987,
+      "step": 17229
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.170408163265306e-05,
+      "loss": 0.0631,
+      "step": 17230
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.168877551020408e-05,
+      "loss": 0.0809,
+      "step": 17231
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.16734693877551e-05,
+      "loss": 0.0423,
+      "step": 17232
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.165816326530612e-05,
+      "loss": 0.105,
+      "step": 17233
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.1642857142857144e-05,
+      "loss": 0.0603,
+      "step": 17234
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.162755102040816e-05,
+      "loss": 0.071,
+      "step": 17235
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.1612244897959184e-05,
+      "loss": 0.1007,
+      "step": 17236
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.15969387755102e-05,
+      "loss": 0.1099,
+      "step": 17237
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.1581632653061225e-05,
+      "loss": 0.1578,
+      "step": 17238
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.156632653061224e-05,
+      "loss": 0.0283,
+      "step": 17239
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.1551020408163266e-05,
+      "loss": 0.1173,
+      "step": 17240
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.153571428571428e-05,
+      "loss": 0.042,
+      "step": 17241
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.15204081632653e-05,
+      "loss": 0.0273,
+      "step": 17242
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.1505102040816324e-05,
+      "loss": 0.1093,
+      "step": 17243
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.148979591836734e-05,
+      "loss": 0.0414,
+      "step": 17244
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.1474489795918364e-05,
+      "loss": 0.1541,
+      "step": 17245
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.145918367346938e-05,
+      "loss": 0.1229,
+      "step": 17246
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.1443877551020405e-05,
+      "loss": 0.1163,
+      "step": 17247
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.142857142857142e-05,
+      "loss": 0.0728,
+      "step": 17248
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.1413265306122446e-05,
+      "loss": 0.1284,
+      "step": 17249
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.139795918367346e-05,
+      "loss": 0.077,
+      "step": 17250
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.138265306122449e-05,
+      "loss": 0.0227,
+      "step": 17251
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.1367346938775504e-05,
+      "loss": 0.0318,
+      "step": 17252
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.135204081632653e-05,
+      "loss": 0.1225,
+      "step": 17253
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.1336734693877544e-05,
+      "loss": 0.1455,
+      "step": 17254
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.132142857142857e-05,
+      "loss": 0.1481,
+      "step": 17255
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.1306122448979585e-05,
+      "loss": 0.1508,
+      "step": 17256
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.129081632653061e-05,
+      "loss": 0.0732,
+      "step": 17257
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.1275510204081626e-05,
+      "loss": 0.0736,
+      "step": 17258
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.126020408163265e-05,
+      "loss": 0.0557,
+      "step": 17259
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.124489795918367e-05,
+      "loss": 0.0289,
+      "step": 17260
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.122959183673469e-05,
+      "loss": 0.1124,
+      "step": 17261
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.121428571428571e-05,
+      "loss": 0.1316,
+      "step": 17262
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.119897959183673e-05,
+      "loss": 0.1553,
+      "step": 17263
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.118367346938775e-05,
+      "loss": 0.1403,
+      "step": 17264
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.1168367346938765e-05,
+      "loss": 0.0774,
+      "step": 17265
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.115306122448979e-05,
+      "loss": 0.0284,
+      "step": 17266
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.1137755102040806e-05,
+      "loss": 0.066,
+      "step": 17267
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.112244897959183e-05,
+      "loss": 0.0965,
+      "step": 17268
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.110714285714285e-05,
+      "loss": 0.0841,
+      "step": 17269
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.109183673469387e-05,
+      "loss": 0.0772,
+      "step": 17270
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.107653061224489e-05,
+      "loss": 0.0749,
+      "step": 17271
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.106122448979591e-05,
+      "loss": 0.1304,
+      "step": 17272
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.104591836734693e-05,
+      "loss": 0.054,
+      "step": 17273
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.103061224489796e-05,
+      "loss": 0.1548,
+      "step": 17274
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.1015306122448976e-05,
+      "loss": 0.0797,
+      "step": 17275
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.1e-05,
+      "loss": 0.1189,
+      "step": 17276
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.0984693877551017e-05,
+      "loss": 0.1117,
+      "step": 17277
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.096938775510204e-05,
+      "loss": 0.073,
+      "step": 17278
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.095408163265306e-05,
+      "loss": 0.1014,
+      "step": 17279
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.093877551020408e-05,
+      "loss": 0.1027,
+      "step": 17280
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.09234693877551e-05,
+      "loss": 0.0363,
+      "step": 17281
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.090816326530612e-05,
+      "loss": 0.108,
+      "step": 17282
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.089285714285714e-05,
+      "loss": 0.0723,
+      "step": 17283
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.087755102040816e-05,
+      "loss": 0.0309,
+      "step": 17284
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.086224489795918e-05,
+      "loss": 0.1106,
+      "step": 17285
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.0846938775510203e-05,
+      "loss": 0.0731,
+      "step": 17286
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.083163265306122e-05,
+      "loss": 0.0796,
+      "step": 17287
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.0816326530612244e-05,
+      "loss": 0.0702,
+      "step": 17288
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.080102040816326e-05,
+      "loss": 0.0741,
+      "step": 17289
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.0785714285714285e-05,
+      "loss": 0.0323,
+      "step": 17290
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.07704081632653e-05,
+      "loss": 0.0331,
+      "step": 17291
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.0755102040816326e-05,
+      "loss": 0.0275,
+      "step": 17292
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.073979591836734e-05,
+      "loss": 0.2231,
+      "step": 17293
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.0724489795918366e-05,
+      "loss": 0.0687,
+      "step": 17294
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.0709183673469383e-05,
+      "loss": 0.1268,
+      "step": 17295
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.06938775510204e-05,
+      "loss": 0.0329,
+      "step": 17296
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.0678571428571424e-05,
+      "loss": 0.0251,
+      "step": 17297
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.066326530612244e-05,
+      "loss": 0.0714,
+      "step": 17298
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.0647959183673465e-05,
+      "loss": 0.2506,
+      "step": 17299
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.063265306122448e-05,
+      "loss": 0.1226,
+      "step": 17300
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.0617346938775506e-05,
+      "loss": 0.0767,
+      "step": 17301
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.060204081632652e-05,
+      "loss": 0.1178,
+      "step": 17302
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.0586734693877547e-05,
+      "loss": 0.1568,
+      "step": 17303
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.0571428571428564e-05,
+      "loss": 0.065,
+      "step": 17304
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.055612244897959e-05,
+      "loss": 0.1235,
+      "step": 17305
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.0540816326530604e-05,
+      "loss": 0.0818,
+      "step": 17306
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.052551020408163e-05,
+      "loss": 0.1998,
+      "step": 17307
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.0510204081632645e-05,
+      "loss": 0.0833,
+      "step": 17308
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.049489795918367e-05,
+      "loss": 0.0667,
+      "step": 17309
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.0479591836734686e-05,
+      "loss": 0.0249,
+      "step": 17310
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.046428571428571e-05,
+      "loss": 0.1091,
+      "step": 17311
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.0448979591836727e-05,
+      "loss": 0.0737,
+      "step": 17312
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.043367346938775e-05,
+      "loss": 0.1194,
+      "step": 17313
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.041836734693877e-05,
+      "loss": 0.031,
+      "step": 17314
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.040306122448979e-05,
+      "loss": 0.0378,
+      "step": 17315
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.038775510204081e-05,
+      "loss": 0.0659,
+      "step": 17316
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.0372448979591825e-05,
+      "loss": 0.0979,
+      "step": 17317
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.035714285714285e-05,
+      "loss": 0.1144,
+      "step": 17318
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.034183673469388e-05,
+      "loss": 0.1309,
+      "step": 17319
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.0326530612244896e-05,
+      "loss": 0.0314,
+      "step": 17320
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.031122448979592e-05,
+      "loss": 0.1116,
+      "step": 17321
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.029591836734694e-05,
+      "loss": 0.0366,
+      "step": 17322
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.028061224489796e-05,
+      "loss": 0.0693,
+      "step": 17323
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.026530612244898e-05,
+      "loss": 0.0891,
+      "step": 17324
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.025e-05,
+      "loss": 0.1986,
+      "step": 17325
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.023469387755102e-05,
+      "loss": 0.0377,
+      "step": 17326
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.0219387755102036e-05,
+      "loss": 0.034,
+      "step": 17327
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.020408163265306e-05,
+      "loss": 0.0463,
+      "step": 17328
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.0188775510204076e-05,
+      "loss": 0.0258,
+      "step": 17329
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.01734693877551e-05,
+      "loss": 0.1961,
+      "step": 17330
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.015816326530612e-05,
+      "loss": 0.084,
+      "step": 17331
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.014285714285714e-05,
+      "loss": 0.0398,
+      "step": 17332
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.012755102040816e-05,
+      "loss": 0.144,
+      "step": 17333
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.011224489795918e-05,
+      "loss": 0.072,
+      "step": 17334
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.00969387755102e-05,
+      "loss": 0.0989,
+      "step": 17335
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.008163265306122e-05,
+      "loss": 0.0663,
+      "step": 17336
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.006632653061224e-05,
+      "loss": 0.0865,
+      "step": 17337
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.005102040816326e-05,
+      "loss": 0.0251,
+      "step": 17338
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.003571428571428e-05,
+      "loss": 0.1153,
+      "step": 17339
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.0020408163265304e-05,
+      "loss": 0.0338,
+      "step": 17340
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 5.000510204081632e-05,
+      "loss": 0.0747,
+      "step": 17341
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.9989795918367345e-05,
+      "loss": 0.0322,
+      "step": 17342
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.997448979591836e-05,
+      "loss": 0.1512,
+      "step": 17343
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.9959183673469386e-05,
+      "loss": 0.0468,
+      "step": 17344
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.99438775510204e-05,
+      "loss": 0.1136,
+      "step": 17345
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.9928571428571426e-05,
+      "loss": 0.0747,
+      "step": 17346
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.991326530612244e-05,
+      "loss": 0.0349,
+      "step": 17347
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.989795918367347e-05,
+      "loss": 0.0341,
+      "step": 17348
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.9882653061224484e-05,
+      "loss": 0.1856,
+      "step": 17349
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.98673469387755e-05,
+      "loss": 0.0709,
+      "step": 17350
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.9852040816326525e-05,
+      "loss": 0.0405,
+      "step": 17351
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.983673469387754e-05,
+      "loss": 0.0346,
+      "step": 17352
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.9821428571428566e-05,
+      "loss": 0.0301,
+      "step": 17353
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.980612244897958e-05,
+      "loss": 0.0527,
+      "step": 17354
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.9790816326530606e-05,
+      "loss": 0.0979,
+      "step": 17355
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.977551020408162e-05,
+      "loss": 0.0337,
+      "step": 17356
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.976020408163265e-05,
+      "loss": 0.0301,
+      "step": 17357
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.9744897959183664e-05,
+      "loss": 0.037,
+      "step": 17358
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.972959183673469e-05,
+      "loss": 0.0746,
+      "step": 17359
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.9714285714285705e-05,
+      "loss": 0.0405,
+      "step": 17360
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.969897959183673e-05,
+      "loss": 0.1029,
+      "step": 17361
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.9683673469387746e-05,
+      "loss": 0.0369,
+      "step": 17362
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.966836734693877e-05,
+      "loss": 0.0877,
+      "step": 17363
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.9653061224489786e-05,
+      "loss": 0.1026,
+      "step": 17364
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.963775510204082e-05,
+      "loss": 0.0257,
+      "step": 17365
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.9622448979591834e-05,
+      "loss": 0.0366,
+      "step": 17366
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.960714285714286e-05,
+      "loss": 0.0331,
+      "step": 17367
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.9591836734693875e-05,
+      "loss": 0.2321,
+      "step": 17368
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.95765306122449e-05,
+      "loss": 0.0646,
+      "step": 17369
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.9561224489795915e-05,
+      "loss": 0.0261,
+      "step": 17370
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.954591836734694e-05,
+      "loss": 0.0339,
+      "step": 17371
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.9530612244897956e-05,
+      "loss": 0.0322,
+      "step": 17372
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.951530612244898e-05,
+      "loss": 0.0592,
+      "step": 17373
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.95e-05,
+      "loss": 0.093,
+      "step": 17374
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.948469387755102e-05,
+      "loss": 0.0747,
+      "step": 17375
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.946938775510204e-05,
+      "loss": 0.0755,
+      "step": 17376
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.945408163265306e-05,
+      "loss": 0.1061,
+      "step": 17377
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.943877551020408e-05,
+      "loss": 0.1098,
+      "step": 17378
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.94234693877551e-05,
+      "loss": 0.0661,
+      "step": 17379
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.940816326530612e-05,
+      "loss": 0.1804,
+      "step": 17380
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.9392857142857136e-05,
+      "loss": 0.1117,
+      "step": 17381
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.937755102040816e-05,
+      "loss": 0.0736,
+      "step": 17382
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.936224489795918e-05,
+      "loss": 0.0703,
+      "step": 17383
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.93469387755102e-05,
+      "loss": 0.0824,
+      "step": 17384
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.933163265306122e-05,
+      "loss": 0.029,
+      "step": 17385
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.931632653061224e-05,
+      "loss": 0.1215,
+      "step": 17386
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.930102040816326e-05,
+      "loss": 0.0772,
+      "step": 17387
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.928571428571428e-05,
+      "loss": 0.1525,
+      "step": 17388
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.92704081632653e-05,
+      "loss": 0.1077,
+      "step": 17389
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.925510204081632e-05,
+      "loss": 0.0405,
+      "step": 17390
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.923979591836734e-05,
+      "loss": 0.0734,
+      "step": 17391
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.9224489795918364e-05,
+      "loss": 0.0968,
+      "step": 17392
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.920918367346938e-05,
+      "loss": 0.2004,
+      "step": 17393
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.9193877551020405e-05,
+      "loss": 0.117,
+      "step": 17394
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.917857142857142e-05,
+      "loss": 0.048,
+      "step": 17395
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.9163265306122445e-05,
+      "loss": 0.0665,
+      "step": 17396
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.914795918367346e-05,
+      "loss": 0.107,
+      "step": 17397
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.9132653061224486e-05,
+      "loss": 0.1034,
+      "step": 17398
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.91173469387755e-05,
+      "loss": 0.1838,
+      "step": 17399
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.910204081632653e-05,
+      "loss": 0.114,
+      "step": 17400
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.9086734693877544e-05,
+      "loss": 0.0664,
+      "step": 17401
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.907142857142857e-05,
+      "loss": 0.0535,
+      "step": 17402
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.9056122448979585e-05,
+      "loss": 0.0305,
+      "step": 17403
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.90408163265306e-05,
+      "loss": 0.0408,
+      "step": 17404
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.9025510204081625e-05,
+      "loss": 0.047,
+      "step": 17405
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.901020408163264e-05,
+      "loss": 0.1183,
+      "step": 17406
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.8994897959183666e-05,
+      "loss": 0.095,
+      "step": 17407
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.897959183673468e-05,
+      "loss": 0.034,
+      "step": 17408
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.896428571428571e-05,
+      "loss": 0.0231,
+      "step": 17409
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.894897959183674e-05,
+      "loss": 0.0298,
+      "step": 17410
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.8933673469387755e-05,
+      "loss": 0.0941,
+      "step": 17411
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.891836734693877e-05,
+      "loss": 0.1122,
+      "step": 17412
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.8903061224489795e-05,
+      "loss": 0.0798,
+      "step": 17413
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.888775510204081e-05,
+      "loss": 0.1022,
+      "step": 17414
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.8872448979591836e-05,
+      "loss": 0.119,
+      "step": 17415
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.885714285714285e-05,
+      "loss": 0.0229,
+      "step": 17416
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.884183673469388e-05,
+      "loss": 0.0521,
+      "step": 17417
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.8826530612244894e-05,
+      "loss": 0.1389,
+      "step": 17418
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.881122448979592e-05,
+      "loss": 0.0754,
+      "step": 17419
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.8795918367346935e-05,
+      "loss": 0.0787,
+      "step": 17420
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.878061224489796e-05,
+      "loss": 0.1385,
+      "step": 17421
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.8765306122448975e-05,
+      "loss": 0.1058,
+      "step": 17422
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.875e-05,
+      "loss": 0.033,
+      "step": 17423
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.8734693877551016e-05,
+      "loss": 0.113,
+      "step": 17424
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.871938775510204e-05,
+      "loss": 0.0322,
+      "step": 17425
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.870408163265306e-05,
+      "loss": 0.0376,
+      "step": 17426
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.868877551020408e-05,
+      "loss": 0.0371,
+      "step": 17427
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.86734693877551e-05,
+      "loss": 0.0351,
+      "step": 17428
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.865816326530612e-05,
+      "loss": 0.0253,
+      "step": 17429
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.864285714285714e-05,
+      "loss": 0.0347,
+      "step": 17430
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.862755102040816e-05,
+      "loss": 0.1528,
+      "step": 17431
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.861224489795918e-05,
+      "loss": 0.0205,
+      "step": 17432
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.85969387755102e-05,
+      "loss": 0.1095,
+      "step": 17433
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.858163265306122e-05,
+      "loss": 0.1073,
+      "step": 17434
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.856632653061224e-05,
+      "loss": 0.0885,
+      "step": 17435
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.855102040816326e-05,
+      "loss": 0.0733,
+      "step": 17436
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.853571428571428e-05,
+      "loss": 0.227,
+      "step": 17437
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.85204081632653e-05,
+      "loss": 0.1191,
+      "step": 17438
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.850510204081632e-05,
+      "loss": 0.0374,
+      "step": 17439
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.848979591836734e-05,
+      "loss": 0.0856,
+      "step": 17440
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.847448979591836e-05,
+      "loss": 0.0511,
+      "step": 17441
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.845918367346938e-05,
+      "loss": 0.0909,
+      "step": 17442
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.84438775510204e-05,
+      "loss": 0.0555,
+      "step": 17443
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.8428571428571424e-05,
+      "loss": 0.0752,
+      "step": 17444
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.841326530612244e-05,
+      "loss": 0.0304,
+      "step": 17445
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.8397959183673465e-05,
+      "loss": 0.1107,
+      "step": 17446
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.838265306122448e-05,
+      "loss": 0.071,
+      "step": 17447
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.8367346938775505e-05,
+      "loss": 0.0818,
+      "step": 17448
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.835204081632652e-05,
+      "loss": 0.0735,
+      "step": 17449
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.8336734693877546e-05,
+      "loss": 0.1208,
+      "step": 17450
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.832142857142856e-05,
+      "loss": 0.1085,
+      "step": 17451
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.830612244897959e-05,
+      "loss": 0.0288,
+      "step": 17452
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.8290816326530604e-05,
+      "loss": 0.0355,
+      "step": 17453
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.827551020408163e-05,
+      "loss": 0.0687,
+      "step": 17454
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.826020408163265e-05,
+      "loss": 0.0833,
+      "step": 17455
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.8244897959183675e-05,
+      "loss": 0.0298,
+      "step": 17456
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.822959183673469e-05,
+      "loss": 0.0352,
+      "step": 17457
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.8214285714285716e-05,
+      "loss": 0.0288,
+      "step": 17458
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.819897959183673e-05,
+      "loss": 0.0993,
+      "step": 17459
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.818367346938776e-05,
+      "loss": 0.0474,
+      "step": 17460
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.8168367346938774e-05,
+      "loss": 0.1078,
+      "step": 17461
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.81530612244898e-05,
+      "loss": 0.0961,
+      "step": 17462
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.8137755102040814e-05,
+      "loss": 0.0354,
+      "step": 17463
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.812244897959184e-05,
+      "loss": 0.1143,
+      "step": 17464
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.8107142857142855e-05,
+      "loss": 0.077,
+      "step": 17465
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.809183673469387e-05,
+      "loss": 0.0253,
+      "step": 17466
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.8076530612244896e-05,
+      "loss": 0.1257,
+      "step": 17467
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.806122448979591e-05,
+      "loss": 0.1276,
+      "step": 17468
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.804591836734694e-05,
+      "loss": 0.1088,
+      "step": 17469
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.8030612244897954e-05,
+      "loss": 0.0358,
+      "step": 17470
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.801530612244898e-05,
+      "loss": 0.0355,
+      "step": 17471
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 0.0284,
+      "step": 17472
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.798469387755102e-05,
+      "loss": 0.1297,
+      "step": 17473
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.7969387755102035e-05,
+      "loss": 0.1277,
+      "step": 17474
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.795408163265306e-05,
+      "loss": 0.1143,
+      "step": 17475
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.7938775510204076e-05,
+      "loss": 0.0278,
+      "step": 17476
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.79234693877551e-05,
+      "loss": 0.0327,
+      "step": 17477
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.790816326530612e-05,
+      "loss": 0.027,
+      "step": 17478
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.789285714285714e-05,
+      "loss": 0.1197,
+      "step": 17479
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.787755102040816e-05,
+      "loss": 0.1631,
+      "step": 17480
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.786224489795918e-05,
+      "loss": 0.0322,
+      "step": 17481
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.78469387755102e-05,
+      "loss": 0.1995,
+      "step": 17482
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.783163265306122e-05,
+      "loss": 0.0754,
+      "step": 17483
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.781632653061224e-05,
+      "loss": 0.0713,
+      "step": 17484
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.780102040816326e-05,
+      "loss": 0.027,
+      "step": 17485
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.778571428571428e-05,
+      "loss": 0.1124,
+      "step": 17486
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.7770408163265304e-05,
+      "loss": 0.1618,
+      "step": 17487
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.775510204081632e-05,
+      "loss": 0.1137,
+      "step": 17488
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.773979591836734e-05,
+      "loss": 0.0287,
+      "step": 17489
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.772448979591836e-05,
+      "loss": 0.0788,
+      "step": 17490
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.770918367346938e-05,
+      "loss": 0.1058,
+      "step": 17491
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.76938775510204e-05,
+      "loss": 0.0925,
+      "step": 17492
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.767857142857142e-05,
+      "loss": 0.1284,
+      "step": 17493
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.766326530612244e-05,
+      "loss": 0.0362,
+      "step": 17494
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.764795918367346e-05,
+      "loss": 0.0398,
+      "step": 17495
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.7632653061224484e-05,
+      "loss": 0.0248,
+      "step": 17496
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.76173469387755e-05,
+      "loss": 0.0444,
+      "step": 17497
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.7602040816326524e-05,
+      "loss": 0.0355,
+      "step": 17498
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.758673469387754e-05,
+      "loss": 0.2396,
+      "step": 17499
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.7571428571428565e-05,
+      "loss": 0.0778,
+      "step": 17500
+    },
+    {
+      "epoch": 16.99,
+      "eval_cer": 0.24472946945255064,
+      "eval_loss": 1.1485048532485962,
+      "eval_runtime": 232.2762,
+      "eval_samples_per_second": 37.946,
+      "eval_steps_per_second": 4.744,
+      "eval_wer": 0.6378516736424655,
+      "step": 17500
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.755612244897959e-05,
+      "loss": 0.0621,
+      "step": 17501
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.754081632653061e-05,
+      "loss": 0.0358,
+      "step": 17502
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.752551020408163e-05,
+      "loss": 0.0678,
+      "step": 17503
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.7510204081632653e-05,
+      "loss": 0.0759,
+      "step": 17504
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.749489795918367e-05,
+      "loss": 0.2373,
+      "step": 17505
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.7479591836734694e-05,
+      "loss": 0.0822,
+      "step": 17506
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.746428571428571e-05,
+      "loss": 0.0333,
+      "step": 17507
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.7448979591836735e-05,
+      "loss": 0.1481,
+      "step": 17508
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.743367346938775e-05,
+      "loss": 0.0498,
+      "step": 17509
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.7418367346938776e-05,
+      "loss": 0.026,
+      "step": 17510
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.740306122448979e-05,
+      "loss": 0.071,
+      "step": 17511
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.7387755102040816e-05,
+      "loss": 0.1045,
+      "step": 17512
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.7372448979591833e-05,
+      "loss": 0.0648,
+      "step": 17513
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.735714285714286e-05,
+      "loss": 0.0498,
+      "step": 17514
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.7341836734693874e-05,
+      "loss": 0.0407,
+      "step": 17515
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.73265306122449e-05,
+      "loss": 0.0669,
+      "step": 17516
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.7311224489795915e-05,
+      "loss": 0.0326,
+      "step": 17517
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.729591836734694e-05,
+      "loss": 0.0323,
+      "step": 17518
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.7280612244897956e-05,
+      "loss": 0.0673,
+      "step": 17519
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.726530612244897e-05,
+      "loss": 0.1073,
+      "step": 17520
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.7249999999999997e-05,
+      "loss": 0.1105,
+      "step": 17521
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.7234693877551014e-05,
+      "loss": 0.0341,
+      "step": 17522
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.721938775510204e-05,
+      "loss": 0.101,
+      "step": 17523
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.7204081632653054e-05,
+      "loss": 0.0715,
+      "step": 17524
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.718877551020408e-05,
+      "loss": 0.0824,
+      "step": 17525
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.7173469387755095e-05,
+      "loss": 0.04,
+      "step": 17526
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.715816326530612e-05,
+      "loss": 0.0639,
+      "step": 17527
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.7142857142857136e-05,
+      "loss": 0.0237,
+      "step": 17528
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.712755102040816e-05,
+      "loss": 0.1246,
+      "step": 17529
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.7112244897959177e-05,
+      "loss": 0.0364,
+      "step": 17530
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.70969387755102e-05,
+      "loss": 0.0244,
+      "step": 17531
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.708163265306122e-05,
+      "loss": 0.0286,
+      "step": 17532
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.706632653061224e-05,
+      "loss": 0.1233,
+      "step": 17533
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.705102040816326e-05,
+      "loss": 0.0454,
+      "step": 17534
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.703571428571428e-05,
+      "loss": 0.0305,
+      "step": 17535
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.70204081632653e-05,
+      "loss": 0.0739,
+      "step": 17536
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.700510204081632e-05,
+      "loss": 0.1098,
+      "step": 17537
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.698979591836734e-05,
+      "loss": 0.0264,
+      "step": 17538
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.6974489795918363e-05,
+      "loss": 0.0267,
+      "step": 17539
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.695918367346938e-05,
+      "loss": 0.0772,
+      "step": 17540
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.69438775510204e-05,
+      "loss": 0.0647,
+      "step": 17541
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.692857142857142e-05,
+      "loss": 0.0657,
+      "step": 17542
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.691326530612244e-05,
+      "loss": 0.0389,
+      "step": 17543
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.689795918367346e-05,
+      "loss": 0.0397,
+      "step": 17544
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.688265306122448e-05,
+      "loss": 0.1089,
+      "step": 17545
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.686734693877551e-05,
+      "loss": 0.0775,
+      "step": 17546
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.685204081632653e-05,
+      "loss": 0.072,
+      "step": 17547
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.683673469387755e-05,
+      "loss": 0.2266,
+      "step": 17548
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.6821428571428574e-05,
+      "loss": 0.076,
+      "step": 17549
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.680612244897959e-05,
+      "loss": 0.031,
+      "step": 17550
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.679081632653061e-05,
+      "loss": 0.1089,
+      "step": 17551
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.677551020408163e-05,
+      "loss": 0.0748,
+      "step": 17552
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.676020408163265e-05,
+      "loss": 0.0639,
+      "step": 17553
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.674489795918367e-05,
+      "loss": 0.1679,
+      "step": 17554
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.672959183673469e-05,
+      "loss": 0.0737,
+      "step": 17555
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.671428571428571e-05,
+      "loss": 0.0279,
+      "step": 17556
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.669897959183673e-05,
+      "loss": 0.1093,
+      "step": 17557
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.6683673469387754e-05,
+      "loss": 0.0893,
+      "step": 17558
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.666836734693877e-05,
+      "loss": 0.103,
+      "step": 17559
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.6653061224489795e-05,
+      "loss": 0.0691,
+      "step": 17560
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.663775510204081e-05,
+      "loss": 0.1137,
+      "step": 17561
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.6622448979591836e-05,
+      "loss": 0.1114,
+      "step": 17562
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.660714285714285e-05,
+      "loss": 0.0294,
+      "step": 17563
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.6591836734693876e-05,
+      "loss": 0.1099,
+      "step": 17564
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.657653061224489e-05,
+      "loss": 0.0348,
+      "step": 17565
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.656122448979592e-05,
+      "loss": 0.0526,
+      "step": 17566
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.6545918367346934e-05,
+      "loss": 0.1236,
+      "step": 17567
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.653061224489796e-05,
+      "loss": 0.0678,
+      "step": 17568
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.6515306122448975e-05,
+      "loss": 0.1057,
+      "step": 17569
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.65e-05,
+      "loss": 0.1001,
+      "step": 17570
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.6484693877551016e-05,
+      "loss": 0.031,
+      "step": 17571
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.646938775510204e-05,
+      "loss": 0.0706,
+      "step": 17572
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.6454081632653056e-05,
+      "loss": 0.0288,
+      "step": 17573
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.643877551020407e-05,
+      "loss": 0.0763,
+      "step": 17574
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.64234693877551e-05,
+      "loss": 0.069,
+      "step": 17575
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.6408163265306114e-05,
+      "loss": 0.0299,
+      "step": 17576
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.639285714285714e-05,
+      "loss": 0.0585,
+      "step": 17577
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.6377551020408155e-05,
+      "loss": 0.0426,
+      "step": 17578
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.636224489795918e-05,
+      "loss": 0.1358,
+      "step": 17579
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.6346938775510196e-05,
+      "loss": 0.1581,
+      "step": 17580
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.633163265306122e-05,
+      "loss": 0.0349,
+      "step": 17581
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.6316326530612236e-05,
+      "loss": 0.0256,
+      "step": 17582
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.630102040816326e-05,
+      "loss": 0.1437,
+      "step": 17583
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.628571428571428e-05,
+      "loss": 0.0348,
+      "step": 17584
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.62704081632653e-05,
+      "loss": 0.1318,
+      "step": 17585
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.625510204081632e-05,
+      "loss": 0.0339,
+      "step": 17586
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.623979591836734e-05,
+      "loss": 0.0362,
+      "step": 17587
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.622448979591836e-05,
+      "loss": 0.1064,
+      "step": 17588
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.620918367346938e-05,
+      "loss": 0.061,
+      "step": 17589
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.61938775510204e-05,
+      "loss": 0.0681,
+      "step": 17590
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.617857142857143e-05,
+      "loss": 0.1024,
+      "step": 17591
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.616326530612245e-05,
+      "loss": 0.2286,
+      "step": 17592
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.614795918367347e-05,
+      "loss": 0.1494,
+      "step": 17593
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.613265306122449e-05,
+      "loss": 0.107,
+      "step": 17594
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.611734693877551e-05,
+      "loss": 0.1455,
+      "step": 17595
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.610204081632653e-05,
+      "loss": 0.0301,
+      "step": 17596
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.608673469387755e-05,
+      "loss": 0.1156,
+      "step": 17597
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.607142857142857e-05,
+      "loss": 0.1057,
+      "step": 17598
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.605612244897959e-05,
+      "loss": 0.1129,
+      "step": 17599
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.604081632653061e-05,
+      "loss": 0.0617,
+      "step": 17600
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.6025510204081634e-05,
+      "loss": 0.0199,
+      "step": 17601
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.601020408163265e-05,
+      "loss": 0.0662,
+      "step": 17602
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.5994897959183675e-05,
+      "loss": 0.0208,
+      "step": 17603
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.597959183673469e-05,
+      "loss": 0.1585,
+      "step": 17604
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.596428571428571e-05,
+      "loss": 0.032,
+      "step": 17605
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.594897959183673e-05,
+      "loss": 0.1033,
+      "step": 17606
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.593367346938775e-05,
+      "loss": 0.0336,
+      "step": 17607
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.591836734693877e-05,
+      "loss": 0.0304,
+      "step": 17608
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.590306122448979e-05,
+      "loss": 0.0414,
+      "step": 17609
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.5887755102040814e-05,
+      "loss": 0.0921,
+      "step": 17610
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.587244897959183e-05,
+      "loss": 0.0772,
+      "step": 17611
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.5857142857142855e-05,
+      "loss": 0.1472,
+      "step": 17612
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.584183673469387e-05,
+      "loss": 0.0352,
+      "step": 17613
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.5826530612244895e-05,
+      "loss": 0.0248,
+      "step": 17614
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.581122448979591e-05,
+      "loss": 0.0213,
+      "step": 17615
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.5795918367346936e-05,
+      "loss": 0.0384,
+      "step": 17616
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.578061224489795e-05,
+      "loss": 0.1569,
+      "step": 17617
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.576530612244898e-05,
+      "loss": 0.032,
+      "step": 17618
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.5749999999999994e-05,
+      "loss": 0.0697,
+      "step": 17619
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.573469387755102e-05,
+      "loss": 0.0956,
+      "step": 17620
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.5719387755102035e-05,
+      "loss": 0.0191,
+      "step": 17621
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.570408163265306e-05,
+      "loss": 0.0932,
+      "step": 17622
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.5688775510204075e-05,
+      "loss": 0.156,
+      "step": 17623
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.56734693877551e-05,
+      "loss": 0.0742,
+      "step": 17624
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.5658163265306116e-05,
+      "loss": 0.0338,
+      "step": 17625
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.564285714285713e-05,
+      "loss": 0.1335,
+      "step": 17626
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.562755102040816e-05,
+      "loss": 0.0278,
+      "step": 17627
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 4.5612244897959174e-05,
+      "loss": 0.074,
+      "step": 17628
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.55969387755102e-05,
+      "loss": 0.0255,
+      "step": 17629
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.5581632653061215e-05,
+      "loss": 0.1051,
+      "step": 17630
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.556632653061224e-05,
+      "loss": 0.0316,
+      "step": 17631
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.5551020408163256e-05,
+      "loss": 0.0275,
+      "step": 17632
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.553571428571428e-05,
+      "loss": 0.1389,
+      "step": 17633
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.5520408163265296e-05,
+      "loss": 0.1086,
+      "step": 17634
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.550510204081632e-05,
+      "loss": 0.0313,
+      "step": 17635
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.548979591836734e-05,
+      "loss": 0.0306,
+      "step": 17636
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.547448979591837e-05,
+      "loss": 0.1081,
+      "step": 17637
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.5459183673469385e-05,
+      "loss": 0.0326,
+      "step": 17638
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.544387755102041e-05,
+      "loss": 0.0674,
+      "step": 17639
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.5428571428571425e-05,
+      "loss": 0.0665,
+      "step": 17640
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.541326530612245e-05,
+      "loss": 0.0682,
+      "step": 17641
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.5397959183673466e-05,
+      "loss": 0.0833,
+      "step": 17642
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.538265306122449e-05,
+      "loss": 0.1146,
+      "step": 17643
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.536734693877551e-05,
+      "loss": 0.0783,
+      "step": 17644
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.535204081632653e-05,
+      "loss": 0.1162,
+      "step": 17645
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.533673469387755e-05,
+      "loss": 0.1065,
+      "step": 17646
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.532142857142857e-05,
+      "loss": 0.0943,
+      "step": 17647
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.530612244897959e-05,
+      "loss": 0.2603,
+      "step": 17648
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.529081632653061e-05,
+      "loss": 0.1114,
+      "step": 17649
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.527551020408163e-05,
+      "loss": 0.0603,
+      "step": 17650
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.526020408163265e-05,
+      "loss": 0.1531,
+      "step": 17651
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.524489795918367e-05,
+      "loss": 0.0857,
+      "step": 17652
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.5229591836734694e-05,
+      "loss": 0.0801,
+      "step": 17653
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.521428571428571e-05,
+      "loss": 0.1434,
+      "step": 17654
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.5198979591836734e-05,
+      "loss": 0.0637,
+      "step": 17655
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.518367346938775e-05,
+      "loss": 0.0662,
+      "step": 17656
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.5168367346938775e-05,
+      "loss": 0.1118,
+      "step": 17657
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.515306122448979e-05,
+      "loss": 0.0274,
+      "step": 17658
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.513775510204081e-05,
+      "loss": 0.062,
+      "step": 17659
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.512244897959183e-05,
+      "loss": 0.0445,
+      "step": 17660
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.510714285714285e-05,
+      "loss": 0.0602,
+      "step": 17661
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.5091836734693874e-05,
+      "loss": 0.1067,
+      "step": 17662
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.507653061224489e-05,
+      "loss": 0.0643,
+      "step": 17663
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.5061224489795915e-05,
+      "loss": 0.0706,
+      "step": 17664
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.504591836734693e-05,
+      "loss": 0.0918,
+      "step": 17665
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.5030612244897955e-05,
+      "loss": 0.0936,
+      "step": 17666
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.501530612244897e-05,
+      "loss": 0.1413,
+      "step": 17667
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 0.0712,
+      "step": 17668
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 4.498469387755101e-05,
+      "loss": 0.0372,
+      "step": 17669
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.496938775510204e-05,
+      "loss": 0.0317,
+      "step": 17670
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.4954081632653054e-05,
+      "loss": 0.0346,
+      "step": 17671
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.493877551020408e-05,
+      "loss": 0.0384,
+      "step": 17672
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.4923469387755095e-05,
+      "loss": 0.2017,
+      "step": 17673
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.490816326530612e-05,
+      "loss": 0.104,
+      "step": 17674
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.4892857142857135e-05,
+      "loss": 0.0265,
+      "step": 17675
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.487755102040816e-05,
+      "loss": 0.0695,
+      "step": 17676
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.4862244897959176e-05,
+      "loss": 0.0645,
+      "step": 17677
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.48469387755102e-05,
+      "loss": 0.0351,
+      "step": 17678
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.483163265306122e-05,
+      "loss": 0.0744,
+      "step": 17679
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.4816326530612234e-05,
+      "loss": 0.0322,
+      "step": 17680
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.480102040816326e-05,
+      "loss": 0.0689,
+      "step": 17681
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.478571428571429e-05,
+      "loss": 0.0309,
+      "step": 17682
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.4770408163265305e-05,
+      "loss": 0.0584,
+      "step": 17683
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.475510204081633e-05,
+      "loss": 0.0325,
+      "step": 17684
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.4739795918367346e-05,
+      "loss": 0.1944,
+      "step": 17685
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.472448979591837e-05,
+      "loss": 0.0295,
+      "step": 17686
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.470918367346939e-05,
+      "loss": 0.0738,
+      "step": 17687
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.469387755102041e-05,
+      "loss": 0.0304,
+      "step": 17688
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.467857142857143e-05,
+      "loss": 0.1702,
+      "step": 17689
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.4663265306122444e-05,
+      "loss": 0.0273,
+      "step": 17690
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.464795918367347e-05,
+      "loss": 0.0361,
+      "step": 17691
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.4632653061224485e-05,
+      "loss": 0.0691,
+      "step": 17692
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.461734693877551e-05,
+      "loss": 0.0279,
+      "step": 17693
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.4602040816326526e-05,
+      "loss": 0.0618,
+      "step": 17694
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.458673469387755e-05,
+      "loss": 0.0315,
+      "step": 17695
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.457142857142857e-05,
+      "loss": 0.0514,
+      "step": 17696
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.455612244897959e-05,
+      "loss": 0.027,
+      "step": 17697
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.454081632653061e-05,
+      "loss": 0.1591,
+      "step": 17698
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.452551020408163e-05,
+      "loss": 0.0718,
+      "step": 17699
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.451020408163265e-05,
+      "loss": 0.0239,
+      "step": 17700
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.449489795918367e-05,
+      "loss": 0.0234,
+      "step": 17701
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.447959183673469e-05,
+      "loss": 0.0248,
+      "step": 17702
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.446428571428571e-05,
+      "loss": 0.0698,
+      "step": 17703
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.444897959183673e-05,
+      "loss": 0.0962,
+      "step": 17704
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.4433673469387754e-05,
+      "loss": 0.0721,
+      "step": 17705
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.441836734693877e-05,
+      "loss": 0.0642,
+      "step": 17706
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.4403061224489794e-05,
+      "loss": 0.0355,
+      "step": 17707
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.438775510204081e-05,
+      "loss": 0.1886,
+      "step": 17708
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.4372448979591835e-05,
+      "loss": 0.035,
+      "step": 17709
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.435714285714285e-05,
+      "loss": 0.0588,
+      "step": 17710
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.434183673469387e-05,
+      "loss": 0.0734,
+      "step": 17711
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.432653061224489e-05,
+      "loss": 0.0711,
+      "step": 17712
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.431122448979591e-05,
+      "loss": 0.07,
+      "step": 17713
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.4295918367346934e-05,
+      "loss": 0.1209,
+      "step": 17714
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.428061224489795e-05,
+      "loss": 0.0608,
+      "step": 17715
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.4265306122448974e-05,
+      "loss": 0.045,
+      "step": 17716
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.424999999999999e-05,
+      "loss": 0.0293,
+      "step": 17717
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.4234693877551015e-05,
+      "loss": 0.0643,
+      "step": 17718
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.421938775510203e-05,
+      "loss": 0.1117,
+      "step": 17719
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.4204081632653056e-05,
+      "loss": 0.062,
+      "step": 17720
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 4.418877551020407e-05,
+      "loss": 0.07,
+      "step": 17721
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.41734693877551e-05,
+      "loss": 0.0249,
+      "step": 17722
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.4158163265306114e-05,
+      "loss": 0.091,
+      "step": 17723
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.414285714285714e-05,
+      "loss": 0.0254,
+      "step": 17724
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.4127551020408154e-05,
+      "loss": 0.0714,
+      "step": 17725
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.411224489795918e-05,
+      "loss": 0.0516,
+      "step": 17726
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.40969387755102e-05,
+      "loss": 0.1212,
+      "step": 17727
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.4081632653061226e-05,
+      "loss": 0.0182,
+      "step": 17728
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.406632653061224e-05,
+      "loss": 0.1096,
+      "step": 17729
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.4051020408163267e-05,
+      "loss": 0.1066,
+      "step": 17730
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.4035714285714284e-05,
+      "loss": 0.0385,
+      "step": 17731
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.402040816326531e-05,
+      "loss": 0.0287,
+      "step": 17732
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.4005102040816324e-05,
+      "loss": 0.0369,
+      "step": 17733
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.398979591836735e-05,
+      "loss": 0.0272,
+      "step": 17734
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.3974489795918365e-05,
+      "loss": 0.0585,
+      "step": 17735
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.395918367346939e-05,
+      "loss": 0.0708,
+      "step": 17736
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.3943877551020406e-05,
+      "loss": 0.1231,
+      "step": 17737
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.392857142857143e-05,
+      "loss": 0.1512,
+      "step": 17738
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.3913265306122447e-05,
+      "loss": 0.1407,
+      "step": 17739
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.389795918367347e-05,
+      "loss": 0.0852,
+      "step": 17740
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.388265306122449e-05,
+      "loss": 0.0378,
+      "step": 17741
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.386734693877551e-05,
+      "loss": 0.1591,
+      "step": 17742
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.385204081632653e-05,
+      "loss": 0.0782,
+      "step": 17743
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.3836734693877545e-05,
+      "loss": 0.0434,
+      "step": 17744
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.382142857142857e-05,
+      "loss": 0.0387,
+      "step": 17745
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.3806122448979586e-05,
+      "loss": 0.0646,
+      "step": 17746
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.379081632653061e-05,
+      "loss": 0.0294,
+      "step": 17747
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.3775510204081627e-05,
+      "loss": 0.1863,
+      "step": 17748
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.376020408163265e-05,
+      "loss": 0.0352,
+      "step": 17749
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.374489795918367e-05,
+      "loss": 0.029,
+      "step": 17750
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.372959183673469e-05,
+      "loss": 0.0415,
+      "step": 17751
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.371428571428571e-05,
+      "loss": 0.0327,
+      "step": 17752
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.369897959183673e-05,
+      "loss": 0.0215,
+      "step": 17753
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.368367346938775e-05,
+      "loss": 0.1693,
+      "step": 17754
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.366836734693877e-05,
+      "loss": 0.0647,
+      "step": 17755
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.365306122448979e-05,
+      "loss": 0.0715,
+      "step": 17756
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.3637755102040813e-05,
+      "loss": 0.0265,
+      "step": 17757
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.362244897959183e-05,
+      "loss": 0.0324,
+      "step": 17758
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.3607142857142854e-05,
+      "loss": 0.0291,
+      "step": 17759
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.359183673469387e-05,
+      "loss": 0.1061,
+      "step": 17760
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.3576530612244895e-05,
+      "loss": 0.0332,
+      "step": 17761
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.356122448979591e-05,
+      "loss": 0.0351,
+      "step": 17762
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.3545918367346936e-05,
+      "loss": 0.1154,
+      "step": 17763
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.353061224489795e-05,
+      "loss": 0.101,
+      "step": 17764
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.351530612244897e-05,
+      "loss": 0.106,
+      "step": 17765
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.3499999999999993e-05,
+      "loss": 0.032,
+      "step": 17766
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.348469387755101e-05,
+      "loss": 0.1022,
+      "step": 17767
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.3469387755102034e-05,
+      "loss": 0.0681,
+      "step": 17768
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.345408163265305e-05,
+      "loss": 0.0729,
+      "step": 17769
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.3438775510204075e-05,
+      "loss": 0.0212,
+      "step": 17770
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.342346938775509e-05,
+      "loss": 0.0267,
+      "step": 17771
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 4.3408163265306116e-05,
+      "loss": 0.0422,
+      "step": 17772
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.3392857142857146e-05,
+      "loss": 0.1604,
+      "step": 17773
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.337755102040816e-05,
+      "loss": 0.1477,
+      "step": 17774
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.336224489795918e-05,
+      "loss": 0.1095,
+      "step": 17775
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.3346938775510204e-05,
+      "loss": 0.0364,
+      "step": 17776
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.333163265306122e-05,
+      "loss": 0.103,
+      "step": 17777
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.3316326530612245e-05,
+      "loss": 0.0235,
+      "step": 17778
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.330102040816326e-05,
+      "loss": 0.1932,
+      "step": 17779
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.3285714285714286e-05,
+      "loss": 0.0651,
+      "step": 17780
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.32704081632653e-05,
+      "loss": 0.0354,
+      "step": 17781
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.3255102040816326e-05,
+      "loss": 0.0319,
+      "step": 17782
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.323979591836734e-05,
+      "loss": 0.0277,
+      "step": 17783
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.322448979591837e-05,
+      "loss": 0.1329,
+      "step": 17784
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.3209183673469384e-05,
+      "loss": 0.1301,
+      "step": 17785
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.319387755102041e-05,
+      "loss": 0.1173,
+      "step": 17786
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.3178571428571425e-05,
+      "loss": 0.0725,
+      "step": 17787
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.316326530612245e-05,
+      "loss": 0.0735,
+      "step": 17788
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.3147959183673466e-05,
+      "loss": 0.0689,
+      "step": 17789
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.313265306122449e-05,
+      "loss": 0.025,
+      "step": 17790
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.3117346938775506e-05,
+      "loss": 0.0765,
+      "step": 17791
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.310204081632653e-05,
+      "loss": 0.0418,
+      "step": 17792
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.308673469387755e-05,
+      "loss": 0.1494,
+      "step": 17793
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.307142857142857e-05,
+      "loss": 0.0232,
+      "step": 17794
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.305612244897959e-05,
+      "loss": 0.0248,
+      "step": 17795
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.3040816326530605e-05,
+      "loss": 0.0629,
+      "step": 17796
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.302551020408163e-05,
+      "loss": 0.0697,
+      "step": 17797
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.3010204081632646e-05,
+      "loss": 0.0932,
+      "step": 17798
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.299489795918367e-05,
+      "loss": 0.112,
+      "step": 17799
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.2979591836734686e-05,
+      "loss": 0.0734,
+      "step": 17800
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.296428571428571e-05,
+      "loss": 0.0199,
+      "step": 17801
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.294897959183673e-05,
+      "loss": 0.0258,
+      "step": 17802
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.293367346938775e-05,
+      "loss": 0.0316,
+      "step": 17803
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.291836734693877e-05,
+      "loss": 0.2232,
+      "step": 17804
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.290306122448979e-05,
+      "loss": 0.1216,
+      "step": 17805
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.288775510204081e-05,
+      "loss": 0.0716,
+      "step": 17806
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.287244897959183e-05,
+      "loss": 0.0727,
+      "step": 17807
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.285714285714285e-05,
+      "loss": 0.0641,
+      "step": 17808
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.284183673469387e-05,
+      "loss": 0.0231,
+      "step": 17809
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.282653061224489e-05,
+      "loss": 0.1137,
+      "step": 17810
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.2811224489795914e-05,
+      "loss": 0.1127,
+      "step": 17811
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.279591836734693e-05,
+      "loss": 0.0779,
+      "step": 17812
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 4.2780612244897955e-05,
+      "loss": 0.0301,
+      "step": 17813
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.276530612244897e-05,
+      "loss": 0.0329,
+      "step": 17814
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.2749999999999996e-05,
+      "loss": 0.0703,
+      "step": 17815
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.273469387755101e-05,
+      "loss": 0.0769,
+      "step": 17816
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.2719387755102036e-05,
+      "loss": 0.0699,
+      "step": 17817
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.270408163265306e-05,
+      "loss": 0.0322,
+      "step": 17818
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.2688775510204084e-05,
+      "loss": 0.0298,
+      "step": 17819
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.26734693877551e-05,
+      "loss": 0.1057,
+      "step": 17820
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.2658163265306125e-05,
+      "loss": 0.0645,
+      "step": 17821
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.264285714285714e-05,
+      "loss": 0.0286,
+      "step": 17822
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.2627551020408165e-05,
+      "loss": 0.0783,
+      "step": 17823
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.261224489795918e-05,
+      "loss": 0.0679,
+      "step": 17824
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.2596938775510206e-05,
+      "loss": 0.0993,
+      "step": 17825
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.258163265306122e-05,
+      "loss": 0.138,
+      "step": 17826
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.256632653061225e-05,
+      "loss": 0.0653,
+      "step": 17827
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.2551020408163264e-05,
+      "loss": 0.1083,
+      "step": 17828
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.253571428571428e-05,
+      "loss": 0.1243,
+      "step": 17829
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.2520408163265305e-05,
+      "loss": 0.0752,
+      "step": 17830
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.250510204081632e-05,
+      "loss": 0.0907,
+      "step": 17831
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.2489795918367345e-05,
+      "loss": 0.0399,
+      "step": 17832
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.247448979591836e-05,
+      "loss": 0.1546,
+      "step": 17833
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.2459183673469386e-05,
+      "loss": 0.0412,
+      "step": 17834
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.24438775510204e-05,
+      "loss": 0.0753,
+      "step": 17835
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.242857142857143e-05,
+      "loss": 0.0738,
+      "step": 17836
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.2413265306122444e-05,
+      "loss": 0.0768,
+      "step": 17837
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.239795918367347e-05,
+      "loss": 0.048,
+      "step": 17838
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.2382653061224485e-05,
+      "loss": 0.0945,
+      "step": 17839
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.236734693877551e-05,
+      "loss": 0.0274,
+      "step": 17840
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.2352040816326526e-05,
+      "loss": 0.0539,
+      "step": 17841
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.233673469387755e-05,
+      "loss": 0.1612,
+      "step": 17842
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.2321428571428566e-05,
+      "loss": 0.0686,
+      "step": 17843
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.230612244897959e-05,
+      "loss": 0.0312,
+      "step": 17844
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.229081632653061e-05,
+      "loss": 0.0715,
+      "step": 17845
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.227551020408163e-05,
+      "loss": 0.0263,
+      "step": 17846
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.226020408163265e-05,
+      "loss": 0.0802,
+      "step": 17847
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.224489795918367e-05,
+      "loss": 0.1627,
+      "step": 17848
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.222959183673469e-05,
+      "loss": 0.0751,
+      "step": 17849
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.2214285714285706e-05,
+      "loss": 0.037,
+      "step": 17850
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.219897959183673e-05,
+      "loss": 0.0692,
+      "step": 17851
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.2183673469387746e-05,
+      "loss": 0.052,
+      "step": 17852
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.216836734693877e-05,
+      "loss": 0.0643,
+      "step": 17853
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.215306122448979e-05,
+      "loss": 0.0779,
+      "step": 17854
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.213775510204081e-05,
+      "loss": 0.1465,
+      "step": 17855
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.212244897959183e-05,
+      "loss": 0.0675,
+      "step": 17856
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.210714285714285e-05,
+      "loss": 0.0813,
+      "step": 17857
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.209183673469387e-05,
+      "loss": 0.028,
+      "step": 17858
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.207653061224489e-05,
+      "loss": 0.0663,
+      "step": 17859
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.206122448979591e-05,
+      "loss": 0.1361,
+      "step": 17860
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.204591836734693e-05,
+      "loss": 0.1271,
+      "step": 17861
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.203061224489795e-05,
+      "loss": 0.0712,
+      "step": 17862
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.2015306122448974e-05,
+      "loss": 0.1085,
+      "step": 17863
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.2e-05,
+      "loss": 0.1044,
+      "step": 17864
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 4.198469387755102e-05,
+      "loss": 0.067,
+      "step": 17865
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.196938775510204e-05,
+      "loss": 0.1204,
+      "step": 17866
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.195408163265306e-05,
+      "loss": 0.1402,
+      "step": 17867
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.193877551020408e-05,
+      "loss": 0.0666,
+      "step": 17868
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.19234693877551e-05,
+      "loss": 0.03,
+      "step": 17869
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.190816326530612e-05,
+      "loss": 0.0729,
+      "step": 17870
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.1892857142857144e-05,
+      "loss": 0.0279,
+      "step": 17871
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.187755102040816e-05,
+      "loss": 0.1158,
+      "step": 17872
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.1862244897959185e-05,
+      "loss": 0.0843,
+      "step": 17873
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.18469387755102e-05,
+      "loss": 0.1048,
+      "step": 17874
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.1831632653061225e-05,
+      "loss": 0.0285,
+      "step": 17875
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.181632653061224e-05,
+      "loss": 0.127,
+      "step": 17876
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.1801020408163266e-05,
+      "loss": 0.0688,
+      "step": 17877
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.178571428571428e-05,
+      "loss": 0.0933,
+      "step": 17878
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.177040816326531e-05,
+      "loss": 0.0468,
+      "step": 17879
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.1755102040816324e-05,
+      "loss": 0.0355,
+      "step": 17880
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.173979591836734e-05,
+      "loss": 0.0699,
+      "step": 17881
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.1724489795918365e-05,
+      "loss": 0.071,
+      "step": 17882
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.170918367346938e-05,
+      "loss": 0.0624,
+      "step": 17883
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.1693877551020405e-05,
+      "loss": 0.0866,
+      "step": 17884
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.167857142857142e-05,
+      "loss": 0.1346,
+      "step": 17885
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.1663265306122446e-05,
+      "loss": 0.0708,
+      "step": 17886
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.164795918367346e-05,
+      "loss": 0.0732,
+      "step": 17887
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.163265306122449e-05,
+      "loss": 0.0635,
+      "step": 17888
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.1617346938775504e-05,
+      "loss": 0.0293,
+      "step": 17889
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.160204081632653e-05,
+      "loss": 0.0884,
+      "step": 17890
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.1586734693877545e-05,
+      "loss": 0.1209,
+      "step": 17891
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.157142857142857e-05,
+      "loss": 0.1318,
+      "step": 17892
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.1556122448979585e-05,
+      "loss": 0.0774,
+      "step": 17893
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.154081632653061e-05,
+      "loss": 0.0337,
+      "step": 17894
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.1525510204081626e-05,
+      "loss": 0.0304,
+      "step": 17895
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.151020408163265e-05,
+      "loss": 0.0697,
+      "step": 17896
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.149489795918367e-05,
+      "loss": 0.0981,
+      "step": 17897
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.147959183673469e-05,
+      "loss": 0.104,
+      "step": 17898
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.146428571428571e-05,
+      "loss": 0.1515,
+      "step": 17899
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.144897959183673e-05,
+      "loss": 0.1528,
+      "step": 17900
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.143367346938775e-05,
+      "loss": 0.0306,
+      "step": 17901
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.141836734693877e-05,
+      "loss": 0.1493,
+      "step": 17902
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.140306122448979e-05,
+      "loss": 0.0401,
+      "step": 17903
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.1387755102040806e-05,
+      "loss": 0.108,
+      "step": 17904
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.137244897959183e-05,
+      "loss": 0.0337,
+      "step": 17905
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.135714285714285e-05,
+      "loss": 0.0233,
+      "step": 17906
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.134183673469387e-05,
+      "loss": 0.0626,
+      "step": 17907
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.132653061224489e-05,
+      "loss": 0.1108,
+      "step": 17908
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.131122448979592e-05,
+      "loss": 0.0966,
+      "step": 17909
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.129591836734694e-05,
+      "loss": 0.1127,
+      "step": 17910
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.128061224489796e-05,
+      "loss": 0.0806,
+      "step": 17911
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.126530612244898e-05,
+      "loss": 0.0749,
+      "step": 17912
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.125e-05,
+      "loss": 0.1058,
+      "step": 17913
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.123469387755102e-05,
+      "loss": 0.0233,
+      "step": 17914
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.121938775510204e-05,
+      "loss": 0.0379,
+      "step": 17915
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.120408163265306e-05,
+      "loss": 0.065,
+      "step": 17916
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.118877551020408e-05,
+      "loss": 0.1788,
+      "step": 17917
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.11734693877551e-05,
+      "loss": 0.1058,
+      "step": 17918
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.115816326530612e-05,
+      "loss": 0.0677,
+      "step": 17919
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.114285714285714e-05,
+      "loss": 0.0238,
+      "step": 17920
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.112755102040816e-05,
+      "loss": 0.0868,
+      "step": 17921
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.111224489795918e-05,
+      "loss": 0.1318,
+      "step": 17922
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.1096938775510204e-05,
+      "loss": 0.1085,
+      "step": 17923
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.108163265306122e-05,
+      "loss": 0.1489,
+      "step": 17924
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.1066326530612244e-05,
+      "loss": 0.0614,
+      "step": 17925
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.105102040816326e-05,
+      "loss": 0.0572,
+      "step": 17926
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.1035714285714285e-05,
+      "loss": 0.0459,
+      "step": 17927
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.10204081632653e-05,
+      "loss": 0.0356,
+      "step": 17928
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.1005102040816326e-05,
+      "loss": 0.0862,
+      "step": 17929
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.098979591836734e-05,
+      "loss": 0.1073,
+      "step": 17930
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.097448979591837e-05,
+      "loss": 0.0361,
+      "step": 17931
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.0959183673469384e-05,
+      "loss": 0.1086,
+      "step": 17932
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.094387755102041e-05,
+      "loss": 0.0537,
+      "step": 17933
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.0928571428571424e-05,
+      "loss": 0.0598,
+      "step": 17934
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.091326530612244e-05,
+      "loss": 0.0364,
+      "step": 17935
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.0897959183673465e-05,
+      "loss": 0.1085,
+      "step": 17936
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.088265306122448e-05,
+      "loss": 0.0717,
+      "step": 17937
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.0867346938775506e-05,
+      "loss": 0.0651,
+      "step": 17938
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.085204081632652e-05,
+      "loss": 0.0633,
+      "step": 17939
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.083673469387755e-05,
+      "loss": 0.1162,
+      "step": 17940
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.0821428571428564e-05,
+      "loss": 0.0344,
+      "step": 17941
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.080612244897959e-05,
+      "loss": 0.08,
+      "step": 17942
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.0790816326530604e-05,
+      "loss": 0.0701,
+      "step": 17943
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.077551020408163e-05,
+      "loss": 0.1658,
+      "step": 17944
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.0760204081632645e-05,
+      "loss": 0.0334,
+      "step": 17945
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.074489795918367e-05,
+      "loss": 0.0749,
+      "step": 17946
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.0729591836734686e-05,
+      "loss": 0.1145,
+      "step": 17947
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.071428571428571e-05,
+      "loss": 0.1296,
+      "step": 17948
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.069897959183673e-05,
+      "loss": 0.043,
+      "step": 17949
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.068367346938775e-05,
+      "loss": 0.0293,
+      "step": 17950
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.066836734693877e-05,
+      "loss": 0.0298,
+      "step": 17951
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.065306122448979e-05,
+      "loss": 0.062,
+      "step": 17952
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.063775510204081e-05,
+      "loss": 0.0976,
+      "step": 17953
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.062244897959184e-05,
+      "loss": 0.063,
+      "step": 17954
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.0607142857142856e-05,
+      "loss": 0.1179,
+      "step": 17955
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.059183673469388e-05,
+      "loss": 0.0642,
+      "step": 17956
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.0576530612244897e-05,
+      "loss": 0.1477,
+      "step": 17957
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.056122448979592e-05,
+      "loss": 0.0736,
+      "step": 17958
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.054591836734694e-05,
+      "loss": 0.0201,
+      "step": 17959
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.053061224489796e-05,
+      "loss": 0.0736,
+      "step": 17960
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.051530612244898e-05,
+      "loss": 0.0299,
+      "step": 17961
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.05e-05,
+      "loss": 0.0792,
+      "step": 17962
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.048469387755102e-05,
+      "loss": 0.1043,
+      "step": 17963
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.046938775510204e-05,
+      "loss": 0.1374,
+      "step": 17964
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.045408163265306e-05,
+      "loss": 0.0393,
+      "step": 17965
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.0438775510204083e-05,
+      "loss": 0.0624,
+      "step": 17966
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.04234693877551e-05,
+      "loss": 0.1156,
+      "step": 17967
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.040816326530612e-05,
+      "loss": 0.0641,
+      "step": 17968
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.039285714285714e-05,
+      "loss": 0.0733,
+      "step": 17969
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.037755102040816e-05,
+      "loss": 0.1004,
+      "step": 17970
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.036224489795918e-05,
+      "loss": 0.0309,
+      "step": 17971
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.03469387755102e-05,
+      "loss": 0.0208,
+      "step": 17972
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.033163265306122e-05,
+      "loss": 0.1741,
+      "step": 17973
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.031632653061224e-05,
+      "loss": 0.0362,
+      "step": 17974
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.0301020408163263e-05,
+      "loss": 0.0258,
+      "step": 17975
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.028571428571428e-05,
+      "loss": 0.0706,
+      "step": 17976
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.0270408163265304e-05,
+      "loss": 0.1372,
+      "step": 17977
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.025510204081632e-05,
+      "loss": 0.0358,
+      "step": 17978
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.0239795918367345e-05,
+      "loss": 0.1647,
+      "step": 17979
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.022448979591836e-05,
+      "loss": 0.0645,
+      "step": 17980
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.0209183673469386e-05,
+      "loss": 0.0812,
+      "step": 17981
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.01938775510204e-05,
+      "loss": 0.0783,
+      "step": 17982
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.0178571428571427e-05,
+      "loss": 0.1,
+      "step": 17983
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.0163265306122443e-05,
+      "loss": 0.0772,
+      "step": 17984
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.014795918367347e-05,
+      "loss": 0.1423,
+      "step": 17985
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.0132653061224484e-05,
+      "loss": 0.0721,
+      "step": 17986
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.011734693877551e-05,
+      "loss": 0.0699,
+      "step": 17987
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.0102040816326525e-05,
+      "loss": 0.025,
+      "step": 17988
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.008673469387754e-05,
+      "loss": 0.0366,
+      "step": 17989
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.0071428571428566e-05,
+      "loss": 0.0284,
+      "step": 17990
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.005612244897958e-05,
+      "loss": 0.0605,
+      "step": 17991
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.0040816326530607e-05,
+      "loss": 0.1216,
+      "step": 17992
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.0025510204081624e-05,
+      "loss": 0.0348,
+      "step": 17993
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.001020408163265e-05,
+      "loss": 0.0208,
+      "step": 17994
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 3.9994897959183664e-05,
+      "loss": 0.0357,
+      "step": 17995
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 3.997959183673469e-05,
+      "loss": 0.0769,
+      "step": 17996
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 3.9964285714285705e-05,
+      "loss": 0.0654,
+      "step": 17997
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 3.994897959183673e-05,
+      "loss": 0.0379,
+      "step": 17998
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 3.9933673469387746e-05,
+      "loss": 0.108,
+      "step": 17999
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.9918367346938776e-05,
+      "loss": 0.0642,
+      "step": 18000
+    },
+    {
+      "epoch": 17.48,
+      "eval_cer": 0.24643532584903344,
+      "eval_loss": 1.2009409666061401,
+      "eval_runtime": 286.3549,
+      "eval_samples_per_second": 30.78,
+      "eval_steps_per_second": 3.848,
+      "eval_wer": 0.6399553988289446,
+      "step": 18000
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.990306122448979e-05,
+      "loss": 0.0164,
+      "step": 18001
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.988775510204082e-05,
+      "loss": 0.0696,
+      "step": 18002
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.9872448979591834e-05,
+      "loss": 0.031,
+      "step": 18003
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.985714285714286e-05,
+      "loss": 0.2305,
+      "step": 18004
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.9841836734693875e-05,
+      "loss": 0.1489,
+      "step": 18005
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.98265306122449e-05,
+      "loss": 0.0765,
+      "step": 18006
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.9811224489795916e-05,
+      "loss": 0.1491,
+      "step": 18007
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.979591836734694e-05,
+      "loss": 0.1404,
+      "step": 18008
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.9780612244897956e-05,
+      "loss": 0.0663,
+      "step": 18009
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.976530612244898e-05,
+      "loss": 0.05,
+      "step": 18010
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.975e-05,
+      "loss": 0.1909,
+      "step": 18011
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.973469387755102e-05,
+      "loss": 0.0713,
+      "step": 18012
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.971938775510204e-05,
+      "loss": 0.1078,
+      "step": 18013
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.970408163265306e-05,
+      "loss": 0.0996,
+      "step": 18014
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.968877551020408e-05,
+      "loss": 0.0383,
+      "step": 18015
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.96734693877551e-05,
+      "loss": 0.0826,
+      "step": 18016
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.965816326530612e-05,
+      "loss": 0.0684,
+      "step": 18017
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.964285714285714e-05,
+      "loss": 0.0785,
+      "step": 18018
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.962755102040816e-05,
+      "loss": 0.0244,
+      "step": 18019
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.961224489795918e-05,
+      "loss": 0.1029,
+      "step": 18020
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.95969387755102e-05,
+      "loss": 0.068,
+      "step": 18021
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.958163265306122e-05,
+      "loss": 0.0385,
+      "step": 18022
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.956632653061224e-05,
+      "loss": 0.1437,
+      "step": 18023
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.955102040816326e-05,
+      "loss": 0.1019,
+      "step": 18024
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.953571428571428e-05,
+      "loss": 0.0262,
+      "step": 18025
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.95204081632653e-05,
+      "loss": 0.0613,
+      "step": 18026
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.950510204081632e-05,
+      "loss": 0.0809,
+      "step": 18027
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.948979591836734e-05,
+      "loss": 0.0274,
+      "step": 18028
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.9474489795918364e-05,
+      "loss": 0.0694,
+      "step": 18029
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.945918367346938e-05,
+      "loss": 0.068,
+      "step": 18030
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.9443877551020405e-05,
+      "loss": 0.0596,
+      "step": 18031
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.942857142857142e-05,
+      "loss": 0.0267,
+      "step": 18032
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.9413265306122446e-05,
+      "loss": 0.1895,
+      "step": 18033
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.939795918367346e-05,
+      "loss": 0.0727,
+      "step": 18034
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.9382653061224486e-05,
+      "loss": 0.1134,
+      "step": 18035
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.93673469387755e-05,
+      "loss": 0.1056,
+      "step": 18036
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.935204081632653e-05,
+      "loss": 0.0284,
+      "step": 18037
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.9336734693877544e-05,
+      "loss": 0.0271,
+      "step": 18038
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.932142857142857e-05,
+      "loss": 0.1039,
+      "step": 18039
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.9306122448979585e-05,
+      "loss": 0.0322,
+      "step": 18040
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.929081632653061e-05,
+      "loss": 0.0599,
+      "step": 18041
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.9275510204081626e-05,
+      "loss": 0.1076,
+      "step": 18042
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.926020408163264e-05,
+      "loss": 0.0305,
+      "step": 18043
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.9244897959183666e-05,
+      "loss": 0.148,
+      "step": 18044
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.92295918367347e-05,
+      "loss": 0.0264,
+      "step": 18045
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.9214285714285714e-05,
+      "loss": 0.0352,
+      "step": 18046
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.919897959183674e-05,
+      "loss": 0.0661,
+      "step": 18047
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.9183673469387755e-05,
+      "loss": 0.0973,
+      "step": 18048
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.916836734693878e-05,
+      "loss": 0.0754,
+      "step": 18049
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.9153061224489795e-05,
+      "loss": 0.181,
+      "step": 18050
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.913775510204082e-05,
+      "loss": 0.0317,
+      "step": 18051
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.9122448979591836e-05,
+      "loss": 0.0277,
+      "step": 18052
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.910714285714285e-05,
+      "loss": 0.0668,
+      "step": 18053
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.909183673469388e-05,
+      "loss": 0.1296,
+      "step": 18054
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.9076530612244894e-05,
+      "loss": 0.1232,
+      "step": 18055
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.906122448979592e-05,
+      "loss": 0.0294,
+      "step": 18056
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.9045918367346935e-05,
+      "loss": 0.032,
+      "step": 18057
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.903061224489796e-05,
+      "loss": 0.1426,
+      "step": 18058
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.9015306122448976e-05,
+      "loss": 0.074,
+      "step": 18059
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.9e-05,
+      "loss": 0.113,
+      "step": 18060
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.8984693877551016e-05,
+      "loss": 0.1571,
+      "step": 18061
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.896938775510204e-05,
+      "loss": 0.0669,
+      "step": 18062
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.895408163265306e-05,
+      "loss": 0.0283,
+      "step": 18063
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.893877551020408e-05,
+      "loss": 0.068,
+      "step": 18064
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.89234693877551e-05,
+      "loss": 0.0682,
+      "step": 18065
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.890816326530612e-05,
+      "loss": 0.0278,
+      "step": 18066
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.889285714285714e-05,
+      "loss": 0.1273,
+      "step": 18067
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.887755102040816e-05,
+      "loss": 0.0401,
+      "step": 18068
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.886224489795918e-05,
+      "loss": 0.0601,
+      "step": 18069
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.88469387755102e-05,
+      "loss": 0.1091,
+      "step": 18070
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.883163265306122e-05,
+      "loss": 0.0368,
+      "step": 18071
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.8816326530612244e-05,
+      "loss": 0.128,
+      "step": 18072
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.880102040816326e-05,
+      "loss": 0.1057,
+      "step": 18073
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.878571428571428e-05,
+      "loss": 0.1085,
+      "step": 18074
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.87704081632653e-05,
+      "loss": 0.0414,
+      "step": 18075
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.875510204081632e-05,
+      "loss": 0.0759,
+      "step": 18076
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.873979591836734e-05,
+      "loss": 0.0821,
+      "step": 18077
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.872448979591836e-05,
+      "loss": 0.0403,
+      "step": 18078
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.870918367346938e-05,
+      "loss": 0.0468,
+      "step": 18079
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.86938775510204e-05,
+      "loss": 0.0763,
+      "step": 18080
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.8678571428571424e-05,
+      "loss": 0.0783,
+      "step": 18081
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.866326530612244e-05,
+      "loss": 0.0574,
+      "step": 18082
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.8647959183673465e-05,
+      "loss": 0.0351,
+      "step": 18083
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.863265306122448e-05,
+      "loss": 0.0256,
+      "step": 18084
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.8617346938775505e-05,
+      "loss": 0.0841,
+      "step": 18085
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.860204081632652e-05,
+      "loss": 0.0827,
+      "step": 18086
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.8586734693877546e-05,
+      "loss": 0.0235,
+      "step": 18087
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.857142857142856e-05,
+      "loss": 0.1022,
+      "step": 18088
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.855612244897959e-05,
+      "loss": 0.1689,
+      "step": 18089
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.854081632653061e-05,
+      "loss": 0.1003,
+      "step": 18090
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.8525510204081635e-05,
+      "loss": 0.0951,
+      "step": 18091
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.851020408163265e-05,
+      "loss": 0.1908,
+      "step": 18092
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.8494897959183675e-05,
+      "loss": 0.0635,
+      "step": 18093
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.847959183673469e-05,
+      "loss": 0.0294,
+      "step": 18094
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.8464285714285716e-05,
+      "loss": 0.0696,
+      "step": 18095
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.844897959183673e-05,
+      "loss": 0.0719,
+      "step": 18096
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.843367346938776e-05,
+      "loss": 0.0251,
+      "step": 18097
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.8418367346938774e-05,
+      "loss": 0.2819,
+      "step": 18098
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.84030612244898e-05,
+      "loss": 0.0633,
+      "step": 18099
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.8387755102040815e-05,
+      "loss": 0.1006,
+      "step": 18100
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.837244897959184e-05,
+      "loss": 0.1047,
+      "step": 18101
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.8357142857142855e-05,
+      "loss": 0.0277,
+      "step": 18102
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.834183673469388e-05,
+      "loss": 0.1412,
+      "step": 18103
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.8326530612244896e-05,
+      "loss": 0.1318,
+      "step": 18104
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.831122448979591e-05,
+      "loss": 0.1584,
+      "step": 18105
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.829591836734694e-05,
+      "loss": 0.0223,
+      "step": 18106
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.8280612244897954e-05,
+      "loss": 0.0292,
+      "step": 18107
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.826530612244898e-05,
+      "loss": 0.061,
+      "step": 18108
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.8249999999999995e-05,
+      "loss": 0.0801,
+      "step": 18109
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.823469387755102e-05,
+      "loss": 0.0471,
+      "step": 18110
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.8219387755102035e-05,
+      "loss": 0.0807,
+      "step": 18111
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.820408163265306e-05,
+      "loss": 0.0452,
+      "step": 18112
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.8188775510204076e-05,
+      "loss": 0.0295,
+      "step": 18113
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.81734693877551e-05,
+      "loss": 0.035,
+      "step": 18114
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.815816326530612e-05,
+      "loss": 0.1068,
+      "step": 18115
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.814285714285714e-05,
+      "loss": 0.0447,
+      "step": 18116
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.812755102040816e-05,
+      "loss": 0.0291,
+      "step": 18117
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.811224489795918e-05,
+      "loss": 0.0748,
+      "step": 18118
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.80969387755102e-05,
+      "loss": 0.1811,
+      "step": 18119
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.808163265306122e-05,
+      "loss": 0.0651,
+      "step": 18120
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.806632653061224e-05,
+      "loss": 0.0302,
+      "step": 18121
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.805102040816326e-05,
+      "loss": 0.0959,
+      "step": 18122
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.803571428571428e-05,
+      "loss": 0.211,
+      "step": 18123
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.8020408163265304e-05,
+      "loss": 0.0351,
+      "step": 18124
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.800510204081632e-05,
+      "loss": 0.0424,
+      "step": 18125
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.7989795918367344e-05,
+      "loss": 0.11,
+      "step": 18126
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.797448979591836e-05,
+      "loss": 0.0228,
+      "step": 18127
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.795918367346938e-05,
+      "loss": 0.0311,
+      "step": 18128
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.79438775510204e-05,
+      "loss": 0.0955,
+      "step": 18129
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.792857142857142e-05,
+      "loss": 0.1152,
+      "step": 18130
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.791326530612244e-05,
+      "loss": 0.0869,
+      "step": 18131
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.789795918367346e-05,
+      "loss": 0.0334,
+      "step": 18132
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.7882653061224484e-05,
+      "loss": 0.0756,
+      "step": 18133
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.78673469387755e-05,
+      "loss": 0.0281,
+      "step": 18134
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.7852040816326525e-05,
+      "loss": 0.1241,
+      "step": 18135
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.7836734693877555e-05,
+      "loss": 0.1163,
+      "step": 18136
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.782142857142857e-05,
+      "loss": 0.0392,
+      "step": 18137
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.780612244897959e-05,
+      "loss": 0.0679,
+      "step": 18138
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.779081632653061e-05,
+      "loss": 0.0193,
+      "step": 18139
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.777551020408163e-05,
+      "loss": 0.0583,
+      "step": 18140
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.7760204081632654e-05,
+      "loss": 0.0595,
+      "step": 18141
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.774489795918367e-05,
+      "loss": 0.1988,
+      "step": 18142
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.7729591836734694e-05,
+      "loss": 0.2645,
+      "step": 18143
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.771428571428571e-05,
+      "loss": 0.1049,
+      "step": 18144
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.7698979591836735e-05,
+      "loss": 0.0747,
+      "step": 18145
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.768367346938775e-05,
+      "loss": 0.0721,
+      "step": 18146
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.7668367346938776e-05,
+      "loss": 0.1086,
+      "step": 18147
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.765306122448979e-05,
+      "loss": 0.2378,
+      "step": 18148
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.763775510204082e-05,
+      "loss": 0.0689,
+      "step": 18149
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.7622448979591834e-05,
+      "loss": 0.0174,
+      "step": 18150
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.760714285714286e-05,
+      "loss": 0.0249,
+      "step": 18151
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.7591836734693874e-05,
+      "loss": 0.1056,
+      "step": 18152
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 3.75765306122449e-05,
+      "loss": 0.0359,
+      "step": 18153
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.7561224489795915e-05,
+      "loss": 0.158,
+      "step": 18154
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.754591836734694e-05,
+      "loss": 0.0708,
+      "step": 18155
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.7530612244897956e-05,
+      "loss": 0.0689,
+      "step": 18156
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.751530612244898e-05,
+      "loss": 0.0677,
+      "step": 18157
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.75e-05,
+      "loss": 0.0266,
+      "step": 18158
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.7484693877551014e-05,
+      "loss": 0.1809,
+      "step": 18159
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.746938775510204e-05,
+      "loss": 0.0642,
+      "step": 18160
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.7454081632653054e-05,
+      "loss": 0.115,
+      "step": 18161
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.743877551020408e-05,
+      "loss": 0.0653,
+      "step": 18162
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.7423469387755095e-05,
+      "loss": 0.0684,
+      "step": 18163
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.740816326530612e-05,
+      "loss": 0.0659,
+      "step": 18164
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.7392857142857136e-05,
+      "loss": 0.0218,
+      "step": 18165
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.737755102040816e-05,
+      "loss": 0.0903,
+      "step": 18166
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.736224489795918e-05,
+      "loss": 0.0902,
+      "step": 18167
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.73469387755102e-05,
+      "loss": 0.0687,
+      "step": 18168
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.733163265306122e-05,
+      "loss": 0.0785,
+      "step": 18169
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.731632653061224e-05,
+      "loss": 0.0695,
+      "step": 18170
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.7301020408163265e-05,
+      "loss": 0.0676,
+      "step": 18171
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.728571428571428e-05,
+      "loss": 0.0452,
+      "step": 18172
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.7270408163265306e-05,
+      "loss": 0.1063,
+      "step": 18173
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.725510204081632e-05,
+      "loss": 0.0244,
+      "step": 18174
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.7239795918367347e-05,
+      "loss": 0.0662,
+      "step": 18175
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.7224489795918364e-05,
+      "loss": 0.0263,
+      "step": 18176
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.720918367346939e-05,
+      "loss": 0.0334,
+      "step": 18177
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.7193877551020404e-05,
+      "loss": 0.0678,
+      "step": 18178
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.717857142857143e-05,
+      "loss": 0.1639,
+      "step": 18179
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.7163265306122445e-05,
+      "loss": 0.1169,
+      "step": 18180
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.714795918367347e-05,
+      "loss": 0.0663,
+      "step": 18181
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.7132653061224486e-05,
+      "loss": 0.0721,
+      "step": 18182
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.711734693877551e-05,
+      "loss": 0.1155,
+      "step": 18183
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.710204081632653e-05,
+      "loss": 0.0452,
+      "step": 18184
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.7086734693877544e-05,
+      "loss": 0.073,
+      "step": 18185
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.707142857142857e-05,
+      "loss": 0.0372,
+      "step": 18186
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.7056122448979584e-05,
+      "loss": 0.0262,
+      "step": 18187
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.704081632653061e-05,
+      "loss": 0.1147,
+      "step": 18188
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.7025510204081625e-05,
+      "loss": 0.0788,
+      "step": 18189
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.701020408163265e-05,
+      "loss": 0.025,
+      "step": 18190
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.6994897959183666e-05,
+      "loss": 0.1146,
+      "step": 18191
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.697959183673469e-05,
+      "loss": 0.088,
+      "step": 18192
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.6964285714285713e-05,
+      "loss": 0.0688,
+      "step": 18193
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 3.694897959183673e-05,
+      "loss": 0.0609,
+      "step": 18194
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.6933673469387754e-05,
+      "loss": 0.1053,
+      "step": 18195
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.691836734693877e-05,
+      "loss": 0.076,
+      "step": 18196
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.6903061224489795e-05,
+      "loss": 0.0711,
+      "step": 18197
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.688775510204081e-05,
+      "loss": 0.0351,
+      "step": 18198
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.6872448979591836e-05,
+      "loss": 0.2724,
+      "step": 18199
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.685714285714285e-05,
+      "loss": 0.049,
+      "step": 18200
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.6841836734693877e-05,
+      "loss": 0.0759,
+      "step": 18201
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.6826530612244894e-05,
+      "loss": 0.037,
+      "step": 18202
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.681122448979592e-05,
+      "loss": 0.0257,
+      "step": 18203
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.6795918367346934e-05,
+      "loss": 0.1131,
+      "step": 18204
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.678061224489796e-05,
+      "loss": 0.0628,
+      "step": 18205
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.6765306122448975e-05,
+      "loss": 0.0727,
+      "step": 18206
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.675e-05,
+      "loss": 0.0745,
+      "step": 18207
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.6734693877551016e-05,
+      "loss": 0.0311,
+      "step": 18208
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.671938775510204e-05,
+      "loss": 0.046,
+      "step": 18209
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.6704081632653057e-05,
+      "loss": 0.1035,
+      "step": 18210
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.668877551020408e-05,
+      "loss": 0.0696,
+      "step": 18211
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.66734693877551e-05,
+      "loss": 0.0665,
+      "step": 18212
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.6658163265306114e-05,
+      "loss": 0.0625,
+      "step": 18213
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.664285714285714e-05,
+      "loss": 0.0591,
+      "step": 18214
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.662755102040816e-05,
+      "loss": 0.029,
+      "step": 18215
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.6612244897959186e-05,
+      "loss": 0.0757,
+      "step": 18216
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.65969387755102e-05,
+      "loss": 0.0388,
+      "step": 18217
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.658163265306122e-05,
+      "loss": 0.0823,
+      "step": 18218
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.6566326530612243e-05,
+      "loss": 0.0322,
+      "step": 18219
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.655102040816326e-05,
+      "loss": 0.0795,
+      "step": 18220
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.6535714285714284e-05,
+      "loss": 0.0636,
+      "step": 18221
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.65204081632653e-05,
+      "loss": 0.1392,
+      "step": 18222
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.6505102040816325e-05,
+      "loss": 0.0391,
+      "step": 18223
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.648979591836734e-05,
+      "loss": 0.0323,
+      "step": 18224
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.6474489795918366e-05,
+      "loss": 0.0591,
+      "step": 18225
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.645918367346938e-05,
+      "loss": 0.0283,
+      "step": 18226
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.6443877551020406e-05,
+      "loss": 0.0295,
+      "step": 18227
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.6428571428571423e-05,
+      "loss": 0.0506,
+      "step": 18228
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.641326530612245e-05,
+      "loss": 0.1417,
+      "step": 18229
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.6397959183673464e-05,
+      "loss": 0.1121,
+      "step": 18230
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.638265306122449e-05,
+      "loss": 0.0671,
+      "step": 18231
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.6367346938775505e-05,
+      "loss": 0.0673,
+      "step": 18232
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.635204081632653e-05,
+      "loss": 0.0319,
+      "step": 18233
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.6336734693877546e-05,
+      "loss": 0.0814,
+      "step": 18234
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.632142857142857e-05,
+      "loss": 0.0662,
+      "step": 18235
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.6306122448979586e-05,
+      "loss": 0.0332,
+      "step": 18236
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.629081632653061e-05,
+      "loss": 0.0262,
+      "step": 18237
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.6275510204081634e-05,
+      "loss": 0.109,
+      "step": 18238
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.626020408163265e-05,
+      "loss": 0.0763,
+      "step": 18239
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.6244897959183675e-05,
+      "loss": 0.0899,
+      "step": 18240
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.622959183673469e-05,
+      "loss": 0.0481,
+      "step": 18241
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.6214285714285716e-05,
+      "loss": 0.1429,
+      "step": 18242
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.619897959183673e-05,
+      "loss": 0.0674,
+      "step": 18243
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.618367346938775e-05,
+      "loss": 0.0351,
+      "step": 18244
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.616836734693877e-05,
+      "loss": 0.0611,
+      "step": 18245
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 3.615306122448979e-05,
+      "loss": 0.1101,
+      "step": 18246
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.6137755102040814e-05,
+      "loss": 0.1276,
+      "step": 18247
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.612244897959183e-05,
+      "loss": 0.0538,
+      "step": 18248
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.6107142857142855e-05,
+      "loss": 0.1178,
+      "step": 18249
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.609183673469387e-05,
+      "loss": 0.1144,
+      "step": 18250
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.6076530612244896e-05,
+      "loss": 0.1064,
+      "step": 18251
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.606122448979591e-05,
+      "loss": 0.0427,
+      "step": 18252
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.6045918367346936e-05,
+      "loss": 0.0716,
+      "step": 18253
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.603061224489795e-05,
+      "loss": 0.0814,
+      "step": 18254
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.601530612244898e-05,
+      "loss": 0.16,
+      "step": 18255
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.5999999999999994e-05,
+      "loss": 0.0663,
+      "step": 18256
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.598469387755102e-05,
+      "loss": 0.0277,
+      "step": 18257
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.5969387755102035e-05,
+      "loss": 0.157,
+      "step": 18258
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.595408163265306e-05,
+      "loss": 0.0881,
+      "step": 18259
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.5938775510204076e-05,
+      "loss": 0.0366,
+      "step": 18260
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.59234693877551e-05,
+      "loss": 0.0239,
+      "step": 18261
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.590816326530612e-05,
+      "loss": 0.1149,
+      "step": 18262
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.589285714285714e-05,
+      "loss": 0.0377,
+      "step": 18263
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.5877551020408164e-05,
+      "loss": 0.0281,
+      "step": 18264
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.586224489795918e-05,
+      "loss": 0.0637,
+      "step": 18265
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.5846938775510205e-05,
+      "loss": 0.0619,
+      "step": 18266
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.583163265306122e-05,
+      "loss": 0.0711,
+      "step": 18267
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.5816326530612245e-05,
+      "loss": 0.066,
+      "step": 18268
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.580102040816326e-05,
+      "loss": 0.0244,
+      "step": 18269
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.578571428571428e-05,
+      "loss": 0.0266,
+      "step": 18270
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.57704081632653e-05,
+      "loss": 0.0232,
+      "step": 18271
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.575510204081632e-05,
+      "loss": 0.0271,
+      "step": 18272
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.5739795918367344e-05,
+      "loss": 0.0786,
+      "step": 18273
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.572448979591836e-05,
+      "loss": 0.0208,
+      "step": 18274
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.5709183673469385e-05,
+      "loss": 0.0741,
+      "step": 18275
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.56938775510204e-05,
+      "loss": 0.076,
+      "step": 18276
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.5678571428571426e-05,
+      "loss": 0.136,
+      "step": 18277
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.566326530612244e-05,
+      "loss": 0.0522,
+      "step": 18278
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.5647959183673466e-05,
+      "loss": 0.0606,
+      "step": 18279
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.563265306122448e-05,
+      "loss": 0.1164,
+      "step": 18280
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.561734693877551e-05,
+      "loss": 0.0322,
+      "step": 18281
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.5602040816326524e-05,
+      "loss": 0.1021,
+      "step": 18282
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.558673469387755e-05,
+      "loss": 0.0713,
+      "step": 18283
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.557142857142857e-05,
+      "loss": 0.0584,
+      "step": 18284
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.555612244897959e-05,
+      "loss": 0.181,
+      "step": 18285
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.554081632653061e-05,
+      "loss": 0.0654,
+      "step": 18286
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.552551020408163e-05,
+      "loss": 0.0726,
+      "step": 18287
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.551020408163265e-05,
+      "loss": 0.0373,
+      "step": 18288
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.549489795918367e-05,
+      "loss": 0.0352,
+      "step": 18289
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.5479591836734694e-05,
+      "loss": 0.0268,
+      "step": 18290
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.546428571428571e-05,
+      "loss": 0.0588,
+      "step": 18291
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.5448979591836735e-05,
+      "loss": 0.0941,
+      "step": 18292
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.543367346938775e-05,
+      "loss": 0.0624,
+      "step": 18293
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.5418367346938775e-05,
+      "loss": 0.0254,
+      "step": 18294
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.540306122448979e-05,
+      "loss": 0.0227,
+      "step": 18295
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.5387755102040816e-05,
+      "loss": 0.1087,
+      "step": 18296
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 3.537244897959183e-05,
+      "loss": 0.0515,
+      "step": 18297
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.535714285714285e-05,
+      "loss": 0.0587,
+      "step": 18298
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.5341836734693874e-05,
+      "loss": 0.0283,
+      "step": 18299
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.532653061224489e-05,
+      "loss": 0.0241,
+      "step": 18300
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.5311224489795915e-05,
+      "loss": 0.0427,
+      "step": 18301
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.529591836734693e-05,
+      "loss": 0.0377,
+      "step": 18302
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.5280612244897955e-05,
+      "loss": 0.031,
+      "step": 18303
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.526530612244897e-05,
+      "loss": 0.0544,
+      "step": 18304
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.5249999999999996e-05,
+      "loss": 0.0235,
+      "step": 18305
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.523469387755102e-05,
+      "loss": 0.1034,
+      "step": 18306
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.521938775510204e-05,
+      "loss": 0.0707,
+      "step": 18307
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.520408163265306e-05,
+      "loss": 0.0642,
+      "step": 18308
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.518877551020408e-05,
+      "loss": 0.0312,
+      "step": 18309
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.51734693877551e-05,
+      "loss": 0.044,
+      "step": 18310
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.515816326530612e-05,
+      "loss": 0.1172,
+      "step": 18311
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.514285714285714e-05,
+      "loss": 0.0639,
+      "step": 18312
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.512755102040816e-05,
+      "loss": 0.1089,
+      "step": 18313
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.511224489795918e-05,
+      "loss": 0.055,
+      "step": 18314
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.50969387755102e-05,
+      "loss": 0.0581,
+      "step": 18315
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.5081632653061224e-05,
+      "loss": 0.022,
+      "step": 18316
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.506632653061224e-05,
+      "loss": 0.0975,
+      "step": 18317
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.5051020408163265e-05,
+      "loss": 0.1066,
+      "step": 18318
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.503571428571428e-05,
+      "loss": 0.0259,
+      "step": 18319
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.5020408163265305e-05,
+      "loss": 0.0313,
+      "step": 18320
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.500510204081632e-05,
+      "loss": 0.0264,
+      "step": 18321
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.4989795918367346e-05,
+      "loss": 0.1238,
+      "step": 18322
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.497448979591836e-05,
+      "loss": 0.1272,
+      "step": 18323
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.495918367346938e-05,
+      "loss": 0.1031,
+      "step": 18324
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.4943877551020404e-05,
+      "loss": 0.029,
+      "step": 18325
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.492857142857142e-05,
+      "loss": 0.1141,
+      "step": 18326
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.4913265306122445e-05,
+      "loss": 0.0284,
+      "step": 18327
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.489795918367346e-05,
+      "loss": 0.0427,
+      "step": 18328
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.4882653061224485e-05,
+      "loss": 0.0879,
+      "step": 18329
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.486734693877551e-05,
+      "loss": 0.1412,
+      "step": 18330
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.4852040816326526e-05,
+      "loss": 0.0814,
+      "step": 18331
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.483673469387755e-05,
+      "loss": 0.0308,
+      "step": 18332
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.482142857142857e-05,
+      "loss": 0.0686,
+      "step": 18333
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.480612244897959e-05,
+      "loss": 0.0265,
+      "step": 18334
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.479081632653061e-05,
+      "loss": 0.1075,
+      "step": 18335
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.477551020408163e-05,
+      "loss": 0.0351,
+      "step": 18336
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.476020408163265e-05,
+      "loss": 0.0631,
+      "step": 18337
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.474489795918367e-05,
+      "loss": 0.0954,
+      "step": 18338
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 3.472959183673469e-05,
+      "loss": 0.0303,
+      "step": 18339
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.471428571428571e-05,
+      "loss": 0.05,
+      "step": 18340
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.469897959183673e-05,
+      "loss": 0.0319,
+      "step": 18341
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.4683673469387754e-05,
+      "loss": 0.0752,
+      "step": 18342
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.466836734693877e-05,
+      "loss": 0.0467,
+      "step": 18343
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.4653061224489795e-05,
+      "loss": 0.0795,
+      "step": 18344
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.463775510204081e-05,
+      "loss": 0.0263,
+      "step": 18345
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.4622448979591835e-05,
+      "loss": 0.0689,
+      "step": 18346
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.460714285714285e-05,
+      "loss": 0.0459,
+      "step": 18347
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.4591836734693876e-05,
+      "loss": 0.1934,
+      "step": 18348
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.457653061224489e-05,
+      "loss": 0.069,
+      "step": 18349
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.456122448979591e-05,
+      "loss": 0.0358,
+      "step": 18350
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.454591836734694e-05,
+      "loss": 0.0672,
+      "step": 18351
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.453061224489796e-05,
+      "loss": 0.0512,
+      "step": 18352
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.451530612244898e-05,
+      "loss": 0.069,
+      "step": 18353
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.45e-05,
+      "loss": 0.1412,
+      "step": 18354
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.4484693877551015e-05,
+      "loss": 0.1056,
+      "step": 18355
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.446938775510204e-05,
+      "loss": 0.0677,
+      "step": 18356
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.4454081632653056e-05,
+      "loss": 0.065,
+      "step": 18357
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.443877551020408e-05,
+      "loss": 0.0799,
+      "step": 18358
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.44234693877551e-05,
+      "loss": 0.054,
+      "step": 18359
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.440816326530612e-05,
+      "loss": 0.0799,
+      "step": 18360
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.439285714285714e-05,
+      "loss": 0.0812,
+      "step": 18361
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.437755102040816e-05,
+      "loss": 0.0727,
+      "step": 18362
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.436224489795918e-05,
+      "loss": 0.0829,
+      "step": 18363
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.43469387755102e-05,
+      "loss": 0.0296,
+      "step": 18364
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.433163265306122e-05,
+      "loss": 0.0351,
+      "step": 18365
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.431632653061224e-05,
+      "loss": 0.1077,
+      "step": 18366
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.430102040816326e-05,
+      "loss": 0.0655,
+      "step": 18367
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.4285714285714284e-05,
+      "loss": 0.1039,
+      "step": 18368
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.42704081632653e-05,
+      "loss": 0.0679,
+      "step": 18369
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.4255102040816324e-05,
+      "loss": 0.0247,
+      "step": 18370
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.423979591836734e-05,
+      "loss": 0.0317,
+      "step": 18371
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.4224489795918365e-05,
+      "loss": 0.0538,
+      "step": 18372
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.420918367346938e-05,
+      "loss": 0.0964,
+      "step": 18373
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.4193877551020406e-05,
+      "loss": 0.0267,
+      "step": 18374
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.417857142857143e-05,
+      "loss": 0.0643,
+      "step": 18375
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.416326530612245e-05,
+      "loss": 0.0647,
+      "step": 18376
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.414795918367347e-05,
+      "loss": 0.0533,
+      "step": 18377
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.413265306122449e-05,
+      "loss": 0.0347,
+      "step": 18378
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.411734693877551e-05,
+      "loss": 0.0371,
+      "step": 18379
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.410204081632653e-05,
+      "loss": 0.072,
+      "step": 18380
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.408673469387755e-05,
+      "loss": 0.0607,
+      "step": 18381
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.407142857142857e-05,
+      "loss": 0.1013,
+      "step": 18382
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.4056122448979586e-05,
+      "loss": 0.0369,
+      "step": 18383
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.404081632653061e-05,
+      "loss": 0.0718,
+      "step": 18384
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.402551020408163e-05,
+      "loss": 0.0705,
+      "step": 18385
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.401020408163265e-05,
+      "loss": 0.1115,
+      "step": 18386
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.399489795918367e-05,
+      "loss": 0.0246,
+      "step": 18387
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.397959183673469e-05,
+      "loss": 0.0615,
+      "step": 18388
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.396428571428571e-05,
+      "loss": 0.0815,
+      "step": 18389
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 3.394897959183673e-05,
+      "loss": 0.0654,
+      "step": 18390
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.393367346938775e-05,
+      "loss": 0.2118,
+      "step": 18391
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.391836734693877e-05,
+      "loss": 0.1121,
+      "step": 18392
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.390306122448979e-05,
+      "loss": 0.0631,
+      "step": 18393
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.3887755102040814e-05,
+      "loss": 0.0306,
+      "step": 18394
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.387244897959183e-05,
+      "loss": 0.0674,
+      "step": 18395
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.3857142857142854e-05,
+      "loss": 0.1075,
+      "step": 18396
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.384183673469388e-05,
+      "loss": 0.0752,
+      "step": 18397
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.3826530612244895e-05,
+      "loss": 0.2598,
+      "step": 18398
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.381122448979592e-05,
+      "loss": 0.0736,
+      "step": 18399
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.3795918367346936e-05,
+      "loss": 0.0659,
+      "step": 18400
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.378061224489796e-05,
+      "loss": 0.0253,
+      "step": 18401
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.376530612244898e-05,
+      "loss": 0.0383,
+      "step": 18402
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.375e-05,
+      "loss": 0.0865,
+      "step": 18403
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.373469387755102e-05,
+      "loss": 0.1876,
+      "step": 18404
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.371938775510204e-05,
+      "loss": 0.1147,
+      "step": 18405
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.370408163265306e-05,
+      "loss": 0.1538,
+      "step": 18406
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.368877551020408e-05,
+      "loss": 0.0697,
+      "step": 18407
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.36734693877551e-05,
+      "loss": 0.0395,
+      "step": 18408
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.3658163265306116e-05,
+      "loss": 0.0671,
+      "step": 18409
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.364285714285714e-05,
+      "loss": 0.0675,
+      "step": 18410
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.362755102040816e-05,
+      "loss": 0.0812,
+      "step": 18411
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.361224489795918e-05,
+      "loss": 0.0355,
+      "step": 18412
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.35969387755102e-05,
+      "loss": 0.066,
+      "step": 18413
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.358163265306122e-05,
+      "loss": 0.0659,
+      "step": 18414
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.356632653061224e-05,
+      "loss": 0.0691,
+      "step": 18415
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.355102040816326e-05,
+      "loss": 0.0373,
+      "step": 18416
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.353571428571428e-05,
+      "loss": 0.167,
+      "step": 18417
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.35204081632653e-05,
+      "loss": 0.145,
+      "step": 18418
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.3505102040816327e-05,
+      "loss": 0.1058,
+      "step": 18419
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.3489795918367344e-05,
+      "loss": 0.0529,
+      "step": 18420
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.347448979591837e-05,
+      "loss": 0.0271,
+      "step": 18421
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.3459183673469384e-05,
+      "loss": 0.0648,
+      "step": 18422
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.344387755102041e-05,
+      "loss": 0.0861,
+      "step": 18423
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.3428571428571425e-05,
+      "loss": 0.1443,
+      "step": 18424
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.341326530612245e-05,
+      "loss": 0.0688,
+      "step": 18425
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.3397959183673466e-05,
+      "loss": 0.0247,
+      "step": 18426
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.338265306122449e-05,
+      "loss": 0.075,
+      "step": 18427
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.3367346938775507e-05,
+      "loss": 0.0559,
+      "step": 18428
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.335204081632653e-05,
+      "loss": 0.2333,
+      "step": 18429
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.333673469387755e-05,
+      "loss": 0.0338,
+      "step": 18430
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.332142857142857e-05,
+      "loss": 0.0278,
+      "step": 18431
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.330612244897959e-05,
+      "loss": 0.0671,
+      "step": 18432
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.329081632653061e-05,
+      "loss": 0.1084,
+      "step": 18433
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.327551020408163e-05,
+      "loss": 0.0632,
+      "step": 18434
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.326020408163265e-05,
+      "loss": 0.1053,
+      "step": 18435
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.324489795918367e-05,
+      "loss": 0.0674,
+      "step": 18436
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.322959183673469e-05,
+      "loss": 0.0651,
+      "step": 18437
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.321428571428571e-05,
+      "loss": 0.1138,
+      "step": 18438
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.319897959183673e-05,
+      "loss": 0.068,
+      "step": 18439
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.318367346938775e-05,
+      "loss": 0.0273,
+      "step": 18440
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.316836734693877e-05,
+      "loss": 0.0217,
+      "step": 18441
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 3.315306122448979e-05,
+      "loss": 0.3248,
+      "step": 18442
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.3137755102040816e-05,
+      "loss": 0.1287,
+      "step": 18443
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.312244897959183e-05,
+      "loss": 0.0271,
+      "step": 18444
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.3107142857142856e-05,
+      "loss": 0.0206,
+      "step": 18445
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.3091836734693873e-05,
+      "loss": 0.0435,
+      "step": 18446
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.30765306122449e-05,
+      "loss": 0.0296,
+      "step": 18447
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.3061224489795914e-05,
+      "loss": 0.0519,
+      "step": 18448
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.304591836734694e-05,
+      "loss": 0.1099,
+      "step": 18449
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.3030612244897955e-05,
+      "loss": 0.0674,
+      "step": 18450
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.301530612244898e-05,
+      "loss": 0.0337,
+      "step": 18451
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.2999999999999996e-05,
+      "loss": 0.0268,
+      "step": 18452
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.298469387755102e-05,
+      "loss": 0.0303,
+      "step": 18453
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.2969387755102037e-05,
+      "loss": 0.1405,
+      "step": 18454
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.295408163265306e-05,
+      "loss": 0.0282,
+      "step": 18455
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.293877551020408e-05,
+      "loss": 0.0691,
+      "step": 18456
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.29234693877551e-05,
+      "loss": 0.0817,
+      "step": 18457
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.290816326530612e-05,
+      "loss": 0.0775,
+      "step": 18458
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.289285714285714e-05,
+      "loss": 0.0227,
+      "step": 18459
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.287755102040816e-05,
+      "loss": 0.0858,
+      "step": 18460
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.286224489795918e-05,
+      "loss": 0.0732,
+      "step": 18461
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.28469387755102e-05,
+      "loss": 0.1536,
+      "step": 18462
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.2831632653061217e-05,
+      "loss": 0.0493,
+      "step": 18463
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.281632653061224e-05,
+      "loss": 0.0491,
+      "step": 18464
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.2801020408163264e-05,
+      "loss": 0.0286,
+      "step": 18465
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.278571428571429e-05,
+      "loss": 0.1149,
+      "step": 18466
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.2770408163265305e-05,
+      "loss": 0.0329,
+      "step": 18467
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.275510204081632e-05,
+      "loss": 0.1525,
+      "step": 18468
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.2739795918367346e-05,
+      "loss": 0.0691,
+      "step": 18469
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.272448979591836e-05,
+      "loss": 0.0766,
+      "step": 18470
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.2709183673469386e-05,
+      "loss": 0.032,
+      "step": 18471
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.2693877551020403e-05,
+      "loss": 0.0839,
+      "step": 18472
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.267857142857143e-05,
+      "loss": 0.1502,
+      "step": 18473
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.2663265306122444e-05,
+      "loss": 0.0313,
+      "step": 18474
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.264795918367347e-05,
+      "loss": 0.0345,
+      "step": 18475
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.2632653061224485e-05,
+      "loss": 0.0361,
+      "step": 18476
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.261734693877551e-05,
+      "loss": 0.0256,
+      "step": 18477
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.2602040816326526e-05,
+      "loss": 0.0187,
+      "step": 18478
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.258673469387755e-05,
+      "loss": 0.1006,
+      "step": 18479
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.2571428571428566e-05,
+      "loss": 0.0425,
+      "step": 18480
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.255612244897959e-05,
+      "loss": 0.0628,
+      "step": 18481
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.255612244897959e-05,
+      "loss": 0.148,
+      "step": 18482
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 3.254081632653061e-05,
+      "loss": 0.064,
+      "step": 18483
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.252551020408163e-05,
+      "loss": 0.0214,
+      "step": 18484
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.251020408163265e-05,
+      "loss": 0.1638,
+      "step": 18485
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.249489795918367e-05,
+      "loss": 0.1505,
+      "step": 18486
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.247959183673469e-05,
+      "loss": 0.0274,
+      "step": 18487
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.246428571428571e-05,
+      "loss": 0.034,
+      "step": 18488
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.2448979591836736e-05,
+      "loss": 0.0692,
+      "step": 18489
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.243367346938775e-05,
+      "loss": 0.08,
+      "step": 18490
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.241836734693878e-05,
+      "loss": 0.053,
+      "step": 18491
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.2403061224489794e-05,
+      "loss": 0.1913,
+      "step": 18492
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.238775510204082e-05,
+      "loss": 0.1113,
+      "step": 18493
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.2372448979591835e-05,
+      "loss": 0.1132,
+      "step": 18494
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.235714285714285e-05,
+      "loss": 0.0745,
+      "step": 18495
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.2341836734693876e-05,
+      "loss": 0.0579,
+      "step": 18496
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.232653061224489e-05,
+      "loss": 0.0279,
+      "step": 18497
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.2311224489795916e-05,
+      "loss": 0.1627,
+      "step": 18498
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.229591836734693e-05,
+      "loss": 0.0671,
+      "step": 18499
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.228061224489796e-05,
+      "loss": 0.0322,
+      "step": 18500
+    },
+    {
+      "epoch": 17.96,
+      "eval_cer": 0.24248313508554956,
+      "eval_loss": 1.2028186321258545,
+      "eval_runtime": 278.3403,
+      "eval_samples_per_second": 31.666,
+      "eval_steps_per_second": 3.959,
+      "eval_wer": 0.6356875281503697,
+      "step": 18500
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.2265306122448974e-05,
+      "loss": 0.0332,
+      "step": 18501
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.225e-05,
+      "loss": 0.0291,
+      "step": 18502
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.2234693877551015e-05,
+      "loss": 0.0626,
+      "step": 18503
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.221938775510204e-05,
+      "loss": 0.0808,
+      "step": 18504
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.2204081632653056e-05,
+      "loss": 0.0745,
+      "step": 18505
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.218877551020408e-05,
+      "loss": 0.1097,
+      "step": 18506
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.2173469387755096e-05,
+      "loss": 0.073,
+      "step": 18507
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.215816326530612e-05,
+      "loss": 0.072,
+      "step": 18508
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.214285714285714e-05,
+      "loss": 0.0188,
+      "step": 18509
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.212755102040816e-05,
+      "loss": 0.1172,
+      "step": 18510
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.2112244897959185e-05,
+      "loss": 0.0344,
+      "step": 18511
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.20969387755102e-05,
+      "loss": 0.069,
+      "step": 18512
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.2081632653061225e-05,
+      "loss": 0.0654,
+      "step": 18513
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.206632653061224e-05,
+      "loss": 0.0463,
+      "step": 18514
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.2051020408163266e-05,
+      "loss": 0.0461,
+      "step": 18515
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.203571428571428e-05,
+      "loss": 0.0927,
+      "step": 18516
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.202040816326531e-05,
+      "loss": 0.0807,
+      "step": 18517
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.2005102040816324e-05,
+      "loss": 0.116,
+      "step": 18518
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.198979591836735e-05,
+      "loss": 0.0279,
+      "step": 18519
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.1974489795918365e-05,
+      "loss": 0.083,
+      "step": 18520
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.195918367346939e-05,
+      "loss": 0.026,
+      "step": 18521
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.1943877551020405e-05,
+      "loss": 0.0249,
+      "step": 18522
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.192857142857142e-05,
+      "loss": 0.1544,
+      "step": 18523
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.1913265306122446e-05,
+      "loss": 0.1116,
+      "step": 18524
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.189795918367346e-05,
+      "loss": 0.0743,
+      "step": 18525
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.188265306122449e-05,
+      "loss": 0.1592,
+      "step": 18526
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.1867346938775504e-05,
+      "loss": 0.0196,
+      "step": 18527
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.185204081632653e-05,
+      "loss": 0.0729,
+      "step": 18528
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.1836734693877545e-05,
+      "loss": 0.1665,
+      "step": 18529
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.182142857142857e-05,
+      "loss": 0.0548,
+      "step": 18530
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.1806122448979586e-05,
+      "loss": 0.0641,
+      "step": 18531
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.179081632653061e-05,
+      "loss": 0.0704,
+      "step": 18532
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.1775510204081626e-05,
+      "loss": 0.0199,
+      "step": 18533
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 3.176020408163265e-05,
+      "loss": 0.1015,
+      "step": 18534
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.1744897959183674e-05,
+      "loss": 0.0682,
+      "step": 18535
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.172959183673469e-05,
+      "loss": 0.0299,
+      "step": 18536
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.1714285714285715e-05,
+      "loss": 0.0804,
+      "step": 18537
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.169897959183673e-05,
+      "loss": 0.1102,
+      "step": 18538
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.1683673469387755e-05,
+      "loss": 0.0182,
+      "step": 18539
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.166836734693877e-05,
+      "loss": 0.0559,
+      "step": 18540
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.1653061224489796e-05,
+      "loss": 0.0739,
+      "step": 18541
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.163775510204081e-05,
+      "loss": 0.0746,
+      "step": 18542
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.162244897959184e-05,
+      "loss": 0.1309,
+      "step": 18543
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.1607142857142854e-05,
+      "loss": 0.1023,
+      "step": 18544
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.159183673469388e-05,
+      "loss": 0.0969,
+      "step": 18545
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.1576530612244895e-05,
+      "loss": 0.0811,
+      "step": 18546
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.156122448979592e-05,
+      "loss": 0.2002,
+      "step": 18547
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.1545918367346935e-05,
+      "loss": 0.0316,
+      "step": 18548
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.153061224489795e-05,
+      "loss": 0.0304,
+      "step": 18549
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.1515306122448976e-05,
+      "loss": 0.0676,
+      "step": 18550
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.149999999999999e-05,
+      "loss": 0.0273,
+      "step": 18551
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.148469387755102e-05,
+      "loss": 0.0195,
+      "step": 18552
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.1469387755102034e-05,
+      "loss": 0.0835,
+      "step": 18553
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.145408163265306e-05,
+      "loss": 0.0993,
+      "step": 18554
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.1438775510204075e-05,
+      "loss": 0.0702,
+      "step": 18555
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.14234693877551e-05,
+      "loss": 0.1189,
+      "step": 18556
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.140816326530612e-05,
+      "loss": 0.0655,
+      "step": 18557
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.139285714285714e-05,
+      "loss": 0.076,
+      "step": 18558
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.137755102040816e-05,
+      "loss": 0.0752,
+      "step": 18559
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.136224489795918e-05,
+      "loss": 0.1107,
+      "step": 18560
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.1346938775510204e-05,
+      "loss": 0.0695,
+      "step": 18561
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.133163265306122e-05,
+      "loss": 0.0519,
+      "step": 18562
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.1316326530612245e-05,
+      "loss": 0.1036,
+      "step": 18563
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.130102040816326e-05,
+      "loss": 0.0895,
+      "step": 18564
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.1285714285714285e-05,
+      "loss": 0.1428,
+      "step": 18565
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.12704081632653e-05,
+      "loss": 0.0312,
+      "step": 18566
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.1255102040816326e-05,
+      "loss": 0.1062,
+      "step": 18567
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.123979591836734e-05,
+      "loss": 0.024,
+      "step": 18568
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.122448979591837e-05,
+      "loss": 0.0698,
+      "step": 18569
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.1209183673469384e-05,
+      "loss": 0.0216,
+      "step": 18570
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.119387755102041e-05,
+      "loss": 0.0873,
+      "step": 18571
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.1178571428571425e-05,
+      "loss": 0.0268,
+      "step": 18572
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.116326530612245e-05,
+      "loss": 0.0662,
+      "step": 18573
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.1147959183673465e-05,
+      "loss": 0.0278,
+      "step": 18574
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.113265306122448e-05,
+      "loss": 0.0242,
+      "step": 18575
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 3.1117346938775506e-05,
+      "loss": 0.031,
+      "step": 18576
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.110204081632652e-05,
+      "loss": 0.0382,
+      "step": 18577
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.108673469387755e-05,
+      "loss": 0.0948,
+      "step": 18578
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.107142857142857e-05,
+      "loss": 0.0311,
+      "step": 18579
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.105612244897959e-05,
+      "loss": 0.0258,
+      "step": 18580
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.104081632653061e-05,
+      "loss": 0.0652,
+      "step": 18581
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.102551020408163e-05,
+      "loss": 0.0638,
+      "step": 18582
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.101020408163265e-05,
+      "loss": 0.0404,
+      "step": 18583
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.099489795918367e-05,
+      "loss": 0.109,
+      "step": 18584
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.097959183673469e-05,
+      "loss": 0.0794,
+      "step": 18585
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.096428571428571e-05,
+      "loss": 0.0323,
+      "step": 18586
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.0948979591836734e-05,
+      "loss": 0.0261,
+      "step": 18587
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.093367346938775e-05,
+      "loss": 0.0967,
+      "step": 18588
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.0918367346938774e-05,
+      "loss": 0.0926,
+      "step": 18589
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.090306122448979e-05,
+      "loss": 0.1081,
+      "step": 18590
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.0887755102040815e-05,
+      "loss": 0.0653,
+      "step": 18591
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.087244897959183e-05,
+      "loss": 0.0292,
+      "step": 18592
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.0857142857142856e-05,
+      "loss": 0.0526,
+      "step": 18593
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.084183673469387e-05,
+      "loss": 0.0682,
+      "step": 18594
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.08265306122449e-05,
+      "loss": 0.0579,
+      "step": 18595
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.0811224489795914e-05,
+      "loss": 0.0575,
+      "step": 18596
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.079591836734694e-05,
+      "loss": 0.0618,
+      "step": 18597
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.0780612244897955e-05,
+      "loss": 0.1127,
+      "step": 18598
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.076530612244898e-05,
+      "loss": 0.1131,
+      "step": 18599
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.0749999999999995e-05,
+      "loss": 0.0282,
+      "step": 18600
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.073469387755102e-05,
+      "loss": 0.0728,
+      "step": 18601
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.071938775510204e-05,
+      "loss": 0.063,
+      "step": 18602
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.070408163265306e-05,
+      "loss": 0.1037,
+      "step": 18603
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.0688775510204084e-05,
+      "loss": 0.084,
+      "step": 18604
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.06734693877551e-05,
+      "loss": 0.0303,
+      "step": 18605
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 3.0658163265306124e-05,
+      "loss": 0.0294,
+      "step": 18606
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.064285714285714e-05,
+      "loss": 0.073,
+      "step": 18607
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.062755102040816e-05,
+      "loss": 0.0296,
+      "step": 18608
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.061224489795918e-05,
+      "loss": 0.0685,
+      "step": 18609
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.05969387755102e-05,
+      "loss": 0.0593,
+      "step": 18610
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.058163265306122e-05,
+      "loss": 0.0661,
+      "step": 18611
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.056632653061224e-05,
+      "loss": 0.1029,
+      "step": 18612
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.0551020408163264e-05,
+      "loss": 0.0705,
+      "step": 18613
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.053571428571428e-05,
+      "loss": 0.0696,
+      "step": 18614
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.0520408163265304e-05,
+      "loss": 0.1167,
+      "step": 18615
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.0505102040816325e-05,
+      "loss": 0.07,
+      "step": 18616
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 3.0489795918367345e-05,
+      "loss": 0.0586,
+      "step": 18617
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.0474489795918362e-05,
+      "loss": 0.0286,
+      "step": 18618
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.0459183673469383e-05,
+      "loss": 0.096,
+      "step": 18619
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.0443877551020403e-05,
+      "loss": 0.1035,
+      "step": 18620
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.0428571428571423e-05,
+      "loss": 0.1,
+      "step": 18621
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.0413265306122444e-05,
+      "loss": 0.0611,
+      "step": 18622
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.0397959183673464e-05,
+      "loss": 0.1814,
+      "step": 18623
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.0382653061224484e-05,
+      "loss": 0.0272,
+      "step": 18624
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.0367346938775508e-05,
+      "loss": 0.0685,
+      "step": 18625
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.035204081632653e-05,
+      "loss": 0.0258,
+      "step": 18626
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 3.033673469387755e-05,
+      "loss": 0.0845,
+      "step": 18627
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.032142857142857e-05,
+      "loss": 0.1293,
+      "step": 18628
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.030612244897959e-05,
+      "loss": 0.075,
+      "step": 18629
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.029081632653061e-05,
+      "loss": 0.0316,
+      "step": 18630
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.027551020408163e-05,
+      "loss": 0.0245,
+      "step": 18631
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.026020408163265e-05,
+      "loss": 0.0619,
+      "step": 18632
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.024489795918367e-05,
+      "loss": 0.0289,
+      "step": 18633
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.022959183673469e-05,
+      "loss": 0.1045,
+      "step": 18634
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.0214285714285712e-05,
+      "loss": 0.0654,
+      "step": 18635
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.0198979591836732e-05,
+      "loss": 0.1517,
+      "step": 18636
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 3.0183673469387753e-05,
+      "loss": 0.0273,
+      "step": 18637
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0168367346938773e-05,
+      "loss": 0.0949,
+      "step": 18638
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0153061224489794e-05,
+      "loss": 0.0441,
+      "step": 18639
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0137755102040814e-05,
+      "loss": 0.0649,
+      "step": 18640
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0122448979591834e-05,
+      "loss": 0.141,
+      "step": 18641
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0107142857142855e-05,
+      "loss": 0.1131,
+      "step": 18642
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0091836734693875e-05,
+      "loss": 0.0735,
+      "step": 18643
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0076530612244892e-05,
+      "loss": 0.0746,
+      "step": 18644
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0061224489795912e-05,
+      "loss": 0.0268,
+      "step": 18645
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.0045918367346933e-05,
+      "loss": 0.1155,
+      "step": 18646
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.003061224489796e-05,
+      "loss": 0.1172,
+      "step": 18647
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 3.001530612244898e-05,
+      "loss": 0.0293,
+      "step": 18648
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 0.0262,
+      "step": 18649
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.9984693877551018e-05,
+      "loss": 0.0305,
+      "step": 18650
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.9969387755102038e-05,
+      "loss": 0.1486,
+      "step": 18651
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.995408163265306e-05,
+      "loss": 0.0407,
+      "step": 18652
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.993877551020408e-05,
+      "loss": 0.1074,
+      "step": 18653
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.99234693877551e-05,
+      "loss": 0.1503,
+      "step": 18654
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.990816326530612e-05,
+      "loss": 0.0658,
+      "step": 18655
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.989285714285714e-05,
+      "loss": 0.0323,
+      "step": 18656
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.987755102040816e-05,
+      "loss": 0.0759,
+      "step": 18657
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.986224489795918e-05,
+      "loss": 0.0323,
+      "step": 18658
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.98469387755102e-05,
+      "loss": 0.1109,
+      "step": 18659
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.983163265306122e-05,
+      "loss": 0.0235,
+      "step": 18660
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.9816326530612242e-05,
+      "loss": 0.0305,
+      "step": 18661
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.9801020408163262e-05,
+      "loss": 0.0223,
+      "step": 18662
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.9785714285714283e-05,
+      "loss": 0.0318,
+      "step": 18663
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.9770408163265303e-05,
+      "loss": 0.0626,
+      "step": 18664
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.9755102040816323e-05,
+      "loss": 0.0495,
+      "step": 18665
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.9739795918367344e-05,
+      "loss": 0.1549,
+      "step": 18666
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.9724489795918364e-05,
+      "loss": 0.0235,
+      "step": 18667
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.9709183673469385e-05,
+      "loss": 0.0362,
+      "step": 18668
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.9693877551020405e-05,
+      "loss": 0.1119,
+      "step": 18669
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.967857142857143e-05,
+      "loss": 0.084,
+      "step": 18670
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.966326530612245e-05,
+      "loss": 0.0337,
+      "step": 18671
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.964795918367347e-05,
+      "loss": 0.0724,
+      "step": 18672
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.963265306122449e-05,
+      "loss": 0.0785,
+      "step": 18673
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.961734693877551e-05,
+      "loss": 0.0293,
+      "step": 18674
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.960204081632653e-05,
+      "loss": 0.0311,
+      "step": 18675
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.9586734693877548e-05,
+      "loss": 0.043,
+      "step": 18676
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.9571428571428568e-05,
+      "loss": 0.0817,
+      "step": 18677
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.955612244897959e-05,
+      "loss": 0.0709,
+      "step": 18678
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.954081632653061e-05,
+      "loss": 0.0267,
+      "step": 18679
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.952551020408163e-05,
+      "loss": 0.0517,
+      "step": 18680
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.951020408163265e-05,
+      "loss": 0.1824,
+      "step": 18681
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.949489795918367e-05,
+      "loss": 0.0662,
+      "step": 18682
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.947959183673469e-05,
+      "loss": 0.0271,
+      "step": 18683
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.946428571428571e-05,
+      "loss": 0.0606,
+      "step": 18684
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.944897959183673e-05,
+      "loss": 0.1297,
+      "step": 18685
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.943367346938775e-05,
+      "loss": 0.0591,
+      "step": 18686
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.9418367346938772e-05,
+      "loss": 0.0236,
+      "step": 18687
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.9403061224489792e-05,
+      "loss": 0.1493,
+      "step": 18688
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.9387755102040813e-05,
+      "loss": 0.1527,
+      "step": 18689
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.9372448979591833e-05,
+      "loss": 0.1141,
+      "step": 18690
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.9357142857142853e-05,
+      "loss": 0.0644,
+      "step": 18691
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.9341836734693874e-05,
+      "loss": 0.0652,
+      "step": 18692
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.9326530612244898e-05,
+      "loss": 0.1774,
+      "step": 18693
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.9311224489795918e-05,
+      "loss": 0.0371,
+      "step": 18694
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.929591836734694e-05,
+      "loss": 0.026,
+      "step": 18695
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.928061224489796e-05,
+      "loss": 0.0251,
+      "step": 18696
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.926530612244898e-05,
+      "loss": 0.092,
+      "step": 18697
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.925e-05,
+      "loss": 0.0602,
+      "step": 18698
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.923469387755102e-05,
+      "loss": 0.0226,
+      "step": 18699
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.921938775510204e-05,
+      "loss": 0.1194,
+      "step": 18700
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.920408163265306e-05,
+      "loss": 0.0305,
+      "step": 18701
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.918877551020408e-05,
+      "loss": 0.0781,
+      "step": 18702
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.9173469387755098e-05,
+      "loss": 0.0595,
+      "step": 18703
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.915816326530612e-05,
+      "loss": 0.0668,
+      "step": 18704
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.914285714285714e-05,
+      "loss": 0.111,
+      "step": 18705
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.912755102040816e-05,
+      "loss": 0.027,
+      "step": 18706
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.911224489795918e-05,
+      "loss": 0.0208,
+      "step": 18707
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.90969387755102e-05,
+      "loss": 0.0773,
+      "step": 18708
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.908163265306122e-05,
+      "loss": 0.1245,
+      "step": 18709
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.906632653061224e-05,
+      "loss": 0.0441,
+      "step": 18710
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.905102040816326e-05,
+      "loss": 0.0246,
+      "step": 18711
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.903571428571428e-05,
+      "loss": 0.08,
+      "step": 18712
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.9020408163265302e-05,
+      "loss": 0.0901,
+      "step": 18713
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.9005102040816322e-05,
+      "loss": 0.0836,
+      "step": 18714
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.8989795918367346e-05,
+      "loss": 0.1283,
+      "step": 18715
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.8974489795918366e-05,
+      "loss": 0.0249,
+      "step": 18716
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.8959183673469387e-05,
+      "loss": 0.0314,
+      "step": 18717
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.8943877551020407e-05,
+      "loss": 0.0338,
+      "step": 18718
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.8928571428571427e-05,
+      "loss": 0.0619,
+      "step": 18719
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.8913265306122448e-05,
+      "loss": 0.0251,
+      "step": 18720
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.8897959183673468e-05,
+      "loss": 0.0522,
+      "step": 18721
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.888265306122449e-05,
+      "loss": 0.2871,
+      "step": 18722
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.886734693877551e-05,
+      "loss": 0.0361,
+      "step": 18723
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.885204081632653e-05,
+      "loss": 0.103,
+      "step": 18724
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.883673469387755e-05,
+      "loss": 0.1254,
+      "step": 18725
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.882142857142857e-05,
+      "loss": 0.0591,
+      "step": 18726
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.880612244897959e-05,
+      "loss": 0.028,
+      "step": 18727
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.879081632653061e-05,
+      "loss": 0.0943,
+      "step": 18728
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.8775510204081628e-05,
+      "loss": 0.0349,
+      "step": 18729
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.8760204081632648e-05,
+      "loss": 0.0997,
+      "step": 18730
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.874489795918367e-05,
+      "loss": 0.0324,
+      "step": 18731
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.872959183673469e-05,
+      "loss": 0.0229,
+      "step": 18732
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.871428571428571e-05,
+      "loss": 0.0276,
+      "step": 18733
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.869897959183673e-05,
+      "loss": 0.0888,
+      "step": 18734
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.868367346938775e-05,
+      "loss": 0.0731,
+      "step": 18735
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.866836734693877e-05,
+      "loss": 0.1395,
+      "step": 18736
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.865306122448979e-05,
+      "loss": 0.0699,
+      "step": 18737
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.8637755102040815e-05,
+      "loss": 0.0902,
+      "step": 18738
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.8622448979591835e-05,
+      "loss": 0.0202,
+      "step": 18739
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.8607142857142856e-05,
+      "loss": 0.2143,
+      "step": 18740
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.8591836734693876e-05,
+      "loss": 0.0715,
+      "step": 18741
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.8576530612244896e-05,
+      "loss": 0.0335,
+      "step": 18742
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.8561224489795917e-05,
+      "loss": 0.0246,
+      "step": 18743
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.8545918367346937e-05,
+      "loss": 0.1495,
+      "step": 18744
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.8530612244897957e-05,
+      "loss": 0.1018,
+      "step": 18745
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.8515306122448978e-05,
+      "loss": 0.0418,
+      "step": 18746
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 0.1447,
+      "step": 18747
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.848469387755102e-05,
+      "loss": 0.0279,
+      "step": 18748
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.846938775510204e-05,
+      "loss": 0.0191,
+      "step": 18749
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.845408163265306e-05,
+      "loss": 0.0247,
+      "step": 18750
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.843877551020408e-05,
+      "loss": 0.0257,
+      "step": 18751
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.84234693877551e-05,
+      "loss": 0.1396,
+      "step": 18752
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.840816326530612e-05,
+      "loss": 0.1521,
+      "step": 18753
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.839285714285714e-05,
+      "loss": 0.0255,
+      "step": 18754
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.837755102040816e-05,
+      "loss": 0.0286,
+      "step": 18755
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.8362244897959178e-05,
+      "loss": 0.0955,
+      "step": 18756
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.83469387755102e-05,
+      "loss": 0.0684,
+      "step": 18757
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.833163265306122e-05,
+      "loss": 0.0565,
+      "step": 18758
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.831632653061224e-05,
+      "loss": 0.1074,
+      "step": 18759
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.830102040816326e-05,
+      "loss": 0.1761,
+      "step": 18760
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.8285714285714284e-05,
+      "loss": 0.0593,
+      "step": 18761
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8270408163265304e-05,
+      "loss": 0.0249,
+      "step": 18762
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8255102040816324e-05,
+      "loss": 0.0947,
+      "step": 18763
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8239795918367345e-05,
+      "loss": 0.0245,
+      "step": 18764
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8224489795918365e-05,
+      "loss": 0.2014,
+      "step": 18765
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8209183673469385e-05,
+      "loss": 0.0317,
+      "step": 18766
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8193877551020406e-05,
+      "loss": 0.103,
+      "step": 18767
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8178571428571426e-05,
+      "loss": 0.0395,
+      "step": 18768
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8163265306122447e-05,
+      "loss": 0.0243,
+      "step": 18769
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8147959183673467e-05,
+      "loss": 0.0515,
+      "step": 18770
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.8132653061224487e-05,
+      "loss": 0.072,
+      "step": 18771
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.8117346938775508e-05,
+      "loss": 0.1084,
+      "step": 18772
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.8102040816326528e-05,
+      "loss": 0.0252,
+      "step": 18773
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.808673469387755e-05,
+      "loss": 0.0587,
+      "step": 18774
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.807142857142857e-05,
+      "loss": 0.1049,
+      "step": 18775
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.805612244897959e-05,
+      "loss": 0.0668,
+      "step": 18776
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.804081632653061e-05,
+      "loss": 0.0535,
+      "step": 18777
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.802551020408163e-05,
+      "loss": 0.2726,
+      "step": 18778
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.801020408163265e-05,
+      "loss": 0.0366,
+      "step": 18779
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.799489795918367e-05,
+      "loss": 0.043,
+      "step": 18780
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.797959183673469e-05,
+      "loss": 0.1005,
+      "step": 18781
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.796428571428571e-05,
+      "loss": 0.0231,
+      "step": 18782
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7948979591836735e-05,
+      "loss": 0.027,
+      "step": 18783
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7933673469387756e-05,
+      "loss": 0.0385,
+      "step": 18784
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7918367346938776e-05,
+      "loss": 0.0193,
+      "step": 18785
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7903061224489796e-05,
+      "loss": 0.0573,
+      "step": 18786
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7887755102040817e-05,
+      "loss": 0.0246,
+      "step": 18787
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7872448979591834e-05,
+      "loss": 0.1498,
+      "step": 18788
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7857142857142854e-05,
+      "loss": 0.0287,
+      "step": 18789
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7841836734693875e-05,
+      "loss": 0.1213,
+      "step": 18790
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7826530612244895e-05,
+      "loss": 0.1626,
+      "step": 18791
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7811224489795915e-05,
+      "loss": 0.0409,
+      "step": 18792
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7795918367346936e-05,
+      "loss": 0.1104,
+      "step": 18793
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7780612244897956e-05,
+      "loss": 0.1139,
+      "step": 18794
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7765306122448977e-05,
+      "loss": 0.0659,
+      "step": 18795
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7749999999999997e-05,
+      "loss": 0.0981,
+      "step": 18796
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7734693877551017e-05,
+      "loss": 0.1094,
+      "step": 18797
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7719387755102038e-05,
+      "loss": 0.0649,
+      "step": 18798
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7704081632653058e-05,
+      "loss": 0.0263,
+      "step": 18799
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.768877551020408e-05,
+      "loss": 0.0725,
+      "step": 18800
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.76734693877551e-05,
+      "loss": 0.0244,
+      "step": 18801
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.765816326530612e-05,
+      "loss": 0.0497,
+      "step": 18802
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.764285714285714e-05,
+      "loss": 0.2186,
+      "step": 18803
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.762755102040816e-05,
+      "loss": 0.1012,
+      "step": 18804
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.761224489795918e-05,
+      "loss": 0.0261,
+      "step": 18805
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.7596938775510204e-05,
+      "loss": 0.0916,
+      "step": 18806
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.7581632653061224e-05,
+      "loss": 0.0242,
+      "step": 18807
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.7566326530612245e-05,
+      "loss": 0.0705,
+      "step": 18808
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.7551020408163265e-05,
+      "loss": 0.1087,
+      "step": 18809
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.7535714285714286e-05,
+      "loss": 0.0679,
+      "step": 18810
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.7520408163265306e-05,
+      "loss": 0.0817,
+      "step": 18811
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.7505102040816326e-05,
+      "loss": 0.0338,
+      "step": 18812
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7489795918367347e-05,
+      "loss": 0.0738,
+      "step": 18813
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7474489795918367e-05,
+      "loss": 0.0278,
+      "step": 18814
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7459183673469384e-05,
+      "loss": 0.0501,
+      "step": 18815
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7443877551020405e-05,
+      "loss": 0.1088,
+      "step": 18816
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7428571428571425e-05,
+      "loss": 0.0569,
+      "step": 18817
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7413265306122445e-05,
+      "loss": 0.0621,
+      "step": 18818
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7397959183673466e-05,
+      "loss": 0.0732,
+      "step": 18819
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7382653061224486e-05,
+      "loss": 0.0256,
+      "step": 18820
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7367346938775506e-05,
+      "loss": 0.1001,
+      "step": 18821
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7352040816326527e-05,
+      "loss": 0.0765,
+      "step": 18822
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7336734693877547e-05,
+      "loss": 0.1092,
+      "step": 18823
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.7321428571428568e-05,
+      "loss": 0.0339,
+      "step": 18824
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.7306122448979588e-05,
+      "loss": 0.0713,
+      "step": 18825
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.729081632653061e-05,
+      "loss": 0.0768,
+      "step": 18826
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.727551020408163e-05,
+      "loss": 0.0285,
+      "step": 18827
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.726020408163265e-05,
+      "loss": 0.2423,
+      "step": 18828
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.7244897959183673e-05,
+      "loss": 0.0658,
+      "step": 18829
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.7229591836734693e-05,
+      "loss": 0.0252,
+      "step": 18830
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.7214285714285714e-05,
+      "loss": 0.0242,
+      "step": 18831
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.7198979591836734e-05,
+      "loss": 0.0348,
+      "step": 18832
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.7183673469387754e-05,
+      "loss": 0.0414,
+      "step": 18833
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7168367346938775e-05,
+      "loss": 0.0656,
+      "step": 18834
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7153061224489795e-05,
+      "loss": 0.1582,
+      "step": 18835
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7137755102040816e-05,
+      "loss": 0.0651,
+      "step": 18836
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7122448979591836e-05,
+      "loss": 0.056,
+      "step": 18837
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7107142857142856e-05,
+      "loss": 0.0572,
+      "step": 18838
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7091836734693877e-05,
+      "loss": 0.0757,
+      "step": 18839
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7076530612244897e-05,
+      "loss": 0.1047,
+      "step": 18840
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7061224489795914e-05,
+      "loss": 0.0894,
+      "step": 18841
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7045918367346934e-05,
+      "loss": 0.0659,
+      "step": 18842
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7030612244897955e-05,
+      "loss": 0.0964,
+      "step": 18843
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.7015306122448975e-05,
+      "loss": 0.0319,
+      "step": 18844
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.6999999999999996e-05,
+      "loss": 0.1031,
+      "step": 18845
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.6984693877551016e-05,
+      "loss": 0.1148,
+      "step": 18846
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.6969387755102036e-05,
+      "loss": 0.0457,
+      "step": 18847
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.6954081632653057e-05,
+      "loss": 0.1117,
+      "step": 18848
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.6938775510204077e-05,
+      "loss": 0.0749,
+      "step": 18849
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.6923469387755098e-05,
+      "loss": 0.0645,
+      "step": 18850
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.690816326530612e-05,
+      "loss": 0.1381,
+      "step": 18851
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.689285714285714e-05,
+      "loss": 0.0594,
+      "step": 18852
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.6877551020408162e-05,
+      "loss": 0.1138,
+      "step": 18853
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.6862244897959182e-05,
+      "loss": 0.1118,
+      "step": 18854
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6846938775510203e-05,
+      "loss": 0.035,
+      "step": 18855
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6831632653061223e-05,
+      "loss": 0.0712,
+      "step": 18856
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6816326530612244e-05,
+      "loss": 0.0605,
+      "step": 18857
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6801020408163264e-05,
+      "loss": 0.088,
+      "step": 18858
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6785714285714284e-05,
+      "loss": 0.0843,
+      "step": 18859
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6770408163265305e-05,
+      "loss": 0.1129,
+      "step": 18860
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6755102040816325e-05,
+      "loss": 0.0708,
+      "step": 18861
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6739795918367345e-05,
+      "loss": 0.1593,
+      "step": 18862
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6724489795918366e-05,
+      "loss": 0.0586,
+      "step": 18863
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.6709183673469386e-05,
+      "loss": 0.0223,
+      "step": 18864
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6693877551020407e-05,
+      "loss": 0.1083,
+      "step": 18865
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6678571428571427e-05,
+      "loss": 0.1159,
+      "step": 18866
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6663265306122447e-05,
+      "loss": 0.1031,
+      "step": 18867
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6647959183673464e-05,
+      "loss": 0.0179,
+      "step": 18868
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6632653061224485e-05,
+      "loss": 0.0287,
+      "step": 18869
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6617346938775505e-05,
+      "loss": 0.0958,
+      "step": 18870
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6602040816326526e-05,
+      "loss": 0.0949,
+      "step": 18871
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6586734693877546e-05,
+      "loss": 0.1077,
+      "step": 18872
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6571428571428566e-05,
+      "loss": 0.0737,
+      "step": 18873
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.655612244897959e-05,
+      "loss": 0.0257,
+      "step": 18874
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.654081632653061e-05,
+      "loss": 0.0562,
+      "step": 18875
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.652551020408163e-05,
+      "loss": 0.0593,
+      "step": 18876
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.651020408163265e-05,
+      "loss": 0.0316,
+      "step": 18877
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.649489795918367e-05,
+      "loss": 0.1719,
+      "step": 18878
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.6479591836734692e-05,
+      "loss": 0.105,
+      "step": 18879
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.6464285714285712e-05,
+      "loss": 0.0286,
+      "step": 18880
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.6448979591836733e-05,
+      "loss": 0.0197,
+      "step": 18881
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.6433673469387753e-05,
+      "loss": 0.0397,
+      "step": 18882
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.6418367346938774e-05,
+      "loss": 0.0369,
+      "step": 18883
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.6403061224489794e-05,
+      "loss": 0.064,
+      "step": 18884
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.6387755102040814e-05,
+      "loss": 0.1125,
+      "step": 18885
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6372448979591835e-05,
+      "loss": 0.0306,
+      "step": 18886
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6357142857142855e-05,
+      "loss": 0.083,
+      "step": 18887
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6341836734693875e-05,
+      "loss": 0.0355,
+      "step": 18888
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6326530612244896e-05,
+      "loss": 0.0347,
+      "step": 18889
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6311224489795916e-05,
+      "loss": 0.1051,
+      "step": 18890
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6295918367346937e-05,
+      "loss": 0.0303,
+      "step": 18891
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6280612244897957e-05,
+      "loss": 0.0841,
+      "step": 18892
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6265306122448977e-05,
+      "loss": 0.0717,
+      "step": 18893
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6249999999999998e-05,
+      "loss": 0.059,
+      "step": 18894
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6234693877551015e-05,
+      "loss": 0.0349,
+      "step": 18895
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.6219387755102035e-05,
+      "loss": 0.0388,
+      "step": 18896
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.6204081632653062e-05,
+      "loss": 0.2397,
+      "step": 18897
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.6188775510204083e-05,
+      "loss": 0.0322,
+      "step": 18898
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.6173469387755103e-05,
+      "loss": 0.1086,
+      "step": 18899
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.615816326530612e-05,
+      "loss": 0.063,
+      "step": 18900
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.614285714285714e-05,
+      "loss": 0.1161,
+      "step": 18901
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.612755102040816e-05,
+      "loss": 0.0616,
+      "step": 18902
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.611224489795918e-05,
+      "loss": 0.1155,
+      "step": 18903
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.60969387755102e-05,
+      "loss": 0.1849,
+      "step": 18904
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.6081632653061222e-05,
+      "loss": 0.0782,
+      "step": 18905
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.6066326530612242e-05,
+      "loss": 0.1112,
+      "step": 18906
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.6051020408163263e-05,
+      "loss": 0.0627,
+      "step": 18907
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.6035714285714283e-05,
+      "loss": 0.0257,
+      "step": 18908
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.6020408163265303e-05,
+      "loss": 0.1093,
+      "step": 18909
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.6005102040816324e-05,
+      "loss": 0.0622,
+      "step": 18910
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.5989795918367344e-05,
+      "loss": 0.0209,
+      "step": 18911
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.5974489795918365e-05,
+      "loss": 0.065,
+      "step": 18912
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.5959183673469385e-05,
+      "loss": 0.0472,
+      "step": 18913
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.5943877551020405e-05,
+      "loss": 0.0452,
+      "step": 18914
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.5928571428571426e-05,
+      "loss": 0.0801,
+      "step": 18915
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.5913265306122446e-05,
+      "loss": 0.1076,
+      "step": 18916
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.5897959183673466e-05,
+      "loss": 0.0582,
+      "step": 18917
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.5882653061224487e-05,
+      "loss": 0.0622,
+      "step": 18918
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.586734693877551e-05,
+      "loss": 0.1,
+      "step": 18919
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.585204081632653e-05,
+      "loss": 0.0597,
+      "step": 18920
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.583673469387755e-05,
+      "loss": 0.0631,
+      "step": 18921
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.5821428571428572e-05,
+      "loss": 0.1067,
+      "step": 18922
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.5806122448979592e-05,
+      "loss": 0.0929,
+      "step": 18923
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.5790816326530613e-05,
+      "loss": 0.1013,
+      "step": 18924
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.5775510204081633e-05,
+      "loss": 0.0697,
+      "step": 18925
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.576020408163265e-05,
+      "loss": 0.0323,
+      "step": 18926
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.574489795918367e-05,
+      "loss": 0.0292,
+      "step": 18927
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.572959183673469e-05,
+      "loss": 0.1091,
+      "step": 18928
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.571428571428571e-05,
+      "loss": 0.1497,
+      "step": 18929
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.569897959183673e-05,
+      "loss": 0.1043,
+      "step": 18930
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.5683673469387752e-05,
+      "loss": 0.0263,
+      "step": 18931
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.5668367346938772e-05,
+      "loss": 0.0589,
+      "step": 18932
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.5653061224489793e-05,
+      "loss": 0.063,
+      "step": 18933
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.5637755102040813e-05,
+      "loss": 0.0704,
+      "step": 18934
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.5622448979591833e-05,
+      "loss": 0.035,
+      "step": 18935
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.5607142857142854e-05,
+      "loss": 0.0646,
+      "step": 18936
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.5591836734693874e-05,
+      "loss": 0.0338,
+      "step": 18937
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.5576530612244895e-05,
+      "loss": 0.0212,
+      "step": 18938
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.5561224489795915e-05,
+      "loss": 0.0257,
+      "step": 18939
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.5545918367346935e-05,
+      "loss": 0.0287,
+      "step": 18940
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.5530612244897956e-05,
+      "loss": 0.0658,
+      "step": 18941
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.551530612244898e-05,
+      "loss": 0.1096,
+      "step": 18942
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.55e-05,
+      "loss": 0.0274,
+      "step": 18943
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.548469387755102e-05,
+      "loss": 0.0278,
+      "step": 18944
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.546938775510204e-05,
+      "loss": 0.0337,
+      "step": 18945
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.545408163265306e-05,
+      "loss": 0.037,
+      "step": 18946
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.543877551020408e-05,
+      "loss": 0.0924,
+      "step": 18947
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.5423469387755102e-05,
+      "loss": 0.0293,
+      "step": 18948
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.5408163265306122e-05,
+      "loss": 0.1037,
+      "step": 18949
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.5392857142857142e-05,
+      "loss": 0.1197,
+      "step": 18950
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.5377551020408163e-05,
+      "loss": 0.0172,
+      "step": 18951
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.5362244897959183e-05,
+      "loss": 0.0655,
+      "step": 18952
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.53469387755102e-05,
+      "loss": 0.1229,
+      "step": 18953
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.533163265306122e-05,
+      "loss": 0.069,
+      "step": 18954
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.531632653061224e-05,
+      "loss": 0.0612,
+      "step": 18955
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.530102040816326e-05,
+      "loss": 0.0165,
+      "step": 18956
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.5285714285714282e-05,
+      "loss": 0.0245,
+      "step": 18957
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.5270408163265302e-05,
+      "loss": 0.0995,
+      "step": 18958
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.5255102040816323e-05,
+      "loss": 0.0901,
+      "step": 18959
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.5239795918367343e-05,
+      "loss": 0.0608,
+      "step": 18960
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.5224489795918363e-05,
+      "loss": 0.0657,
+      "step": 18961
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.5209183673469384e-05,
+      "loss": 0.0416,
+      "step": 18962
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.5193877551020404e-05,
+      "loss": 0.0308,
+      "step": 18963
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.5178571428571424e-05,
+      "loss": 0.0549,
+      "step": 18964
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.5163265306122448e-05,
+      "loss": 0.0568,
+      "step": 18965
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.514795918367347e-05,
+      "loss": 0.0362,
+      "step": 18966
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.513265306122449e-05,
+      "loss": 0.0977,
+      "step": 18967
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.511734693877551e-05,
+      "loss": 0.1061,
+      "step": 18968
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.510204081632653e-05,
+      "loss": 0.0233,
+      "step": 18969
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.508673469387755e-05,
+      "loss": 0.0284,
+      "step": 18970
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.507142857142857e-05,
+      "loss": 0.0834,
+      "step": 18971
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.505612244897959e-05,
+      "loss": 0.0451,
+      "step": 18972
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.504081632653061e-05,
+      "loss": 0.0237,
+      "step": 18973
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.502551020408163e-05,
+      "loss": 0.1374,
+      "step": 18974
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.5010204081632652e-05,
+      "loss": 0.112,
+      "step": 18975
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.4994897959183672e-05,
+      "loss": 0.072,
+      "step": 18976
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.4979591836734693e-05,
+      "loss": 0.0784,
+      "step": 18977
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.4964285714285713e-05,
+      "loss": 0.1516,
+      "step": 18978
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.4948979591836734e-05,
+      "loss": 0.0991,
+      "step": 18979
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.493367346938775e-05,
+      "loss": 0.0742,
+      "step": 18980
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.491836734693877e-05,
+      "loss": 0.0221,
+      "step": 18981
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.490306122448979e-05,
+      "loss": 0.0956,
+      "step": 18982
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.488775510204081e-05,
+      "loss": 0.0407,
+      "step": 18983
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.4872448979591832e-05,
+      "loss": 0.0259,
+      "step": 18984
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.4857142857142852e-05,
+      "loss": 0.0389,
+      "step": 18985
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.4841836734693873e-05,
+      "loss": 0.0276,
+      "step": 18986
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.4826530612244893e-05,
+      "loss": 0.1453,
+      "step": 18987
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.4811224489795917e-05,
+      "loss": 0.0399,
+      "step": 18988
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.4795918367346937e-05,
+      "loss": 0.0309,
+      "step": 18989
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.4780612244897958e-05,
+      "loss": 0.0699,
+      "step": 18990
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.4765306122448978e-05,
+      "loss": 0.0328,
+      "step": 18991
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.475e-05,
+      "loss": 0.0152,
+      "step": 18992
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.473469387755102e-05,
+      "loss": 0.0625,
+      "step": 18993
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.471938775510204e-05,
+      "loss": 0.0974,
+      "step": 18994
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.470408163265306e-05,
+      "loss": 0.028,
+      "step": 18995
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.468877551020408e-05,
+      "loss": 0.1039,
+      "step": 18996
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.46734693877551e-05,
+      "loss": 0.1007,
+      "step": 18997
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.465816326530612e-05,
+      "loss": 0.1095,
+      "step": 18998
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.464285714285714e-05,
+      "loss": 0.0642,
+      "step": 18999
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.462755102040816e-05,
+      "loss": 0.031,
+      "step": 19000
+    },
+    {
+      "epoch": 18.45,
+      "eval_cer": 0.24158930114718843,
+      "eval_loss": 1.238124966621399,
+      "eval_runtime": 231.0957,
+      "eval_samples_per_second": 38.14,
+      "eval_steps_per_second": 4.769,
+      "eval_wer": 0.6285359610673522,
+      "step": 19000
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.4612244897959182e-05,
+      "loss": 0.1036,
+      "step": 19001
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.4596938775510202e-05,
+      "loss": 0.0669,
+      "step": 19002
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.4581632653061223e-05,
+      "loss": 0.0625,
+      "step": 19003
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.4566326530612243e-05,
+      "loss": 0.0333,
+      "step": 19004
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.4551020408163263e-05,
+      "loss": 0.101,
+      "step": 19005
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.4535714285714284e-05,
+      "loss": 0.0291,
+      "step": 19006
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.45204081632653e-05,
+      "loss": 0.0281,
+      "step": 19007
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.450510204081632e-05,
+      "loss": 0.0324,
+      "step": 19008
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.448979591836734e-05,
+      "loss": 0.0846,
+      "step": 19009
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.447448979591837e-05,
+      "loss": 0.1034,
+      "step": 19010
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.4459183673469386e-05,
+      "loss": 0.1055,
+      "step": 19011
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.4443877551020406e-05,
+      "loss": 0.0578,
+      "step": 19012
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.4428571428571427e-05,
+      "loss": 0.073,
+      "step": 19013
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.4413265306122447e-05,
+      "loss": 0.0639,
+      "step": 19014
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.4397959183673467e-05,
+      "loss": 0.0838,
+      "step": 19015
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.4382653061224488e-05,
+      "loss": 0.034,
+      "step": 19016
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.4367346938775508e-05,
+      "loss": 0.1483,
+      "step": 19017
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.435204081632653e-05,
+      "loss": 0.0711,
+      "step": 19018
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.433673469387755e-05,
+      "loss": 0.0348,
+      "step": 19019
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.432142857142857e-05,
+      "loss": 0.0627,
+      "step": 19020
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.430612244897959e-05,
+      "loss": 0.1079,
+      "step": 19021
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.429081632653061e-05,
+      "loss": 0.0343,
+      "step": 19022
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.427551020408163e-05,
+      "loss": 0.1121,
+      "step": 19023
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.426020408163265e-05,
+      "loss": 0.0623,
+      "step": 19024
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.424489795918367e-05,
+      "loss": 0.1069,
+      "step": 19025
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.422959183673469e-05,
+      "loss": 0.0644,
+      "step": 19026
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.4214285714285712e-05,
+      "loss": 0.0267,
+      "step": 19027
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.4198979591836732e-05,
+      "loss": 0.05,
+      "step": 19028
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.4183673469387753e-05,
+      "loss": 0.1163,
+      "step": 19029
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.4168367346938773e-05,
+      "loss": 0.0289,
+      "step": 19030
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.4153061224489793e-05,
+      "loss": 0.0652,
+      "step": 19031
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.4137755102040814e-05,
+      "loss": 0.0255,
+      "step": 19032
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.4122448979591838e-05,
+      "loss": 0.0251,
+      "step": 19033
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.4107142857142858e-05,
+      "loss": 0.1673,
+      "step": 19034
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.409183673469388e-05,
+      "loss": 0.1123,
+      "step": 19035
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.40765306122449e-05,
+      "loss": 0.065,
+      "step": 19036
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.406122448979592e-05,
+      "loss": 0.0355,
+      "step": 19037
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.4045918367346936e-05,
+      "loss": 0.0557,
+      "step": 19038
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.4030612244897956e-05,
+      "loss": 0.113,
+      "step": 19039
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.4015306122448977e-05,
+      "loss": 0.0983,
+      "step": 19040
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.3999999999999997e-05,
+      "loss": 0.1869,
+      "step": 19041
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.3984693877551018e-05,
+      "loss": 0.1386,
+      "step": 19042
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.3969387755102038e-05,
+      "loss": 0.0277,
+      "step": 19043
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.395408163265306e-05,
+      "loss": 0.1043,
+      "step": 19044
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.393877551020408e-05,
+      "loss": 0.0561,
+      "step": 19045
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.39234693877551e-05,
+      "loss": 0.0534,
+      "step": 19046
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.390816326530612e-05,
+      "loss": 0.0702,
+      "step": 19047
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.389285714285714e-05,
+      "loss": 0.1012,
+      "step": 19048
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.387755102040816e-05,
+      "loss": 0.0714,
+      "step": 19049
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.386224489795918e-05,
+      "loss": 0.1086,
+      "step": 19050
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.38469387755102e-05,
+      "loss": 0.0235,
+      "step": 19051
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.383163265306122e-05,
+      "loss": 0.0395,
+      "step": 19052
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.3816326530612242e-05,
+      "loss": 0.1881,
+      "step": 19053
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.3801020408163262e-05,
+      "loss": 0.0291,
+      "step": 19054
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.3785714285714283e-05,
+      "loss": 0.1441,
+      "step": 19055
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.3770408163265306e-05,
+      "loss": 0.057,
+      "step": 19056
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.3755102040816327e-05,
+      "loss": 0.0368,
+      "step": 19057
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.3739795918367347e-05,
+      "loss": 0.123,
+      "step": 19058
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.3724489795918367e-05,
+      "loss": 0.1292,
+      "step": 19059
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.3709183673469388e-05,
+      "loss": 0.0326,
+      "step": 19060
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.3693877551020408e-05,
+      "loss": 0.0334,
+      "step": 19061
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.367857142857143e-05,
+      "loss": 0.093,
+      "step": 19062
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.366326530612245e-05,
+      "loss": 0.1135,
+      "step": 19063
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.364795918367347e-05,
+      "loss": 0.0387,
+      "step": 19064
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.3632653061224486e-05,
+      "loss": 0.0728,
+      "step": 19065
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.3617346938775507e-05,
+      "loss": 0.1215,
+      "step": 19066
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.3602040816326527e-05,
+      "loss": 0.0617,
+      "step": 19067
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.3586734693877548e-05,
+      "loss": 0.0668,
+      "step": 19068
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.3571428571428568e-05,
+      "loss": 0.061,
+      "step": 19069
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.3556122448979588e-05,
+      "loss": 0.0299,
+      "step": 19070
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.354081632653061e-05,
+      "loss": 0.0771,
+      "step": 19071
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.352551020408163e-05,
+      "loss": 0.0504,
+      "step": 19072
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.351020408163265e-05,
+      "loss": 0.0644,
+      "step": 19073
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.349489795918367e-05,
+      "loss": 0.0671,
+      "step": 19074
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.347959183673469e-05,
+      "loss": 0.0621,
+      "step": 19075
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.346428571428571e-05,
+      "loss": 0.1488,
+      "step": 19076
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.344897959183673e-05,
+      "loss": 0.0232,
+      "step": 19077
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.3433673469387755e-05,
+      "loss": 0.1027,
+      "step": 19078
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.3418367346938775e-05,
+      "loss": 0.0254,
+      "step": 19079
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.3403061224489796e-05,
+      "loss": 0.1238,
+      "step": 19080
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.3387755102040816e-05,
+      "loss": 0.1048,
+      "step": 19081
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.3372448979591836e-05,
+      "loss": 0.0275,
+      "step": 19082
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.3357142857142857e-05,
+      "loss": 0.0546,
+      "step": 19083
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.3341836734693877e-05,
+      "loss": 0.1419,
+      "step": 19084
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.3326530612244897e-05,
+      "loss": 0.1101,
+      "step": 19085
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.3311224489795918e-05,
+      "loss": 0.1081,
+      "step": 19086
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.3295918367346938e-05,
+      "loss": 0.0718,
+      "step": 19087
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.328061224489796e-05,
+      "loss": 0.027,
+      "step": 19088
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.326530612244898e-05,
+      "loss": 0.0268,
+      "step": 19089
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.325e-05,
+      "loss": 0.0343,
+      "step": 19090
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.323469387755102e-05,
+      "loss": 0.1494,
+      "step": 19091
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.3219387755102037e-05,
+      "loss": 0.1471,
+      "step": 19092
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.3204081632653057e-05,
+      "loss": 0.0991,
+      "step": 19093
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.3188775510204077e-05,
+      "loss": 0.029,
+      "step": 19094
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.3173469387755098e-05,
+      "loss": 0.1046,
+      "step": 19095
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.3158163265306118e-05,
+      "loss": 0.0707,
+      "step": 19096
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.314285714285714e-05,
+      "loss": 0.074,
+      "step": 19097
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.312755102040816e-05,
+      "loss": 0.0734,
+      "step": 19098
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.311224489795918e-05,
+      "loss": 0.0254,
+      "step": 19099
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.30969387755102e-05,
+      "loss": 0.0814,
+      "step": 19100
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.3081632653061224e-05,
+      "loss": 0.0947,
+      "step": 19101
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.3066326530612244e-05,
+      "loss": 0.0286,
+      "step": 19102
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.3051020408163264e-05,
+      "loss": 0.1098,
+      "step": 19103
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.3035714285714285e-05,
+      "loss": 0.0245,
+      "step": 19104
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.3020408163265305e-05,
+      "loss": 0.0648,
+      "step": 19105
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.3005102040816325e-05,
+      "loss": 0.0644,
+      "step": 19106
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.2989795918367346e-05,
+      "loss": 0.0911,
+      "step": 19107
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.2974489795918366e-05,
+      "loss": 0.0473,
+      "step": 19108
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.2959183673469387e-05,
+      "loss": 0.1188,
+      "step": 19109
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.2943877551020407e-05,
+      "loss": 0.0318,
+      "step": 19110
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.2928571428571427e-05,
+      "loss": 0.0604,
+      "step": 19111
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.2913265306122448e-05,
+      "loss": 0.0769,
+      "step": 19112
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.2897959183673468e-05,
+      "loss": 0.075,
+      "step": 19113
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.288265306122449e-05,
+      "loss": 0.0427,
+      "step": 19114
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.286734693877551e-05,
+      "loss": 0.1248,
+      "step": 19115
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.285204081632653e-05,
+      "loss": 0.1156,
+      "step": 19116
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.283673469387755e-05,
+      "loss": 0.0276,
+      "step": 19117
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.2821428571428567e-05,
+      "loss": 0.0763,
+      "step": 19118
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.2806122448979587e-05,
+      "loss": 0.029,
+      "step": 19119
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.2790816326530607e-05,
+      "loss": 0.0609,
+      "step": 19120
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.2775510204081628e-05,
+      "loss": 0.0374,
+      "step": 19121
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2760204081632648e-05,
+      "loss": 0.1959,
+      "step": 19122
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.274489795918367e-05,
+      "loss": 0.111,
+      "step": 19123
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2729591836734692e-05,
+      "loss": 0.0722,
+      "step": 19124
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2714285714285713e-05,
+      "loss": 0.0228,
+      "step": 19125
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2698979591836733e-05,
+      "loss": 0.0207,
+      "step": 19126
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2683673469387753e-05,
+      "loss": 0.0415,
+      "step": 19127
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2668367346938774e-05,
+      "loss": 0.1004,
+      "step": 19128
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2653061224489794e-05,
+      "loss": 0.0294,
+      "step": 19129
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2637755102040815e-05,
+      "loss": 0.1603,
+      "step": 19130
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2622448979591835e-05,
+      "loss": 0.1602,
+      "step": 19131
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2607142857142855e-05,
+      "loss": 0.0316,
+      "step": 19132
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.2591836734693876e-05,
+      "loss": 0.0272,
+      "step": 19133
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.2576530612244896e-05,
+      "loss": 0.0975,
+      "step": 19134
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.2561224489795917e-05,
+      "loss": 0.0784,
+      "step": 19135
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.2545918367346937e-05,
+      "loss": 0.0709,
+      "step": 19136
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.2530612244897957e-05,
+      "loss": 0.1122,
+      "step": 19137
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.2515306122448978e-05,
+      "loss": 0.0183,
+      "step": 19138
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.2499999999999998e-05,
+      "loss": 0.0294,
+      "step": 19139
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.248469387755102e-05,
+      "loss": 0.0999,
+      "step": 19140
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.246938775510204e-05,
+      "loss": 0.1491,
+      "step": 19141
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.245408163265306e-05,
+      "loss": 0.0385,
+      "step": 19142
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.243877551020408e-05,
+      "loss": 0.0606,
+      "step": 19143
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.24234693877551e-05,
+      "loss": 0.0324,
+      "step": 19144
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.2408163265306117e-05,
+      "loss": 0.0609,
+      "step": 19145
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.2392857142857144e-05,
+      "loss": 0.0652,
+      "step": 19146
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.2377551020408164e-05,
+      "loss": 0.1289,
+      "step": 19147
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.2362244897959185e-05,
+      "loss": 0.0608,
+      "step": 19148
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.2346938775510205e-05,
+      "loss": 0.1045,
+      "step": 19149
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.2331632653061222e-05,
+      "loss": 0.0649,
+      "step": 19150
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.2316326530612243e-05,
+      "loss": 0.0319,
+      "step": 19151
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.2301020408163263e-05,
+      "loss": 0.0327,
+      "step": 19152
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2285714285714283e-05,
+      "loss": 0.1178,
+      "step": 19153
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2270408163265304e-05,
+      "loss": 0.0266,
+      "step": 19154
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2255102040816324e-05,
+      "loss": 0.0675,
+      "step": 19155
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2239795918367345e-05,
+      "loss": 0.0177,
+      "step": 19156
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2224489795918365e-05,
+      "loss": 0.045,
+      "step": 19157
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2209183673469385e-05,
+      "loss": 0.0631,
+      "step": 19158
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2193877551020406e-05,
+      "loss": 0.0483,
+      "step": 19159
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2178571428571426e-05,
+      "loss": 0.0296,
+      "step": 19160
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2163265306122446e-05,
+      "loss": 0.0701,
+      "step": 19161
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2147959183673467e-05,
+      "loss": 0.0594,
+      "step": 19162
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.2132653061224487e-05,
+      "loss": 0.0157,
+      "step": 19163
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.2117346938775508e-05,
+      "loss": 0.1039,
+      "step": 19164
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.2102040816326528e-05,
+      "loss": 0.0711,
+      "step": 19165
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.208673469387755e-05,
+      "loss": 0.076,
+      "step": 19166
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.207142857142857e-05,
+      "loss": 0.0664,
+      "step": 19167
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.205612244897959e-05,
+      "loss": 0.0282,
+      "step": 19168
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.2040816326530613e-05,
+      "loss": 0.0787,
+      "step": 19169
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.2025510204081633e-05,
+      "loss": 0.0205,
+      "step": 19170
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.2010204081632654e-05,
+      "loss": 0.0533,
+      "step": 19171
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.1994897959183674e-05,
+      "loss": 0.1877,
+      "step": 19172
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.1979591836734694e-05,
+      "loss": 0.0228,
+      "step": 19173
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1964285714285715e-05,
+      "loss": 0.1412,
+      "step": 19174
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1948979591836735e-05,
+      "loss": 0.0408,
+      "step": 19175
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1933673469387756e-05,
+      "loss": 0.091,
+      "step": 19176
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1918367346938773e-05,
+      "loss": 0.0535,
+      "step": 19177
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1903061224489793e-05,
+      "loss": 0.0312,
+      "step": 19178
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1887755102040813e-05,
+      "loss": 0.1123,
+      "step": 19179
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1872448979591834e-05,
+      "loss": 0.0651,
+      "step": 19180
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1857142857142854e-05,
+      "loss": 0.0843,
+      "step": 19181
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1841836734693874e-05,
+      "loss": 0.0851,
+      "step": 19182
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1826530612244895e-05,
+      "loss": 0.0276,
+      "step": 19183
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1811224489795915e-05,
+      "loss": 0.074,
+      "step": 19184
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1795918367346936e-05,
+      "loss": 0.1039,
+      "step": 19185
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1780612244897956e-05,
+      "loss": 0.1003,
+      "step": 19186
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1765306122448976e-05,
+      "loss": 0.0722,
+      "step": 19187
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1749999999999997e-05,
+      "loss": 0.0619,
+      "step": 19188
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1734693877551017e-05,
+      "loss": 0.0233,
+      "step": 19189
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1719387755102038e-05,
+      "loss": 0.0323,
+      "step": 19190
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1704081632653058e-05,
+      "loss": 0.188,
+      "step": 19191
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.168877551020408e-05,
+      "loss": 0.2148,
+      "step": 19192
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1673469387755102e-05,
+      "loss": 0.0654,
+      "step": 19193
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1658163265306122e-05,
+      "loss": 0.0293,
+      "step": 19194
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1642857142857143e-05,
+      "loss": 0.1259,
+      "step": 19195
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1627551020408163e-05,
+      "loss": 0.1406,
+      "step": 19196
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1612244897959184e-05,
+      "loss": 0.0956,
+      "step": 19197
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1596938775510204e-05,
+      "loss": 0.0281,
+      "step": 19198
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1581632653061224e-05,
+      "loss": 0.0624,
+      "step": 19199
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1566326530612245e-05,
+      "loss": 0.0961,
+      "step": 19200
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1551020408163265e-05,
+      "loss": 0.0361,
+      "step": 19201
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1535714285714285e-05,
+      "loss": 0.0319,
+      "step": 19202
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1520408163265302e-05,
+      "loss": 0.1074,
+      "step": 19203
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1505102040816323e-05,
+      "loss": 0.0697,
+      "step": 19204
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1489795918367343e-05,
+      "loss": 0.1077,
+      "step": 19205
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1474489795918364e-05,
+      "loss": 0.1054,
+      "step": 19206
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1459183673469384e-05,
+      "loss": 0.041,
+      "step": 19207
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1443877551020404e-05,
+      "loss": 0.0696,
+      "step": 19208
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1428571428571425e-05,
+      "loss": 0.1433,
+      "step": 19209
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1413265306122445e-05,
+      "loss": 0.0706,
+      "step": 19210
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1397959183673466e-05,
+      "loss": 0.0794,
+      "step": 19211
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1382653061224486e-05,
+      "loss": 0.0398,
+      "step": 19212
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.1367346938775506e-05,
+      "loss": 0.097,
+      "step": 19213
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.135204081632653e-05,
+      "loss": 0.0224,
+      "step": 19214
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.133673469387755e-05,
+      "loss": 0.1108,
+      "step": 19215
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.132142857142857e-05,
+      "loss": 0.074,
+      "step": 19216
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.130612244897959e-05,
+      "loss": 0.0357,
+      "step": 19217
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.129081632653061e-05,
+      "loss": 0.0209,
+      "step": 19218
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.1275510204081632e-05,
+      "loss": 0.0283,
+      "step": 19219
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.1260204081632652e-05,
+      "loss": 0.0177,
+      "step": 19220
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.1244897959183673e-05,
+      "loss": 0.1012,
+      "step": 19221
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.1229591836734693e-05,
+      "loss": 0.0716,
+      "step": 19222
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.1214285714285713e-05,
+      "loss": 0.0346,
+      "step": 19223
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.1198979591836734e-05,
+      "loss": 0.1099,
+      "step": 19224
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1183673469387754e-05,
+      "loss": 0.0487,
+      "step": 19225
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1168367346938775e-05,
+      "loss": 0.0555,
+      "step": 19226
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1153061224489795e-05,
+      "loss": 0.0718,
+      "step": 19227
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1137755102040815e-05,
+      "loss": 0.1639,
+      "step": 19228
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1122448979591836e-05,
+      "loss": 0.1429,
+      "step": 19229
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1107142857142853e-05,
+      "loss": 0.0881,
+      "step": 19230
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1091836734693873e-05,
+      "loss": 0.0877,
+      "step": 19231
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1076530612244894e-05,
+      "loss": 0.0323,
+      "step": 19232
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1061224489795914e-05,
+      "loss": 0.0774,
+      "step": 19233
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1045918367346934e-05,
+      "loss": 0.0878,
+      "step": 19234
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.1030612244897955e-05,
+      "loss": 0.069,
+      "step": 19235
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.1015306122448975e-05,
+      "loss": 0.0682,
+      "step": 19236
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.1e-05,
+      "loss": 0.028,
+      "step": 19237
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.098469387755102e-05,
+      "loss": 0.0591,
+      "step": 19238
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.096938775510204e-05,
+      "loss": 0.0234,
+      "step": 19239
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.095408163265306e-05,
+      "loss": 0.1545,
+      "step": 19240
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.093877551020408e-05,
+      "loss": 0.0791,
+      "step": 19241
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.09234693877551e-05,
+      "loss": 0.0209,
+      "step": 19242
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.090816326530612e-05,
+      "loss": 0.0177,
+      "step": 19243
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.089285714285714e-05,
+      "loss": 0.1071,
+      "step": 19244
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.0877551020408162e-05,
+      "loss": 0.0498,
+      "step": 19245
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0862244897959182e-05,
+      "loss": 0.0669,
+      "step": 19246
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0846938775510203e-05,
+      "loss": 0.2105,
+      "step": 19247
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0831632653061223e-05,
+      "loss": 0.0658,
+      "step": 19248
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0816326530612243e-05,
+      "loss": 0.0285,
+      "step": 19249
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0801020408163264e-05,
+      "loss": 0.102,
+      "step": 19250
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0785714285714284e-05,
+      "loss": 0.0279,
+      "step": 19251
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0770408163265305e-05,
+      "loss": 0.0578,
+      "step": 19252
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0755102040816325e-05,
+      "loss": 0.0835,
+      "step": 19253
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0739795918367345e-05,
+      "loss": 0.0701,
+      "step": 19254
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0724489795918366e-05,
+      "loss": 0.0309,
+      "step": 19255
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.0709183673469386e-05,
+      "loss": 0.1089,
+      "step": 19256
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.0693877551020403e-05,
+      "loss": 0.0307,
+      "step": 19257
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.0678571428571423e-05,
+      "loss": 0.1007,
+      "step": 19258
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.0663265306122444e-05,
+      "loss": 0.0819,
+      "step": 19259
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.064795918367347e-05,
+      "loss": 0.0669,
+      "step": 19260
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.063265306122449e-05,
+      "loss": 0.0629,
+      "step": 19261
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.061734693877551e-05,
+      "loss": 0.078,
+      "step": 19262
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.060204081632653e-05,
+      "loss": 0.1299,
+      "step": 19263
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.058673469387755e-05,
+      "loss": 0.0677,
+      "step": 19264
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.057142857142857e-05,
+      "loss": 0.0822,
+      "step": 19265
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.055612244897959e-05,
+      "loss": 0.1078,
+      "step": 19266
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.054081632653061e-05,
+      "loss": 0.0303,
+      "step": 19267
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.052551020408163e-05,
+      "loss": 0.026,
+      "step": 19268
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.051020408163265e-05,
+      "loss": 0.0244,
+      "step": 19269
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.049489795918367e-05,
+      "loss": 0.0399,
+      "step": 19270
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.0479591836734692e-05,
+      "loss": 0.095,
+      "step": 19271
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.0464285714285712e-05,
+      "loss": 0.0419,
+      "step": 19272
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.0448979591836733e-05,
+      "loss": 0.0615,
+      "step": 19273
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.0433673469387753e-05,
+      "loss": 0.0659,
+      "step": 19274
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.0418367346938773e-05,
+      "loss": 0.0692,
+      "step": 19275
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.0403061224489794e-05,
+      "loss": 0.0914,
+      "step": 19276
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.0387755102040814e-05,
+      "loss": 0.0718,
+      "step": 19277
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.0372448979591834e-05,
+      "loss": 0.0598,
+      "step": 19278
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.0357142857142855e-05,
+      "loss": 0.1138,
+      "step": 19279
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.0341836734693875e-05,
+      "loss": 0.0658,
+      "step": 19280
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.0326530612244896e-05,
+      "loss": 0.1004,
+      "step": 19281
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.031122448979592e-05,
+      "loss": 0.0183,
+      "step": 19282
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.029591836734694e-05,
+      "loss": 0.0573,
+      "step": 19283
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.028061224489796e-05,
+      "loss": 0.125,
+      "step": 19284
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.026530612244898e-05,
+      "loss": 0.11,
+      "step": 19285
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.025e-05,
+      "loss": 0.0696,
+      "step": 19286
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.023469387755102e-05,
+      "loss": 0.0609,
+      "step": 19287
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.0219387755102042e-05,
+      "loss": 0.1419,
+      "step": 19288
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.020408163265306e-05,
+      "loss": 0.1177,
+      "step": 19289
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.018877551020408e-05,
+      "loss": 0.0572,
+      "step": 19290
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.01734693877551e-05,
+      "loss": 0.107,
+      "step": 19291
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.015816326530612e-05,
+      "loss": 0.0223,
+      "step": 19292
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.014285714285714e-05,
+      "loss": 0.1139,
+      "step": 19293
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.012755102040816e-05,
+      "loss": 0.062,
+      "step": 19294
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.011224489795918e-05,
+      "loss": 0.0734,
+      "step": 19295
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.00969387755102e-05,
+      "loss": 0.1438,
+      "step": 19296
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.0081632653061222e-05,
+      "loss": 0.0888,
+      "step": 19297
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 2.0066326530612242e-05,
+      "loss": 0.1094,
+      "step": 19298
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 2.0051020408163263e-05,
+      "loss": 0.0315,
+      "step": 19299
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 2.0035714285714283e-05,
+      "loss": 0.1485,
+      "step": 19300
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 2.0020408163265303e-05,
+      "loss": 0.0325,
+      "step": 19301
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 2.0005102040816324e-05,
+      "loss": 0.0306,
+      "step": 19302
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.9989795918367344e-05,
+      "loss": 0.1541,
+      "step": 19303
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.9974489795918364e-05,
+      "loss": 0.0277,
+      "step": 19304
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.9959183673469388e-05,
+      "loss": 0.064,
+      "step": 19305
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.994387755102041e-05,
+      "loss": 0.0966,
+      "step": 19306
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.992857142857143e-05,
+      "loss": 0.0242,
+      "step": 19307
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.991326530612245e-05,
+      "loss": 0.0544,
+      "step": 19308
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.989795918367347e-05,
+      "loss": 0.0913,
+      "step": 19309
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.988265306122449e-05,
+      "loss": 0.0326,
+      "step": 19310
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.986734693877551e-05,
+      "loss": 0.1491,
+      "step": 19311
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.985204081632653e-05,
+      "loss": 0.0681,
+      "step": 19312
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.983673469387755e-05,
+      "loss": 0.0723,
+      "step": 19313
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.982142857142857e-05,
+      "loss": 0.0316,
+      "step": 19314
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.980612244897959e-05,
+      "loss": 0.0661,
+      "step": 19315
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.979081632653061e-05,
+      "loss": 0.1575,
+      "step": 19316
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.977551020408163e-05,
+      "loss": 0.0274,
+      "step": 19317
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.976020408163265e-05,
+      "loss": 0.0297,
+      "step": 19318
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.974489795918367e-05,
+      "loss": 0.0678,
+      "step": 19319
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.972959183673469e-05,
+      "loss": 0.0251,
+      "step": 19320
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.971428571428571e-05,
+      "loss": 0.0505,
+      "step": 19321
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.969897959183673e-05,
+      "loss": 0.0932,
+      "step": 19322
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.968367346938775e-05,
+      "loss": 0.0306,
+      "step": 19323
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.9668367346938772e-05,
+      "loss": 0.0582,
+      "step": 19324
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.9653061224489792e-05,
+      "loss": 0.0204,
+      "step": 19325
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.9637755102040813e-05,
+      "loss": 0.0289,
+      "step": 19326
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.9622448979591833e-05,
+      "loss": 0.0354,
+      "step": 19327
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.9607142857142857e-05,
+      "loss": 0.1404,
+      "step": 19328
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.9591836734693877e-05,
+      "loss": 0.0758,
+      "step": 19329
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.9576530612244898e-05,
+      "loss": 0.0331,
+      "step": 19330
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.9561224489795918e-05,
+      "loss": 0.0268,
+      "step": 19331
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.954591836734694e-05,
+      "loss": 0.1088,
+      "step": 19332
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.953061224489796e-05,
+      "loss": 0.1302,
+      "step": 19333
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.951530612244898e-05,
+      "loss": 0.0788,
+      "step": 19334
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.95e-05,
+      "loss": 0.067,
+      "step": 19335
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.948469387755102e-05,
+      "loss": 0.067,
+      "step": 19336
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.946938775510204e-05,
+      "loss": 0.1118,
+      "step": 19337
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.945408163265306e-05,
+      "loss": 0.1003,
+      "step": 19338
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.943877551020408e-05,
+      "loss": 0.0263,
+      "step": 19339
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.94234693877551e-05,
+      "loss": 0.0991,
+      "step": 19340
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.9408163265306122e-05,
+      "loss": 0.0718,
+      "step": 19341
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.939285714285714e-05,
+      "loss": 0.0662,
+      "step": 19342
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.937755102040816e-05,
+      "loss": 0.0283,
+      "step": 19343
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.936224489795918e-05,
+      "loss": 0.0781,
+      "step": 19344
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.93469387755102e-05,
+      "loss": 0.068,
+      "step": 19345
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.933163265306122e-05,
+      "loss": 0.0748,
+      "step": 19346
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.931632653061224e-05,
+      "loss": 0.1118,
+      "step": 19347
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.930102040816326e-05,
+      "loss": 0.122,
+      "step": 19348
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.928571428571428e-05,
+      "loss": 0.0601,
+      "step": 19349
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.9270408163265305e-05,
+      "loss": 0.0855,
+      "step": 19350
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.9255102040816326e-05,
+      "loss": 0.0364,
+      "step": 19351
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.9239795918367346e-05,
+      "loss": 0.0239,
+      "step": 19352
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.9224489795918367e-05,
+      "loss": 0.1173,
+      "step": 19353
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.9209183673469387e-05,
+      "loss": 0.1071,
+      "step": 19354
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.9193877551020407e-05,
+      "loss": 0.0246,
+      "step": 19355
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.9178571428571428e-05,
+      "loss": 0.0583,
+      "step": 19356
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.9163265306122448e-05,
+      "loss": 0.1206,
+      "step": 19357
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.914795918367347e-05,
+      "loss": 0.0186,
+      "step": 19358
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.913265306122449e-05,
+      "loss": 0.1227,
+      "step": 19359
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.911734693877551e-05,
+      "loss": 0.0346,
+      "step": 19360
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.910204081632653e-05,
+      "loss": 0.0229,
+      "step": 19361
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.908673469387755e-05,
+      "loss": 0.0246,
+      "step": 19362
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.907142857142857e-05,
+      "loss": 0.0341,
+      "step": 19363
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.905612244897959e-05,
+      "loss": 0.0687,
+      "step": 19364
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.904081632653061e-05,
+      "loss": 0.1018,
+      "step": 19365
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.902551020408163e-05,
+      "loss": 0.2785,
+      "step": 19366
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.9010204081632652e-05,
+      "loss": 0.1056,
+      "step": 19367
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.8994897959183672e-05,
+      "loss": 0.0319,
+      "step": 19368
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.897959183673469e-05,
+      "loss": 0.0218,
+      "step": 19369
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.896428571428571e-05,
+      "loss": 0.0261,
+      "step": 19370
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.894897959183673e-05,
+      "loss": 0.0281,
+      "step": 19371
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.893367346938775e-05,
+      "loss": 0.1276,
+      "step": 19372
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8918367346938778e-05,
+      "loss": 0.0691,
+      "step": 19373
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8903061224489795e-05,
+      "loss": 0.0974,
+      "step": 19374
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8887755102040815e-05,
+      "loss": 0.0812,
+      "step": 19375
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8872448979591835e-05,
+      "loss": 0.0259,
+      "step": 19376
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8857142857142856e-05,
+      "loss": 0.0662,
+      "step": 19377
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8841836734693876e-05,
+      "loss": 0.1511,
+      "step": 19378
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8826530612244896e-05,
+      "loss": 0.0631,
+      "step": 19379
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8811224489795917e-05,
+      "loss": 0.0205,
+      "step": 19380
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8795918367346937e-05,
+      "loss": 0.062,
+      "step": 19381
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8780612244897958e-05,
+      "loss": 0.0256,
+      "step": 19382
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8765306122448978e-05,
+      "loss": 0.1244,
+      "step": 19383
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.875e-05,
+      "loss": 0.2041,
+      "step": 19384
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.873469387755102e-05,
+      "loss": 0.0278,
+      "step": 19385
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.871938775510204e-05,
+      "loss": 0.1162,
+      "step": 19386
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.870408163265306e-05,
+      "loss": 0.0661,
+      "step": 19387
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.868877551020408e-05,
+      "loss": 0.021,
+      "step": 19388
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.86734693877551e-05,
+      "loss": 0.0217,
+      "step": 19389
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.865816326530612e-05,
+      "loss": 0.0726,
+      "step": 19390
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.864285714285714e-05,
+      "loss": 0.0857,
+      "step": 19391
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.862755102040816e-05,
+      "loss": 0.0701,
+      "step": 19392
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8612244897959182e-05,
+      "loss": 0.0657,
+      "step": 19393
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8596938775510202e-05,
+      "loss": 0.0246,
+      "step": 19394
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8581632653061223e-05,
+      "loss": 0.0173,
+      "step": 19395
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8566326530612243e-05,
+      "loss": 0.0871,
+      "step": 19396
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8551020408163263e-05,
+      "loss": 0.1134,
+      "step": 19397
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8535714285714284e-05,
+      "loss": 0.1005,
+      "step": 19398
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8520408163265304e-05,
+      "loss": 0.0541,
+      "step": 19399
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8505102040816324e-05,
+      "loss": 0.0238,
+      "step": 19400
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8489795918367345e-05,
+      "loss": 0.0725,
+      "step": 19401
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8474489795918365e-05,
+      "loss": 0.0876,
+      "step": 19402
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8459183673469386e-05,
+      "loss": 0.0278,
+      "step": 19403
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8443877551020406e-05,
+      "loss": 0.0265,
+      "step": 19404
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8428571428571426e-05,
+      "loss": 0.1099,
+      "step": 19405
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8413265306122447e-05,
+      "loss": 0.0991,
+      "step": 19406
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8397959183673467e-05,
+      "loss": 0.0637,
+      "step": 19407
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8382653061224488e-05,
+      "loss": 0.103,
+      "step": 19408
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8367346938775508e-05,
+      "loss": 0.0759,
+      "step": 19409
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8352040816326528e-05,
+      "loss": 0.035,
+      "step": 19410
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.833673469387755e-05,
+      "loss": 0.07,
+      "step": 19411
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.832142857142857e-05,
+      "loss": 0.0256,
+      "step": 19412
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.8306122448979593e-05,
+      "loss": 0.03,
+      "step": 19413
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.829081632653061e-05,
+      "loss": 0.0271,
+      "step": 19414
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.827551020408163e-05,
+      "loss": 0.1062,
+      "step": 19415
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.826020408163265e-05,
+      "loss": 0.1247,
+      "step": 19416
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.824489795918367e-05,
+      "loss": 0.0619,
+      "step": 19417
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.822959183673469e-05,
+      "loss": 0.0296,
+      "step": 19418
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.8214285714285712e-05,
+      "loss": 0.1483,
+      "step": 19419
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.8198979591836732e-05,
+      "loss": 0.027,
+      "step": 19420
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8183673469387752e-05,
+      "loss": 0.0993,
+      "step": 19421
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8168367346938773e-05,
+      "loss": 0.142,
+      "step": 19422
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8153061224489793e-05,
+      "loss": 0.0265,
+      "step": 19423
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8137755102040817e-05,
+      "loss": 0.0483,
+      "step": 19424
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8122448979591837e-05,
+      "loss": 0.0331,
+      "step": 19425
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8107142857142858e-05,
+      "loss": 0.0572,
+      "step": 19426
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8091836734693875e-05,
+      "loss": 0.0848,
+      "step": 19427
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8076530612244895e-05,
+      "loss": 0.0859,
+      "step": 19428
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8061224489795916e-05,
+      "loss": 0.1397,
+      "step": 19429
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8045918367346936e-05,
+      "loss": 0.075,
+      "step": 19430
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.8030612244897956e-05,
+      "loss": 0.1875,
+      "step": 19431
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.8015306122448977e-05,
+      "loss": 0.0998,
+      "step": 19432
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.7999999999999997e-05,
+      "loss": 0.0578,
+      "step": 19433
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.7984693877551017e-05,
+      "loss": 0.1165,
+      "step": 19434
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.7969387755102038e-05,
+      "loss": 0.0613,
+      "step": 19435
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.795408163265306e-05,
+      "loss": 0.0657,
+      "step": 19436
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.7938775510204082e-05,
+      "loss": 0.0309,
+      "step": 19437
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.7923469387755102e-05,
+      "loss": 0.0865,
+      "step": 19438
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.7908163265306123e-05,
+      "loss": 0.0273,
+      "step": 19439
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.789285714285714e-05,
+      "loss": 0.1487,
+      "step": 19440
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.787755102040816e-05,
+      "loss": 0.1094,
+      "step": 19441
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.786224489795918e-05,
+      "loss": 0.0583,
+      "step": 19442
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.78469387755102e-05,
+      "loss": 0.03,
+      "step": 19443
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.783163265306122e-05,
+      "loss": 0.0331,
+      "step": 19444
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.781632653061224e-05,
+      "loss": 0.0325,
+      "step": 19445
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.7801020408163262e-05,
+      "loss": 0.0193,
+      "step": 19446
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.7785714285714286e-05,
+      "loss": 0.1956,
+      "step": 19447
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.7770408163265306e-05,
+      "loss": 0.0214,
+      "step": 19448
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.7755102040816327e-05,
+      "loss": 0.0249,
+      "step": 19449
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.7739795918367347e-05,
+      "loss": 0.0313,
+      "step": 19450
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.7724489795918367e-05,
+      "loss": 0.0325,
+      "step": 19451
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.7709183673469388e-05,
+      "loss": 0.0609,
+      "step": 19452
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.7693877551020408e-05,
+      "loss": 0.1221,
+      "step": 19453
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.7678571428571425e-05,
+      "loss": 0.0695,
+      "step": 19454
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.7663265306122445e-05,
+      "loss": 0.072,
+      "step": 19455
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.7647959183673466e-05,
+      "loss": 0.0718,
+      "step": 19456
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.7632653061224486e-05,
+      "loss": 0.0408,
+      "step": 19457
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.761734693877551e-05,
+      "loss": 0.0504,
+      "step": 19458
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.760204081632653e-05,
+      "loss": 0.1077,
+      "step": 19459
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.758673469387755e-05,
+      "loss": 0.1033,
+      "step": 19460
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.757142857142857e-05,
+      "loss": 0.0672,
+      "step": 19461
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.755612244897959e-05,
+      "loss": 0.0978,
+      "step": 19462
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.7540816326530612e-05,
+      "loss": 0.0234,
+      "step": 19463
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.7525510204081632e-05,
+      "loss": 0.0244,
+      "step": 19464
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.7510204081632653e-05,
+      "loss": 0.078,
+      "step": 19465
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.7494897959183673e-05,
+      "loss": 0.0808,
+      "step": 19466
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.747959183673469e-05,
+      "loss": 0.0797,
+      "step": 19467
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.746428571428571e-05,
+      "loss": 0.0658,
+      "step": 19468
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.744897959183673e-05,
+      "loss": 0.115,
+      "step": 19469
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.7433673469387755e-05,
+      "loss": 0.0567,
+      "step": 19470
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.7418367346938775e-05,
+      "loss": 0.0443,
+      "step": 19471
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.7403061224489795e-05,
+      "loss": 0.1701,
+      "step": 19472
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.7387755102040816e-05,
+      "loss": 0.0621,
+      "step": 19473
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.7372448979591836e-05,
+      "loss": 0.0241,
+      "step": 19474
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.7357142857142856e-05,
+      "loss": 0.0296,
+      "step": 19475
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.7341836734693877e-05,
+      "loss": 0.0332,
+      "step": 19476
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.7326530612244897e-05,
+      "loss": 0.0682,
+      "step": 19477
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.7311224489795918e-05,
+      "loss": 0.1217,
+      "step": 19478
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.7295918367346938e-05,
+      "loss": 0.0626,
+      "step": 19479
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.7280612244897955e-05,
+      "loss": 0.0314,
+      "step": 19480
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.726530612244898e-05,
+      "loss": 0.0263,
+      "step": 19481
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.725e-05,
+      "loss": 0.0281,
+      "step": 19482
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.723469387755102e-05,
+      "loss": 0.03,
+      "step": 19483
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.721938775510204e-05,
+      "loss": 0.1403,
+      "step": 19484
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.720408163265306e-05,
+      "loss": 0.0681,
+      "step": 19485
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.718877551020408e-05,
+      "loss": 0.0226,
+      "step": 19486
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.71734693877551e-05,
+      "loss": 0.157,
+      "step": 19487
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.715816326530612e-05,
+      "loss": 0.0717,
+      "step": 19488
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.7142857142857142e-05,
+      "loss": 0.023,
+      "step": 19489
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.7127551020408162e-05,
+      "loss": 0.062,
+      "step": 19490
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.7112244897959183e-05,
+      "loss": 0.1851,
+      "step": 19491
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.7096938775510203e-05,
+      "loss": 0.0214,
+      "step": 19492
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.7081632653061223e-05,
+      "loss": 0.0725,
+      "step": 19493
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.7066326530612244e-05,
+      "loss": 0.0676,
+      "step": 19494
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.7051020408163264e-05,
+      "loss": 0.0545,
+      "step": 19495
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.7035714285714285e-05,
+      "loss": 0.0211,
+      "step": 19496
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.7020408163265305e-05,
+      "loss": 0.0708,
+      "step": 19497
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.7005102040816325e-05,
+      "loss": 0.0659,
+      "step": 19498
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.6989795918367346e-05,
+      "loss": 0.0254,
+      "step": 19499
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.6974489795918366e-05,
+      "loss": 0.0579,
+      "step": 19500
+    },
+    {
+      "epoch": 18.93,
+      "eval_cer": 0.2408730155276449,
+      "eval_loss": 1.2299330234527588,
+      "eval_runtime": 228.1442,
+      "eval_samples_per_second": 38.633,
+      "eval_steps_per_second": 4.83,
+      "eval_wer": 0.6265091344516582,
+      "step": 19500
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.6959183673469386e-05,
+      "loss": 0.0176,
+      "step": 19501
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.6943877551020407e-05,
+      "loss": 0.0446,
+      "step": 19502
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.6928571428571427e-05,
+      "loss": 0.18,
+      "step": 19503
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.6913265306122448e-05,
+      "loss": 0.0689,
+      "step": 19504
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.6897959183673468e-05,
+      "loss": 0.0678,
+      "step": 19505
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.688265306122449e-05,
+      "loss": 0.1088,
+      "step": 19506
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.686734693877551e-05,
+      "loss": 0.0644,
+      "step": 19507
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.685204081632653e-05,
+      "loss": 0.069,
+      "step": 19508
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.683673469387755e-05,
+      "loss": 0.1199,
+      "step": 19509
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.682142857142857e-05,
+      "loss": 0.0683,
+      "step": 19510
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.680612244897959e-05,
+      "loss": 0.0616,
+      "step": 19511
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.679081632653061e-05,
+      "loss": 0.0179,
+      "step": 19512
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.677551020408163e-05,
+      "loss": 0.0311,
+      "step": 19513
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.676020408163265e-05,
+      "loss": 0.0701,
+      "step": 19514
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6744897959183672e-05,
+      "loss": 0.1035,
+      "step": 19515
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6729591836734692e-05,
+      "loss": 0.0292,
+      "step": 19516
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6714285714285713e-05,
+      "loss": 0.1373,
+      "step": 19517
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6698979591836733e-05,
+      "loss": 0.1045,
+      "step": 19518
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6683673469387753e-05,
+      "loss": 0.0992,
+      "step": 19519
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6668367346938774e-05,
+      "loss": 0.0155,
+      "step": 19520
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6653061224489794e-05,
+      "loss": 0.0601,
+      "step": 19521
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6637755102040814e-05,
+      "loss": 0.1633,
+      "step": 19522
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.6622448979591835e-05,
+      "loss": 0.0647,
+      "step": 19523
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6607142857142855e-05,
+      "loss": 0.0181,
+      "step": 19524
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6591836734693876e-05,
+      "loss": 0.0629,
+      "step": 19525
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6576530612244896e-05,
+      "loss": 0.1044,
+      "step": 19526
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6561224489795916e-05,
+      "loss": 0.0428,
+      "step": 19527
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6545918367346937e-05,
+      "loss": 0.0877,
+      "step": 19528
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6530612244897957e-05,
+      "loss": 0.0266,
+      "step": 19529
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6515306122448977e-05,
+      "loss": 0.0222,
+      "step": 19530
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6499999999999998e-05,
+      "loss": 0.097,
+      "step": 19531
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.6484693877551018e-05,
+      "loss": 0.1497,
+      "step": 19532
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.646938775510204e-05,
+      "loss": 0.0235,
+      "step": 19533
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.645408163265306e-05,
+      "loss": 0.0877,
+      "step": 19534
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.643877551020408e-05,
+      "loss": 0.0302,
+      "step": 19535
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.64234693877551e-05,
+      "loss": 0.2129,
+      "step": 19536
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.640816326530612e-05,
+      "loss": 0.0274,
+      "step": 19537
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.6392857142857144e-05,
+      "loss": 0.1045,
+      "step": 19538
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.637755102040816e-05,
+      "loss": 0.0821,
+      "step": 19539
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.636224489795918e-05,
+      "loss": 0.0906,
+      "step": 19540
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.6346938775510202e-05,
+      "loss": 0.1123,
+      "step": 19541
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.6331632653061222e-05,
+      "loss": 0.1101,
+      "step": 19542
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.6316326530612242e-05,
+      "loss": 0.0292,
+      "step": 19543
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.6301020408163263e-05,
+      "loss": 0.0266,
+      "step": 19544
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.6285714285714283e-05,
+      "loss": 0.0392,
+      "step": 19545
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.6270408163265304e-05,
+      "loss": 0.0453,
+      "step": 19546
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.6255102040816324e-05,
+      "loss": 0.0255,
+      "step": 19547
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.6239795918367344e-05,
+      "loss": 0.0332,
+      "step": 19548
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.6224489795918368e-05,
+      "loss": 0.0688,
+      "step": 19549
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.620918367346939e-05,
+      "loss": 0.0666,
+      "step": 19550
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.619387755102041e-05,
+      "loss": 0.0634,
+      "step": 19551
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.6178571428571426e-05,
+      "loss": 0.0808,
+      "step": 19552
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.6163265306122446e-05,
+      "loss": 0.0932,
+      "step": 19553
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.6147959183673467e-05,
+      "loss": 0.1456,
+      "step": 19554
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.6132653061224487e-05,
+      "loss": 0.0652,
+      "step": 19555
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.6117346938775507e-05,
+      "loss": 0.0267,
+      "step": 19556
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.6102040816326528e-05,
+      "loss": 0.0674,
+      "step": 19557
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.6086734693877548e-05,
+      "loss": 0.0663,
+      "step": 19558
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.607142857142857e-05,
+      "loss": 0.0866,
+      "step": 19559
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.6056122448979592e-05,
+      "loss": 0.0244,
+      "step": 19560
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.6040816326530613e-05,
+      "loss": 0.1018,
+      "step": 19561
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.6025510204081633e-05,
+      "loss": 0.0631,
+      "step": 19562
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.6010204081632653e-05,
+      "loss": 0.1102,
+      "step": 19563
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.5994897959183674e-05,
+      "loss": 0.0234,
+      "step": 19564
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5979591836734694e-05,
+      "loss": 0.0643,
+      "step": 19565
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.596428571428571e-05,
+      "loss": 0.1843,
+      "step": 19566
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.594897959183673e-05,
+      "loss": 0.0261,
+      "step": 19567
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5933673469387752e-05,
+      "loss": 0.1076,
+      "step": 19568
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5918367346938772e-05,
+      "loss": 0.0761,
+      "step": 19569
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5903061224489793e-05,
+      "loss": 0.0596,
+      "step": 19570
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5887755102040813e-05,
+      "loss": 0.0621,
+      "step": 19571
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5872448979591837e-05,
+      "loss": 0.1004,
+      "step": 19572
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5857142857142857e-05,
+      "loss": 0.026,
+      "step": 19573
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5841836734693878e-05,
+      "loss": 0.0557,
+      "step": 19574
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.5826530612244898e-05,
+      "loss": 0.0527,
+      "step": 19575
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.581122448979592e-05,
+      "loss": 0.0626,
+      "step": 19576
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.579591836734694e-05,
+      "loss": 0.0545,
+      "step": 19577
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.578061224489796e-05,
+      "loss": 0.1121,
+      "step": 19578
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.5765306122448976e-05,
+      "loss": 0.0728,
+      "step": 19579
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.5749999999999997e-05,
+      "loss": 0.0691,
+      "step": 19580
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.5734693877551017e-05,
+      "loss": 0.1452,
+      "step": 19581
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.5719387755102037e-05,
+      "loss": 0.0327,
+      "step": 19582
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.570408163265306e-05,
+      "loss": 0.0781,
+      "step": 19583
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.568877551020408e-05,
+      "loss": 0.0648,
+      "step": 19584
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.5673469387755102e-05,
+      "loss": 0.0216,
+      "step": 19585
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5658163265306122e-05,
+      "loss": 0.0728,
+      "step": 19586
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5642857142857143e-05,
+      "loss": 0.1045,
+      "step": 19587
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5627551020408163e-05,
+      "loss": 0.0987,
+      "step": 19588
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5612244897959183e-05,
+      "loss": 0.0245,
+      "step": 19589
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5596938775510204e-05,
+      "loss": 0.0674,
+      "step": 19590
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5581632653061224e-05,
+      "loss": 0.1499,
+      "step": 19591
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.556632653061224e-05,
+      "loss": 0.0229,
+      "step": 19592
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.555102040816326e-05,
+      "loss": 0.0207,
+      "step": 19593
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5535714285714285e-05,
+      "loss": 0.0232,
+      "step": 19594
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5520408163265306e-05,
+      "loss": 0.0577,
+      "step": 19595
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.5505102040816326e-05,
+      "loss": 0.1941,
+      "step": 19596
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.5489795918367346e-05,
+      "loss": 0.1057,
+      "step": 19597
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.5474489795918367e-05,
+      "loss": 0.0262,
+      "step": 19598
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.5459183673469387e-05,
+      "loss": 0.0332,
+      "step": 19599
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.5443877551020408e-05,
+      "loss": 0.0707,
+      "step": 19600
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.5428571428571428e-05,
+      "loss": 0.1307,
+      "step": 19601
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.541326530612245e-05,
+      "loss": 0.1065,
+      "step": 19602
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.539795918367347e-05,
+      "loss": 0.0622,
+      "step": 19603
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.538265306122449e-05,
+      "loss": 0.0633,
+      "step": 19604
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.536734693877551e-05,
+      "loss": 0.0387,
+      "step": 19605
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.535204081632653e-05,
+      "loss": 0.0359,
+      "step": 19606
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.533673469387755e-05,
+      "loss": 0.0353,
+      "step": 19607
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.532142857142857e-05,
+      "loss": 0.1428,
+      "step": 19608
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.530612244897959e-05,
+      "loss": 0.0668,
+      "step": 19609
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.529081632653061e-05,
+      "loss": 0.0986,
+      "step": 19610
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.5275510204081632e-05,
+      "loss": 0.0248,
+      "step": 19611
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.5260204081632652e-05,
+      "loss": 0.0553,
+      "step": 19612
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.5244897959183673e-05,
+      "loss": 0.045,
+      "step": 19613
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.5229591836734691e-05,
+      "loss": 0.1384,
+      "step": 19614
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.5214285714285712e-05,
+      "loss": 0.064,
+      "step": 19615
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.5198979591836732e-05,
+      "loss": 0.0266,
+      "step": 19616
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5183673469387754e-05,
+      "loss": 0.0238,
+      "step": 19617
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5168367346938774e-05,
+      "loss": 0.0615,
+      "step": 19618
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5153061224489795e-05,
+      "loss": 0.0288,
+      "step": 19619
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5137755102040815e-05,
+      "loss": 0.1146,
+      "step": 19620
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5122448979591836e-05,
+      "loss": 0.0215,
+      "step": 19621
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5107142857142856e-05,
+      "loss": 0.029,
+      "step": 19622
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5091836734693876e-05,
+      "loss": 0.0339,
+      "step": 19623
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5076530612244897e-05,
+      "loss": 0.0279,
+      "step": 19624
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5061224489795917e-05,
+      "loss": 0.0247,
+      "step": 19625
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.5045918367346938e-05,
+      "loss": 0.0272,
+      "step": 19626
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.5030612244897956e-05,
+      "loss": 0.2176,
+      "step": 19627
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.501530612244898e-05,
+      "loss": 0.0736,
+      "step": 19628
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 0.0238,
+      "step": 19629
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.4984693877551019e-05,
+      "loss": 0.0245,
+      "step": 19630
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.496938775510204e-05,
+      "loss": 0.0267,
+      "step": 19631
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.495408163265306e-05,
+      "loss": 0.1022,
+      "step": 19632
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.493877551020408e-05,
+      "loss": 0.0896,
+      "step": 19633
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.49234693877551e-05,
+      "loss": 0.1391,
+      "step": 19634
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.4908163265306121e-05,
+      "loss": 0.0635,
+      "step": 19635
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.4892857142857141e-05,
+      "loss": 0.0208,
+      "step": 19636
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4877551020408162e-05,
+      "loss": 0.0283,
+      "step": 19637
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4862244897959182e-05,
+      "loss": 0.065,
+      "step": 19638
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4846938775510203e-05,
+      "loss": 0.0952,
+      "step": 19639
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4831632653061225e-05,
+      "loss": 0.1505,
+      "step": 19640
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4816326530612245e-05,
+      "loss": 0.1108,
+      "step": 19641
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4801020408163265e-05,
+      "loss": 0.065,
+      "step": 19642
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4785714285714284e-05,
+      "loss": 0.0275,
+      "step": 19643
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4770408163265304e-05,
+      "loss": 0.0628,
+      "step": 19644
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4755102040816325e-05,
+      "loss": 0.0699,
+      "step": 19645
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4739795918367345e-05,
+      "loss": 0.0717,
+      "step": 19646
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4724489795918366e-05,
+      "loss": 0.0594,
+      "step": 19647
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4709183673469386e-05,
+      "loss": 0.0657,
+      "step": 19648
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4693877551020406e-05,
+      "loss": 0.0747,
+      "step": 19649
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4678571428571427e-05,
+      "loss": 0.021,
+      "step": 19650
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4663265306122449e-05,
+      "loss": 0.0489,
+      "step": 19651
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.464795918367347e-05,
+      "loss": 0.0627,
+      "step": 19652
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.463265306122449e-05,
+      "loss": 0.0616,
+      "step": 19653
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.461734693877551e-05,
+      "loss": 0.0653,
+      "step": 19654
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.460204081632653e-05,
+      "loss": 0.0788,
+      "step": 19655
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4586734693877549e-05,
+      "loss": 0.0275,
+      "step": 19656
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.457142857142857e-05,
+      "loss": 0.1264,
+      "step": 19657
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.455612244897959e-05,
+      "loss": 0.2115,
+      "step": 19658
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.454081632653061e-05,
+      "loss": 0.0733,
+      "step": 19659
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.452551020408163e-05,
+      "loss": 0.067,
+      "step": 19660
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4510204081632651e-05,
+      "loss": 0.0964,
+      "step": 19661
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4494897959183673e-05,
+      "loss": 0.028,
+      "step": 19662
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4479591836734693e-05,
+      "loss": 0.0717,
+      "step": 19663
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4464285714285714e-05,
+      "loss": 0.0861,
+      "step": 19664
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4448979591836734e-05,
+      "loss": 0.0339,
+      "step": 19665
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4433673469387755e-05,
+      "loss": 0.0934,
+      "step": 19666
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4418367346938775e-05,
+      "loss": 0.0617,
+      "step": 19667
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4403061224489795e-05,
+      "loss": 0.0244,
+      "step": 19668
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4387755102040814e-05,
+      "loss": 0.0291,
+      "step": 19669
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4372448979591834e-05,
+      "loss": 0.1172,
+      "step": 19670
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4357142857142855e-05,
+      "loss": 0.0255,
+      "step": 19671
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4341836734693875e-05,
+      "loss": 0.0193,
+      "step": 19672
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4326530612244895e-05,
+      "loss": 0.0637,
+      "step": 19673
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4311224489795918e-05,
+      "loss": 0.025,
+      "step": 19674
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4295918367346938e-05,
+      "loss": 0.0314,
+      "step": 19675
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4280612244897958e-05,
+      "loss": 0.0344,
+      "step": 19676
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4265306122448979e-05,
+      "loss": 0.0398,
+      "step": 19677
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4249999999999999e-05,
+      "loss": 0.1126,
+      "step": 19678
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.423469387755102e-05,
+      "loss": 0.1021,
+      "step": 19679
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.421938775510204e-05,
+      "loss": 0.1484,
+      "step": 19680
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.420408163265306e-05,
+      "loss": 0.135,
+      "step": 19681
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.418877551020408e-05,
+      "loss": 0.0893,
+      "step": 19682
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.41734693877551e-05,
+      "loss": 0.1571,
+      "step": 19683
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.415816326530612e-05,
+      "loss": 0.0342,
+      "step": 19684
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4142857142857142e-05,
+      "loss": 0.0514,
+      "step": 19685
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4127551020408162e-05,
+      "loss": 0.0592,
+      "step": 19686
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4112244897959183e-05,
+      "loss": 0.0623,
+      "step": 19687
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4096938775510203e-05,
+      "loss": 0.0282,
+      "step": 19688
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4081632653061223e-05,
+      "loss": 0.1735,
+      "step": 19689
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4066326530612244e-05,
+      "loss": 0.0975,
+      "step": 19690
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4051020408163264e-05,
+      "loss": 0.1802,
+      "step": 19691
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4035714285714284e-05,
+      "loss": 0.0625,
+      "step": 19692
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4020408163265305e-05,
+      "loss": 0.0249,
+      "step": 19693
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4005102040816325e-05,
+      "loss": 0.0207,
+      "step": 19694
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.3989795918367346e-05,
+      "loss": 0.156,
+      "step": 19695
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.3974489795918368e-05,
+      "loss": 0.0708,
+      "step": 19696
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.3959183673469388e-05,
+      "loss": 0.06,
+      "step": 19697
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.3943877551020408e-05,
+      "loss": 0.0676,
+      "step": 19698
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3928571428571427e-05,
+      "loss": 0.0289,
+      "step": 19699
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3913265306122447e-05,
+      "loss": 0.0711,
+      "step": 19700
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3897959183673468e-05,
+      "loss": 0.0635,
+      "step": 19701
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3882653061224488e-05,
+      "loss": 0.1992,
+      "step": 19702
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3867346938775509e-05,
+      "loss": 0.072,
+      "step": 19703
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3852040816326529e-05,
+      "loss": 0.0249,
+      "step": 19704
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.383673469387755e-05,
+      "loss": 0.0282,
+      "step": 19705
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.382142857142857e-05,
+      "loss": 0.0377,
+      "step": 19706
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.380612244897959e-05,
+      "loss": 0.0323,
+      "step": 19707
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3790816326530612e-05,
+      "loss": 0.0494,
+      "step": 19708
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3775510204081633e-05,
+      "loss": 0.0279,
+      "step": 19709
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3760204081632653e-05,
+      "loss": 0.0678,
+      "step": 19710
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3744897959183673e-05,
+      "loss": 0.0659,
+      "step": 19711
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3729591836734692e-05,
+      "loss": 0.0592,
+      "step": 19712
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3714285714285712e-05,
+      "loss": 0.0453,
+      "step": 19713
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3698979591836733e-05,
+      "loss": 0.1526,
+      "step": 19714
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3683673469387753e-05,
+      "loss": 0.062,
+      "step": 19715
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3668367346938774e-05,
+      "loss": 0.1036,
+      "step": 19716
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3653061224489794e-05,
+      "loss": 0.0277,
+      "step": 19717
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3637755102040814e-05,
+      "loss": 0.0159,
+      "step": 19718
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.3622448979591836e-05,
+      "loss": 0.0215,
+      "step": 19719
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3607142857142857e-05,
+      "loss": 0.0472,
+      "step": 19720
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3591836734693877e-05,
+      "loss": 0.1572,
+      "step": 19721
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3576530612244898e-05,
+      "loss": 0.066,
+      "step": 19722
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3561224489795918e-05,
+      "loss": 0.0543,
+      "step": 19723
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3545918367346938e-05,
+      "loss": 0.1097,
+      "step": 19724
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3530612244897957e-05,
+      "loss": 0.0248,
+      "step": 19725
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3515306122448977e-05,
+      "loss": 0.0626,
+      "step": 19726
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3499999999999998e-05,
+      "loss": 0.1054,
+      "step": 19727
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3484693877551018e-05,
+      "loss": 0.0991,
+      "step": 19728
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3469387755102039e-05,
+      "loss": 0.0549,
+      "step": 19729
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.345408163265306e-05,
+      "loss": 0.155,
+      "step": 19730
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3438775510204081e-05,
+      "loss": 0.0327,
+      "step": 19731
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3423469387755101e-05,
+      "loss": 0.0846,
+      "step": 19732
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3408163265306122e-05,
+      "loss": 0.1074,
+      "step": 19733
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3392857142857142e-05,
+      "loss": 0.0713,
+      "step": 19734
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3377551020408163e-05,
+      "loss": 0.0713,
+      "step": 19735
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3362244897959183e-05,
+      "loss": 0.1409,
+      "step": 19736
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3346938775510203e-05,
+      "loss": 0.0521,
+      "step": 19737
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3331632653061224e-05,
+      "loss": 0.0562,
+      "step": 19738
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3316326530612242e-05,
+      "loss": 0.0437,
+      "step": 19739
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3301020408163263e-05,
+      "loss": 0.0245,
+      "step": 19740
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3285714285714283e-05,
+      "loss": 0.1404,
+      "step": 19741
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3270408163265305e-05,
+      "loss": 0.0558,
+      "step": 19742
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3255102040816326e-05,
+      "loss": 0.0251,
+      "step": 19743
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3239795918367346e-05,
+      "loss": 0.0315,
+      "step": 19744
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3224489795918366e-05,
+      "loss": 0.0563,
+      "step": 19745
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3209183673469387e-05,
+      "loss": 0.1118,
+      "step": 19746
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3193877551020407e-05,
+      "loss": 0.0662,
+      "step": 19747
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3178571428571428e-05,
+      "loss": 0.0568,
+      "step": 19748
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3163265306122448e-05,
+      "loss": 0.1056,
+      "step": 19749
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.3147959183673468e-05,
+      "loss": 0.0981,
+      "step": 19750
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.3132653061224489e-05,
+      "loss": 0.0485,
+      "step": 19751
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.3117346938775507e-05,
+      "loss": 0.1847,
+      "step": 19752
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.3102040816326531e-05,
+      "loss": 0.029,
+      "step": 19753
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.3086734693877551e-05,
+      "loss": 0.0437,
+      "step": 19754
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.307142857142857e-05,
+      "loss": 0.017,
+      "step": 19755
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.305612244897959e-05,
+      "loss": 0.0674,
+      "step": 19756
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.3040816326530611e-05,
+      "loss": 0.0825,
+      "step": 19757
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.3025510204081631e-05,
+      "loss": 0.1732,
+      "step": 19758
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.3025510204081631e-05,
+      "loss": 0.0679,
+      "step": 19759
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.3010204081632652e-05,
+      "loss": 0.0286,
+      "step": 19760
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2994897959183672e-05,
+      "loss": 0.107,
+      "step": 19761
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2979591836734692e-05,
+      "loss": 0.0305,
+      "step": 19762
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2964285714285713e-05,
+      "loss": 0.0305,
+      "step": 19763
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2948979591836733e-05,
+      "loss": 0.1415,
+      "step": 19764
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2933673469387755e-05,
+      "loss": 0.0656,
+      "step": 19765
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2918367346938776e-05,
+      "loss": 0.1413,
+      "step": 19766
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2903061224489796e-05,
+      "loss": 0.0301,
+      "step": 19767
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2887755102040816e-05,
+      "loss": 0.0931,
+      "step": 19768
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2872448979591835e-05,
+      "loss": 0.0628,
+      "step": 19769
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2857142857142856e-05,
+      "loss": 0.1594,
+      "step": 19770
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2841836734693876e-05,
+      "loss": 0.1505,
+      "step": 19771
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2826530612244896e-05,
+      "loss": 0.0226,
+      "step": 19772
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2811224489795917e-05,
+      "loss": 0.1502,
+      "step": 19773
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2795918367346937e-05,
+      "loss": 0.0703,
+      "step": 19774
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2780612244897957e-05,
+      "loss": 0.0562,
+      "step": 19775
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2765306122448978e-05,
+      "loss": 0.0917,
+      "step": 19776
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.275e-05,
+      "loss": 0.095,
+      "step": 19777
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.273469387755102e-05,
+      "loss": 0.0224,
+      "step": 19778
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.271938775510204e-05,
+      "loss": 0.148,
+      "step": 19779
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2704081632653061e-05,
+      "loss": 0.0218,
+      "step": 19780
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2688775510204081e-05,
+      "loss": 0.065,
+      "step": 19781
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.26734693877551e-05,
+      "loss": 0.0209,
+      "step": 19782
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.265816326530612e-05,
+      "loss": 0.1719,
+      "step": 19783
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2642857142857141e-05,
+      "loss": 0.059,
+      "step": 19784
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2627551020408161e-05,
+      "loss": 0.109,
+      "step": 19785
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2612244897959182e-05,
+      "loss": 0.0252,
+      "step": 19786
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2596938775510202e-05,
+      "loss": 0.1129,
+      "step": 19787
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2581632653061224e-05,
+      "loss": 0.0485,
+      "step": 19788
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2566326530612244e-05,
+      "loss": 0.1064,
+      "step": 19789
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2551020408163265e-05,
+      "loss": 0.0662,
+      "step": 19790
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2535714285714285e-05,
+      "loss": 0.0716,
+      "step": 19791
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2520408163265306e-05,
+      "loss": 0.1055,
+      "step": 19792
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2505102040816326e-05,
+      "loss": 0.0765,
+      "step": 19793
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2489795918367346e-05,
+      "loss": 0.022,
+      "step": 19794
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2474489795918367e-05,
+      "loss": 0.1204,
+      "step": 19795
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2459183673469385e-05,
+      "loss": 0.0237,
+      "step": 19796
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2443877551020406e-05,
+      "loss": 0.057,
+      "step": 19797
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2428571428571426e-05,
+      "loss": 0.0835,
+      "step": 19798
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2413265306122447e-05,
+      "loss": 0.1364,
+      "step": 19799
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2397959183673469e-05,
+      "loss": 0.0819,
+      "step": 19800
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2382653061224489e-05,
+      "loss": 0.0398,
+      "step": 19801
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.236734693877551e-05,
+      "loss": 0.127,
+      "step": 19802
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.235204081632653e-05,
+      "loss": 0.1089,
+      "step": 19803
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.233673469387755e-05,
+      "loss": 0.0222,
+      "step": 19804
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.232142857142857e-05,
+      "loss": 0.0531,
+      "step": 19805
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.2306122448979591e-05,
+      "loss": 0.0802,
+      "step": 19806
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.2290816326530611e-05,
+      "loss": 0.0592,
+      "step": 19807
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.2275510204081632e-05,
+      "loss": 0.086,
+      "step": 19808
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.226020408163265e-05,
+      "loss": 0.0622,
+      "step": 19809
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.224489795918367e-05,
+      "loss": 0.0224,
+      "step": 19810
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.2229591836734693e-05,
+      "loss": 0.0185,
+      "step": 19811
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.2214285714285713e-05,
+      "loss": 0.0241,
+      "step": 19812
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2198979591836734e-05,
+      "loss": 0.0886,
+      "step": 19813
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2183673469387754e-05,
+      "loss": 0.0756,
+      "step": 19814
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2168367346938774e-05,
+      "loss": 0.1503,
+      "step": 19815
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2153061224489795e-05,
+      "loss": 0.0641,
+      "step": 19816
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2137755102040815e-05,
+      "loss": 0.0305,
+      "step": 19817
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2122448979591836e-05,
+      "loss": 0.0422,
+      "step": 19818
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2107142857142856e-05,
+      "loss": 0.0258,
+      "step": 19819
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2091836734693876e-05,
+      "loss": 0.1172,
+      "step": 19820
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2076530612244897e-05,
+      "loss": 0.0629,
+      "step": 19821
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2061224489795919e-05,
+      "loss": 0.0667,
+      "step": 19822
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.204591836734694e-05,
+      "loss": 0.0635,
+      "step": 19823
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.203061224489796e-05,
+      "loss": 0.143,
+      "step": 19824
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.2015306122448978e-05,
+      "loss": 0.0393,
+      "step": 19825
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.1999999999999999e-05,
+      "loss": 0.0374,
+      "step": 19826
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.1984693877551019e-05,
+      "loss": 0.1071,
+      "step": 19827
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.196938775510204e-05,
+      "loss": 0.0287,
+      "step": 19828
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.195408163265306e-05,
+      "loss": 0.0269,
+      "step": 19829
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.193877551020408e-05,
+      "loss": 0.0233,
+      "step": 19830
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.19234693877551e-05,
+      "loss": 0.0333,
+      "step": 19831
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.1908163265306121e-05,
+      "loss": 0.0328,
+      "step": 19832
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1892857142857141e-05,
+      "loss": 0.104,
+      "step": 19833
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1877551020408163e-05,
+      "loss": 0.0695,
+      "step": 19834
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1862244897959184e-05,
+      "loss": 0.0667,
+      "step": 19835
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1846938775510204e-05,
+      "loss": 0.0255,
+      "step": 19836
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1831632653061225e-05,
+      "loss": 0.0638,
+      "step": 19837
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1816326530612243e-05,
+      "loss": 0.0449,
+      "step": 19838
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1801020408163264e-05,
+      "loss": 0.104,
+      "step": 19839
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1785714285714284e-05,
+      "loss": 0.0363,
+      "step": 19840
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1770408163265304e-05,
+      "loss": 0.0271,
+      "step": 19841
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1755102040816325e-05,
+      "loss": 0.0219,
+      "step": 19842
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1739795918367345e-05,
+      "loss": 0.0659,
+      "step": 19843
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1724489795918365e-05,
+      "loss": 0.0246,
+      "step": 19844
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1709183673469388e-05,
+      "loss": 0.0628,
+      "step": 19845
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1693877551020408e-05,
+      "loss": 0.1298,
+      "step": 19846
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1678571428571428e-05,
+      "loss": 0.0266,
+      "step": 19847
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1663265306122449e-05,
+      "loss": 0.0206,
+      "step": 19848
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1647959183673469e-05,
+      "loss": 0.0667,
+      "step": 19849
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.163265306122449e-05,
+      "loss": 0.0656,
+      "step": 19850
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.161734693877551e-05,
+      "loss": 0.1198,
+      "step": 19851
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1602040816326529e-05,
+      "loss": 0.07,
+      "step": 19852
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1586734693877549e-05,
+      "loss": 0.0298,
+      "step": 19853
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.157142857142857e-05,
+      "loss": 0.0614,
+      "step": 19854
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.155612244897959e-05,
+      "loss": 0.0959,
+      "step": 19855
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.1540816326530612e-05,
+      "loss": 0.0632,
+      "step": 19856
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.1525510204081632e-05,
+      "loss": 0.0627,
+      "step": 19857
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.1510204081632653e-05,
+      "loss": 0.1175,
+      "step": 19858
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.1494897959183673e-05,
+      "loss": 0.0683,
+      "step": 19859
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.1479591836734693e-05,
+      "loss": 0.063,
+      "step": 19860
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.1464285714285714e-05,
+      "loss": 0.1186,
+      "step": 19861
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.1448979591836734e-05,
+      "loss": 0.0512,
+      "step": 19862
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.1433673469387754e-05,
+      "loss": 0.0272,
+      "step": 19863
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1418367346938775e-05,
+      "loss": 0.1496,
+      "step": 19864
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1403061224489794e-05,
+      "loss": 0.1223,
+      "step": 19865
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1387755102040814e-05,
+      "loss": 0.0236,
+      "step": 19866
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1372448979591834e-05,
+      "loss": 0.0584,
+      "step": 19867
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1357142857142856e-05,
+      "loss": 0.1819,
+      "step": 19868
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1341836734693877e-05,
+      "loss": 0.0313,
+      "step": 19869
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1326530612244897e-05,
+      "loss": 0.125,
+      "step": 19870
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1311224489795917e-05,
+      "loss": 0.1448,
+      "step": 19871
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1295918367346938e-05,
+      "loss": 0.0197,
+      "step": 19872
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1280612244897958e-05,
+      "loss": 0.1193,
+      "step": 19873
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1265306122448979e-05,
+      "loss": 0.0643,
+      "step": 19874
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1249999999999999e-05,
+      "loss": 0.1295,
+      "step": 19875
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.123469387755102e-05,
+      "loss": 0.0926,
+      "step": 19876
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.121938775510204e-05,
+      "loss": 0.0688,
+      "step": 19877
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1204081632653058e-05,
+      "loss": 0.1043,
+      "step": 19878
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1188775510204082e-05,
+      "loss": 0.1397,
+      "step": 19879
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1173469387755103e-05,
+      "loss": 0.0215,
+      "step": 19880
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1158163265306121e-05,
+      "loss": 0.0656,
+      "step": 19881
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1142857142857142e-05,
+      "loss": 0.0374,
+      "step": 19882
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1127551020408162e-05,
+      "loss": 0.1095,
+      "step": 19883
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.1112244897959182e-05,
+      "loss": 0.041,
+      "step": 19884
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.1096938775510203e-05,
+      "loss": 0.0277,
+      "step": 19885
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.1081632653061223e-05,
+      "loss": 0.1067,
+      "step": 19886
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.1066326530612244e-05,
+      "loss": 0.0609,
+      "step": 19887
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.1051020408163264e-05,
+      "loss": 0.0561,
+      "step": 19888
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.1035714285714284e-05,
+      "loss": 0.1416,
+      "step": 19889
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.1020408163265306e-05,
+      "loss": 0.0994,
+      "step": 19890
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.1005102040816327e-05,
+      "loss": 0.0727,
+      "step": 19891
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.0989795918367347e-05,
+      "loss": 0.0296,
+      "step": 19892
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.0974489795918368e-05,
+      "loss": 0.0649,
+      "step": 19893
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.0959183673469386e-05,
+      "loss": 0.029,
+      "step": 19894
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0943877551020407e-05,
+      "loss": 0.0672,
+      "step": 19895
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0928571428571427e-05,
+      "loss": 0.1437,
+      "step": 19896
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0913265306122447e-05,
+      "loss": 0.0379,
+      "step": 19897
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0897959183673468e-05,
+      "loss": 0.0882,
+      "step": 19898
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0882653061224488e-05,
+      "loss": 0.0614,
+      "step": 19899
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0867346938775509e-05,
+      "loss": 0.0444,
+      "step": 19900
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0852040816326529e-05,
+      "loss": 0.0965,
+      "step": 19901
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0836734693877551e-05,
+      "loss": 0.1168,
+      "step": 19902
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0821428571428571e-05,
+      "loss": 0.1035,
+      "step": 19903
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0806122448979592e-05,
+      "loss": 0.0166,
+      "step": 19904
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0790816326530612e-05,
+      "loss": 0.0139,
+      "step": 19905
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0775510204081633e-05,
+      "loss": 0.0651,
+      "step": 19906
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0760204081632651e-05,
+      "loss": 0.0364,
+      "step": 19907
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0744897959183672e-05,
+      "loss": 0.1879,
+      "step": 19908
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0729591836734692e-05,
+      "loss": 0.0282,
+      "step": 19909
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0714285714285712e-05,
+      "loss": 0.0308,
+      "step": 19910
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0698979591836733e-05,
+      "loss": 0.2057,
+      "step": 19911
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0683673469387753e-05,
+      "loss": 0.0354,
+      "step": 19912
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0668367346938775e-05,
+      "loss": 0.0253,
+      "step": 19913
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0653061224489796e-05,
+      "loss": 0.0424,
+      "step": 19914
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.0637755102040816e-05,
+      "loss": 0.0652,
+      "step": 19915
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0622448979591836e-05,
+      "loss": 0.0293,
+      "step": 19916
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0607142857142857e-05,
+      "loss": 0.0643,
+      "step": 19917
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0591836734693877e-05,
+      "loss": 0.1014,
+      "step": 19918
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0576530612244898e-05,
+      "loss": 0.098,
+      "step": 19919
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0561224489795918e-05,
+      "loss": 0.0856,
+      "step": 19920
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0545918367346937e-05,
+      "loss": 0.0197,
+      "step": 19921
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0530612244897957e-05,
+      "loss": 0.1125,
+      "step": 19922
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0515306122448977e-05,
+      "loss": 0.0244,
+      "step": 19923
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.05e-05,
+      "loss": 0.0601,
+      "step": 19924
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.048469387755102e-05,
+      "loss": 0.0533,
+      "step": 19925
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.046938775510204e-05,
+      "loss": 0.0511,
+      "step": 19926
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.045408163265306e-05,
+      "loss": 0.1524,
+      "step": 19927
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0438775510204081e-05,
+      "loss": 0.026,
+      "step": 19928
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0423469387755101e-05,
+      "loss": 0.0736,
+      "step": 19929
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0408163265306122e-05,
+      "loss": 0.0681,
+      "step": 19930
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0392857142857142e-05,
+      "loss": 0.0324,
+      "step": 19931
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0377551020408162e-05,
+      "loss": 0.0275,
+      "step": 19932
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0362244897959183e-05,
+      "loss": 0.1071,
+      "step": 19933
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0346938775510202e-05,
+      "loss": 0.027,
+      "step": 19934
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0331632653061222e-05,
+      "loss": 0.1093,
+      "step": 19935
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0316326530612246e-05,
+      "loss": 0.0999,
+      "step": 19936
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0301020408163264e-05,
+      "loss": 0.0233,
+      "step": 19937
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0285714285714285e-05,
+      "loss": 0.0546,
+      "step": 19938
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0270408163265305e-05,
+      "loss": 0.1125,
+      "step": 19939
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0255102040816326e-05,
+      "loss": 0.0608,
+      "step": 19940
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0239795918367346e-05,
+      "loss": 0.0289,
+      "step": 19941
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0224489795918366e-05,
+      "loss": 0.0635,
+      "step": 19942
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0209183673469387e-05,
+      "loss": 0.026,
+      "step": 19943
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0193877551020407e-05,
+      "loss": 0.0651,
+      "step": 19944
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.0178571428571427e-05,
+      "loss": 0.0762,
+      "step": 19945
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.0163265306122448e-05,
+      "loss": 0.0727,
+      "step": 19946
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.014795918367347e-05,
+      "loss": 0.0244,
+      "step": 19947
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.013265306122449e-05,
+      "loss": 0.066,
+      "step": 19948
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.011734693877551e-05,
+      "loss": 0.0684,
+      "step": 19949
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.010204081632653e-05,
+      "loss": 0.0253,
+      "step": 19950
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.008673469387755e-05,
+      "loss": 0.0286,
+      "step": 19951
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.007142857142857e-05,
+      "loss": 0.0582,
+      "step": 19952
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.005612244897959e-05,
+      "loss": 0.0285,
+      "step": 19953
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.0040816326530611e-05,
+      "loss": 0.0674,
+      "step": 19954
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.0025510204081631e-05,
+      "loss": 0.0786,
+      "step": 19955
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.0010204081632652e-05,
+      "loss": 0.0594,
+      "step": 19956
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.994897959183672e-06,
+      "loss": 0.1403,
+      "step": 19957
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.979591836734694e-06,
+      "loss": 0.1107,
+      "step": 19958
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.964285714285714e-06,
+      "loss": 0.0226,
+      "step": 19959
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.948979591836735e-06,
+      "loss": 0.0226,
+      "step": 19960
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.933673469387755e-06,
+      "loss": 0.0468,
+      "step": 19961
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.918367346938776e-06,
+      "loss": 0.1074,
+      "step": 19962
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.903061224489794e-06,
+      "loss": 0.0589,
+      "step": 19963
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.887755102040815e-06,
+      "loss": 0.0624,
+      "step": 19964
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.872448979591835e-06,
+      "loss": 0.0988,
+      "step": 19965
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 9.857142857142855e-06,
+      "loss": 0.0556,
+      "step": 19966
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.841836734693876e-06,
+      "loss": 0.0278,
+      "step": 19967
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.826530612244896e-06,
+      "loss": 0.0999,
+      "step": 19968
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.811224489795917e-06,
+      "loss": 0.0173,
+      "step": 19969
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.795918367346939e-06,
+      "loss": 0.0283,
+      "step": 19970
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.780612244897959e-06,
+      "loss": 0.0331,
+      "step": 19971
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.76530612244898e-06,
+      "loss": 0.0693,
+      "step": 19972
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.75e-06,
+      "loss": 0.0286,
+      "step": 19973
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.73469387755102e-06,
+      "loss": 0.0328,
+      "step": 19974
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.71938775510204e-06,
+      "loss": 0.0485,
+      "step": 19975
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.704081632653061e-06,
+      "loss": 0.0734,
+      "step": 19976
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.68877551020408e-06,
+      "loss": 0.0813,
+      "step": 19977
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.6734693877551e-06,
+      "loss": 0.107,
+      "step": 19978
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.65816326530612e-06,
+      "loss": 0.0357,
+      "step": 19979
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.64285714285714e-06,
+      "loss": 0.027,
+      "step": 19980
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.627551020408163e-06,
+      "loss": 0.0778,
+      "step": 19981
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.612244897959183e-06,
+      "loss": 0.0956,
+      "step": 19982
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.596938775510204e-06,
+      "loss": 0.0926,
+      "step": 19983
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.581632653061224e-06,
+      "loss": 0.0604,
+      "step": 19984
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.566326530612244e-06,
+      "loss": 0.0636,
+      "step": 19985
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.551020408163265e-06,
+      "loss": 0.0657,
+      "step": 19986
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.535714285714285e-06,
+      "loss": 0.0222,
+      "step": 19987
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.520408163265306e-06,
+      "loss": 0.0646,
+      "step": 19988
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.505102040816326e-06,
+      "loss": 0.1446,
+      "step": 19989
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.489795918367345e-06,
+      "loss": 0.0796,
+      "step": 19990
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.474489795918365e-06,
+      "loss": 0.0589,
+      "step": 19991
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.459183673469389e-06,
+      "loss": 0.2184,
+      "step": 19992
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.443877551020407e-06,
+      "loss": 0.0665,
+      "step": 19993
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.428571428571428e-06,
+      "loss": 0.0404,
+      "step": 19994
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.413265306122448e-06,
+      "loss": 0.0435,
+      "step": 19995
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.397959183673469e-06,
+      "loss": 0.0706,
+      "step": 19996
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.382653061224489e-06,
+      "loss": 0.0642,
+      "step": 19997
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.36734693877551e-06,
+      "loss": 0.0326,
+      "step": 19998
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.35204081632653e-06,
+      "loss": 0.0234,
+      "step": 19999
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.33673469387755e-06,
+      "loss": 0.0628,
+      "step": 20000
+    },
+    {
+      "epoch": 19.42,
+      "eval_cer": 0.2394700356750274,
+      "eval_loss": 1.2582364082336426,
+      "eval_runtime": 228.3986,
+      "eval_samples_per_second": 38.59,
+      "eval_steps_per_second": 4.825,
+      "eval_wer": 0.6276516275033232,
+      "step": 20000
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.32142857142857e-06,
+      "loss": 0.1024,
+      "step": 20001
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.306122448979591e-06,
+      "loss": 0.137,
+      "step": 20002
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.290816326530611e-06,
+      "loss": 0.034,
+      "step": 20003
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.275510204081632e-06,
+      "loss": 0.063,
+      "step": 20004
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.260204081632652e-06,
+      "loss": 0.051,
+      "step": 20005
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.244897959183672e-06,
+      "loss": 0.0296,
+      "step": 20006
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.229591836734693e-06,
+      "loss": 0.1048,
+      "step": 20007
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.214285714285713e-06,
+      "loss": 0.1161,
+      "step": 20008
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.198979591836734e-06,
+      "loss": 0.029,
+      "step": 20009
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.183673469387754e-06,
+      "loss": 0.1513,
+      "step": 20010
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.168367346938774e-06,
+      "loss": 0.0369,
+      "step": 20011
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.153061224489796e-06,
+      "loss": 0.0336,
+      "step": 20012
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.137755102040815e-06,
+      "loss": 0.0581,
+      "step": 20013
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.122448979591835e-06,
+      "loss": 0.1394,
+      "step": 20014
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.107142857142856e-06,
+      "loss": 0.0246,
+      "step": 20015
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.091836734693876e-06,
+      "loss": 0.0736,
+      "step": 20016
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.076530612244897e-06,
+      "loss": 0.0551,
+      "step": 20017
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 9.061224489795919e-06,
+      "loss": 0.0235,
+      "step": 20018
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 9.045918367346937e-06,
+      "loss": 0.0215,
+      "step": 20019
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 9.030612244897958e-06,
+      "loss": 0.0478,
+      "step": 20020
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 9.015306122448978e-06,
+      "loss": 0.0762,
+      "step": 20021
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 8.999999999999999e-06,
+      "loss": 0.0973,
+      "step": 20022
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 8.984693877551019e-06,
+      "loss": 0.0235,
+      "step": 20023
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 8.969387755102041e-06,
+      "loss": 0.0576,
+      "step": 20024
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 8.954081632653061e-06,
+      "loss": 0.0203,
+      "step": 20025
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 8.93877551020408e-06,
+      "loss": 0.0351,
+      "step": 20026
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 8.9234693877551e-06,
+      "loss": 0.0696,
+      "step": 20027
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 8.90816326530612e-06,
+      "loss": 0.0307,
+      "step": 20028
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.892857142857143e-06,
+      "loss": 0.0253,
+      "step": 20029
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.877551020408163e-06,
+      "loss": 0.1012,
+      "step": 20030
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.862244897959184e-06,
+      "loss": 0.0187,
+      "step": 20031
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.846938775510204e-06,
+      "loss": 0.0379,
+      "step": 20032
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.831632653061223e-06,
+      "loss": 0.0278,
+      "step": 20033
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.816326530612243e-06,
+      "loss": 0.1002,
+      "step": 20034
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.801020408163265e-06,
+      "loss": 0.0257,
+      "step": 20035
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.785714285714286e-06,
+      "loss": 0.0754,
+      "step": 20036
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.770408163265306e-06,
+      "loss": 0.0696,
+      "step": 20037
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.755102040816326e-06,
+      "loss": 0.0324,
+      "step": 20038
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.739795918367345e-06,
+      "loss": 0.0749,
+      "step": 20039
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.724489795918365e-06,
+      "loss": 0.0348,
+      "step": 20040
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.709183673469387e-06,
+      "loss": 0.0731,
+      "step": 20041
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.693877551020408e-06,
+      "loss": 0.0289,
+      "step": 20042
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.678571428571428e-06,
+      "loss": 0.1023,
+      "step": 20043
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.663265306122449e-06,
+      "loss": 0.0162,
+      "step": 20044
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.647959183673469e-06,
+      "loss": 0.123,
+      "step": 20045
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.63265306122449e-06,
+      "loss": 0.0612,
+      "step": 20046
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.61734693877551e-06,
+      "loss": 0.0234,
+      "step": 20047
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.60204081632653e-06,
+      "loss": 0.0662,
+      "step": 20048
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.58673469387755e-06,
+      "loss": 0.032,
+      "step": 20049
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.0583,
+      "step": 20050
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.556122448979591e-06,
+      "loss": 0.0604,
+      "step": 20051
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.540816326530612e-06,
+      "loss": 0.0796,
+      "step": 20052
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.525510204081632e-06,
+      "loss": 0.0632,
+      "step": 20053
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.510204081632652e-06,
+      "loss": 0.0634,
+      "step": 20054
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.494897959183673e-06,
+      "loss": 0.1063,
+      "step": 20055
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.479591836734693e-06,
+      "loss": 0.15,
+      "step": 20056
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.464285714285714e-06,
+      "loss": 0.0325,
+      "step": 20057
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.448979591836734e-06,
+      "loss": 0.0302,
+      "step": 20058
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 8.433673469387754e-06,
+      "loss": 0.0662,
+      "step": 20059
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.418367346938775e-06,
+      "loss": 0.0557,
+      "step": 20060
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.403061224489795e-06,
+      "loss": 0.0632,
+      "step": 20061
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.387755102040816e-06,
+      "loss": 0.0613,
+      "step": 20062
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.372448979591836e-06,
+      "loss": 0.0265,
+      "step": 20063
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.357142857142856e-06,
+      "loss": 0.076,
+      "step": 20064
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.341836734693877e-06,
+      "loss": 0.111,
+      "step": 20065
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.326530612244897e-06,
+      "loss": 0.0676,
+      "step": 20066
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.311224489795917e-06,
+      "loss": 0.0666,
+      "step": 20067
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.295918367346938e-06,
+      "loss": 0.0306,
+      "step": 20068
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.280612244897958e-06,
+      "loss": 0.0204,
+      "step": 20069
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.265306122448979e-06,
+      "loss": 0.0533,
+      "step": 20070
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.249999999999999e-06,
+      "loss": 0.1435,
+      "step": 20071
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.23469387755102e-06,
+      "loss": 0.1016,
+      "step": 20072
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.21938775510204e-06,
+      "loss": 0.0236,
+      "step": 20073
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.20408163265306e-06,
+      "loss": 0.1061,
+      "step": 20074
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.18877551020408e-06,
+      "loss": 0.0705,
+      "step": 20075
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.173469387755101e-06,
+      "loss": 0.0382,
+      "step": 20076
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.158163265306121e-06,
+      "loss": 0.1416,
+      "step": 20077
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.142857142857142e-06,
+      "loss": 0.0605,
+      "step": 20078
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.127551020408162e-06,
+      "loss": 0.0629,
+      "step": 20079
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 8.112244897959184e-06,
+      "loss": 0.0246,
+      "step": 20080
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 8.096938775510204e-06,
+      "loss": 0.062,
+      "step": 20081
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 8.081632653061223e-06,
+      "loss": 0.0722,
+      "step": 20082
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 8.066326530612244e-06,
+      "loss": 0.0788,
+      "step": 20083
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 8.051020408163264e-06,
+      "loss": 0.0773,
+      "step": 20084
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 8.035714285714284e-06,
+      "loss": 0.0622,
+      "step": 20085
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 8.020408163265306e-06,
+      "loss": 0.0985,
+      "step": 20086
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 8.005102040816327e-06,
+      "loss": 0.0781,
+      "step": 20087
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 7.989795918367347e-06,
+      "loss": 0.0233,
+      "step": 20088
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 7.974489795918366e-06,
+      "loss": 0.0206,
+      "step": 20089
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 7.959183673469386e-06,
+      "loss": 0.1464,
+      "step": 20090
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.943877551020407e-06,
+      "loss": 0.0805,
+      "step": 20091
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.928571428571429e-06,
+      "loss": 0.0309,
+      "step": 20092
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.913265306122449e-06,
+      "loss": 0.1898,
+      "step": 20093
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.89795918367347e-06,
+      "loss": 0.0294,
+      "step": 20094
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.882653061224488e-06,
+      "loss": 0.1166,
+      "step": 20095
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.867346938775508e-06,
+      "loss": 0.1459,
+      "step": 20096
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.85204081632653e-06,
+      "loss": 0.0622,
+      "step": 20097
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.836734693877551e-06,
+      "loss": 0.0835,
+      "step": 20098
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.821428571428571e-06,
+      "loss": 0.0259,
+      "step": 20099
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.806122448979592e-06,
+      "loss": 0.1185,
+      "step": 20100
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.790816326530612e-06,
+      "loss": 0.0234,
+      "step": 20101
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.77551020408163e-06,
+      "loss": 0.0684,
+      "step": 20102
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.760204081632653e-06,
+      "loss": 0.051,
+      "step": 20103
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.744897959183673e-06,
+      "loss": 0.0635,
+      "step": 20104
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.729591836734694e-06,
+      "loss": 0.0328,
+      "step": 20105
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.714285714285714e-06,
+      "loss": 0.0662,
+      "step": 20106
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.698979591836734e-06,
+      "loss": 0.0821,
+      "step": 20107
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.683673469387755e-06,
+      "loss": 0.0576,
+      "step": 20108
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.668367346938775e-06,
+      "loss": 0.0624,
+      "step": 20109
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 7.653061224489796e-06,
+      "loss": 0.0517,
+      "step": 20110
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.637755102040816e-06,
+      "loss": 0.0673,
+      "step": 20111
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.622448979591836e-06,
+      "loss": 0.0268,
+      "step": 20112
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.607142857142856e-06,
+      "loss": 0.0259,
+      "step": 20113
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.591836734693877e-06,
+      "loss": 0.0887,
+      "step": 20114
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.5765306122448974e-06,
+      "loss": 0.0276,
+      "step": 20115
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.561224489795918e-06,
+      "loss": 0.1314,
+      "step": 20116
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.545918367346938e-06,
+      "loss": 0.0293,
+      "step": 20117
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.530612244897959e-06,
+      "loss": 0.0223,
+      "step": 20118
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.515306122448978e-06,
+      "loss": 0.0327,
+      "step": 20119
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.499999999999999e-06,
+      "loss": 0.0888,
+      "step": 20120
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.48469387755102e-06,
+      "loss": 0.1829,
+      "step": 20121
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.46938775510204e-06,
+      "loss": 0.0988,
+      "step": 20122
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.4540816326530605e-06,
+      "loss": 0.0305,
+      "step": 20123
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.438775510204081e-06,
+      "loss": 0.0305,
+      "step": 20124
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.423469387755101e-06,
+      "loss": 0.0307,
+      "step": 20125
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.4081632653061225e-06,
+      "loss": 0.0631,
+      "step": 20126
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.392857142857142e-06,
+      "loss": 0.2233,
+      "step": 20127
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.377551020408162e-06,
+      "loss": 0.0737,
+      "step": 20128
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.362244897959183e-06,
+      "loss": 0.0631,
+      "step": 20129
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.346938775510203e-06,
+      "loss": 0.0704,
+      "step": 20130
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.331632653061224e-06,
+      "loss": 0.1422,
+      "step": 20131
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.316326530612245e-06,
+      "loss": 0.0631,
+      "step": 20132
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.301020408163265e-06,
+      "loss": 0.0623,
+      "step": 20133
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.285714285714285e-06,
+      "loss": 0.182,
+      "step": 20134
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.270408163265305e-06,
+      "loss": 0.0987,
+      "step": 20135
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.2551020408163255e-06,
+      "loss": 0.0326,
+      "step": 20136
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.239795918367347e-06,
+      "loss": 0.1592,
+      "step": 20137
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.224489795918367e-06,
+      "loss": 0.0714,
+      "step": 20138
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.2091836734693874e-06,
+      "loss": 0.1407,
+      "step": 20139
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.193877551020407e-06,
+      "loss": 0.058,
+      "step": 20140
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.178571428571427e-06,
+      "loss": 0.0627,
+      "step": 20141
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.163265306122448e-06,
+      "loss": 0.027,
+      "step": 20142
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.147959183673469e-06,
+      "loss": 0.0632,
+      "step": 20143
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.132653061224489e-06,
+      "loss": 0.0696,
+      "step": 20144
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.11734693877551e-06,
+      "loss": 0.0451,
+      "step": 20145
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.10204081632653e-06,
+      "loss": 0.1376,
+      "step": 20146
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.08673469387755e-06,
+      "loss": 0.1054,
+      "step": 20147
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.071428571428571e-06,
+      "loss": 0.0248,
+      "step": 20148
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.056122448979591e-06,
+      "loss": 0.0224,
+      "step": 20149
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.040816326530612e-06,
+      "loss": 0.0277,
+      "step": 20150
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 7.025510204081632e-06,
+      "loss": 0.0665,
+      "step": 20151
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 7.010204081632652e-06,
+      "loss": 0.1105,
+      "step": 20152
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.994897959183673e-06,
+      "loss": 0.0258,
+      "step": 20153
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.979591836734694e-06,
+      "loss": 0.0621,
+      "step": 20154
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.9642857142857136e-06,
+      "loss": 0.0402,
+      "step": 20155
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.948979591836734e-06,
+      "loss": 0.0949,
+      "step": 20156
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.933673469387754e-06,
+      "loss": 0.0338,
+      "step": 20157
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.918367346938775e-06,
+      "loss": 0.0555,
+      "step": 20158
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.903061224489795e-06,
+      "loss": 0.1369,
+      "step": 20159
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.887755102040816e-06,
+      "loss": 0.1077,
+      "step": 20160
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.872448979591837e-06,
+      "loss": 0.0574,
+      "step": 20161
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.857142857142856e-06,
+      "loss": 0.0341,
+      "step": 20162
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.841836734693877e-06,
+      "loss": 0.0689,
+      "step": 20163
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.826530612244897e-06,
+      "loss": 0.0176,
+      "step": 20164
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.811224489795918e-06,
+      "loss": 0.0196,
+      "step": 20165
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.795918367346939e-06,
+      "loss": 0.1366,
+      "step": 20166
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.780612244897959e-06,
+      "loss": 0.0619,
+      "step": 20167
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.7653061224489785e-06,
+      "loss": 0.0248,
+      "step": 20168
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.749999999999999e-06,
+      "loss": 0.0238,
+      "step": 20169
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.734693877551019e-06,
+      "loss": 0.0949,
+      "step": 20170
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.7193877551020405e-06,
+      "loss": 0.0388,
+      "step": 20171
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.704081632653061e-06,
+      "loss": 0.0568,
+      "step": 20172
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.688775510204081e-06,
+      "loss": 0.0208,
+      "step": 20173
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.673469387755102e-06,
+      "loss": 0.0606,
+      "step": 20174
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.658163265306121e-06,
+      "loss": 0.0509,
+      "step": 20175
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.6428571428571416e-06,
+      "loss": 0.1374,
+      "step": 20176
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.627551020408163e-06,
+      "loss": 0.1483,
+      "step": 20177
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.612244897959183e-06,
+      "loss": 0.1035,
+      "step": 20178
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.5969387755102036e-06,
+      "loss": 0.0234,
+      "step": 20179
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.581632653061224e-06,
+      "loss": 0.0281,
+      "step": 20180
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.566326530612244e-06,
+      "loss": 0.0181,
+      "step": 20181
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.5510204081632656e-06,
+      "loss": 0.0365,
+      "step": 20182
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.535714285714285e-06,
+      "loss": 0.1499,
+      "step": 20183
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.5204081632653055e-06,
+      "loss": 0.1461,
+      "step": 20184
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.505102040816326e-06,
+      "loss": 0.0443,
+      "step": 20185
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.489795918367346e-06,
+      "loss": 0.0574,
+      "step": 20186
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.474489795918367e-06,
+      "loss": 0.0225,
+      "step": 20187
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.459183673469388e-06,
+      "loss": 0.0734,
+      "step": 20188
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.443877551020408e-06,
+      "loss": 0.0538,
+      "step": 20189
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.428571428571428e-06,
+      "loss": 0.0216,
+      "step": 20190
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.413265306122448e-06,
+      "loss": 0.0328,
+      "step": 20191
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.3979591836734685e-06,
+      "loss": 0.1685,
+      "step": 20192
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.382653061224489e-06,
+      "loss": 0.0249,
+      "step": 20193
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.36734693877551e-06,
+      "loss": 0.0981,
+      "step": 20194
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.3520408163265305e-06,
+      "loss": 0.0934,
+      "step": 20195
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.33673469387755e-06,
+      "loss": 0.1495,
+      "step": 20196
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.3214285714285704e-06,
+      "loss": 0.0575,
+      "step": 20197
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.306122448979591e-06,
+      "loss": 0.0673,
+      "step": 20198
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.290816326530612e-06,
+      "loss": 0.096,
+      "step": 20199
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.2755102040816324e-06,
+      "loss": 0.0272,
+      "step": 20200
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.260204081632653e-06,
+      "loss": 0.0619,
+      "step": 20201
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.244897959183673e-06,
+      "loss": 0.1405,
+      "step": 20202
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 6.229591836734693e-06,
+      "loss": 0.0296,
+      "step": 20203
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.214285714285713e-06,
+      "loss": 0.0347,
+      "step": 20204
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.198979591836734e-06,
+      "loss": 0.0249,
+      "step": 20205
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.183673469387755e-06,
+      "loss": 0.1148,
+      "step": 20206
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.168367346938775e-06,
+      "loss": 0.0641,
+      "step": 20207
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.1530612244897955e-06,
+      "loss": 0.1146,
+      "step": 20208
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.137755102040816e-06,
+      "loss": 0.0628,
+      "step": 20209
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.122448979591835e-06,
+      "loss": 0.0759,
+      "step": 20210
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.107142857142857e-06,
+      "loss": 0.0983,
+      "step": 20211
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.091836734693877e-06,
+      "loss": 0.0193,
+      "step": 20212
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.076530612244897e-06,
+      "loss": 0.0511,
+      "step": 20213
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 6.061224489795918e-06,
+      "loss": 0.0862,
+      "step": 20214
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 6.045918367346938e-06,
+      "loss": 0.0246,
+      "step": 20215
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 6.030612244897959e-06,
+      "loss": 0.0631,
+      "step": 20216
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 6.01530612244898e-06,
+      "loss": 0.0619,
+      "step": 20217
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 5.999999999999999e-06,
+      "loss": 0.0595,
+      "step": 20218
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 5.98469387755102e-06,
+      "loss": 0.0425,
+      "step": 20219
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 5.96938775510204e-06,
+      "loss": 0.0799,
+      "step": 20220
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 5.9540816326530605e-06,
+      "loss": 0.1113,
+      "step": 20221
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 5.938775510204082e-06,
+      "loss": 0.0262,
+      "step": 20222
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 5.923469387755102e-06,
+      "loss": 0.0518,
+      "step": 20223
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 5.908163265306122e-06,
+      "loss": 0.0257,
+      "step": 20224
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.892857142857142e-06,
+      "loss": 0.0258,
+      "step": 20225
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.877551020408162e-06,
+      "loss": 0.0282,
+      "step": 20226
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.862244897959183e-06,
+      "loss": 0.0974,
+      "step": 20227
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.846938775510204e-06,
+      "loss": 0.0589,
+      "step": 20228
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.831632653061224e-06,
+      "loss": 0.1027,
+      "step": 20229
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.816326530612245e-06,
+      "loss": 0.0291,
+      "step": 20230
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.801020408163264e-06,
+      "loss": 0.101,
+      "step": 20231
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.785714285714285e-06,
+      "loss": 0.0249,
+      "step": 20232
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.770408163265306e-06,
+      "loss": 0.0805,
+      "step": 20233
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.755102040816326e-06,
+      "loss": 0.0251,
+      "step": 20234
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.739795918367347e-06,
+      "loss": 0.1076,
+      "step": 20235
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.724489795918367e-06,
+      "loss": 0.0227,
+      "step": 20236
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.709183673469387e-06,
+      "loss": 0.1008,
+      "step": 20237
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.693877551020407e-06,
+      "loss": 0.024,
+      "step": 20238
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.678571428571428e-06,
+      "loss": 0.023,
+      "step": 20239
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.6632653061224486e-06,
+      "loss": 0.0594,
+      "step": 20240
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.647959183673469e-06,
+      "loss": 0.0205,
+      "step": 20241
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.632653061224489e-06,
+      "loss": 0.1037,
+      "step": 20242
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.61734693877551e-06,
+      "loss": 0.0258,
+      "step": 20243
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.602040816326529e-06,
+      "loss": 0.0322,
+      "step": 20244
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.586734693877551e-06,
+      "loss": 0.1061,
+      "step": 20245
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.571428571428571e-06,
+      "loss": 0.0312,
+      "step": 20246
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.556122448979591e-06,
+      "loss": 0.1025,
+      "step": 20247
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.540816326530612e-06,
+      "loss": 0.0589,
+      "step": 20248
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.525510204081632e-06,
+      "loss": 0.0459,
+      "step": 20249
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.510204081632653e-06,
+      "loss": 0.0263,
+      "step": 20250
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.494897959183674e-06,
+      "loss": 0.057,
+      "step": 20251
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.479591836734693e-06,
+      "loss": 0.0978,
+      "step": 20252
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.4642857142857135e-06,
+      "loss": 0.0265,
+      "step": 20253
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.448979591836734e-06,
+      "loss": 0.0194,
+      "step": 20254
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.433673469387754e-06,
+      "loss": 0.0636,
+      "step": 20255
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.4183673469387755e-06,
+      "loss": 0.0243,
+      "step": 20256
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.403061224489796e-06,
+      "loss": 0.104,
+      "step": 20257
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.387755102040816e-06,
+      "loss": 0.0269,
+      "step": 20258
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.372448979591836e-06,
+      "loss": 0.143,
+      "step": 20259
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.357142857142856e-06,
+      "loss": 0.065,
+      "step": 20260
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.3418367346938766e-06,
+      "loss": 0.0546,
+      "step": 20261
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.326530612244898e-06,
+      "loss": 0.0636,
+      "step": 20262
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.311224489795918e-06,
+      "loss": 0.0617,
+      "step": 20263
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.2959183673469386e-06,
+      "loss": 0.0445,
+      "step": 20264
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.280612244897959e-06,
+      "loss": 0.0283,
+      "step": 20265
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.2653061224489785e-06,
+      "loss": 0.0694,
+      "step": 20266
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.25e-06,
+      "loss": 0.0257,
+      "step": 20267
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.23469387755102e-06,
+      "loss": 0.0316,
+      "step": 20268
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.2193877551020405e-06,
+      "loss": 0.0647,
+      "step": 20269
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.204081632653061e-06,
+      "loss": 0.1494,
+      "step": 20270
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.188775510204081e-06,
+      "loss": 0.1439,
+      "step": 20271
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.173469387755101e-06,
+      "loss": 0.0354,
+      "step": 20272
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.158163265306123e-06,
+      "loss": 0.0227,
+      "step": 20273
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.142857142857142e-06,
+      "loss": 0.0224,
+      "step": 20274
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.127551020408163e-06,
+      "loss": 0.0599,
+      "step": 20275
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 5.112244897959183e-06,
+      "loss": 0.0361,
+      "step": 20276
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 5.0969387755102035e-06,
+      "loss": 0.1359,
+      "step": 20277
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 5.081632653061224e-06,
+      "loss": 0.0986,
+      "step": 20278
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 5.066326530612245e-06,
+      "loss": 0.0281,
+      "step": 20279
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 5.051020408163265e-06,
+      "loss": 0.0584,
+      "step": 20280
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 5.035714285714285e-06,
+      "loss": 0.0656,
+      "step": 20281
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 5.0204081632653054e-06,
+      "loss": 0.1002,
+      "step": 20282
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 5.005102040816326e-06,
+      "loss": 0.028,
+      "step": 20283
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 4.989795918367347e-06,
+      "loss": 0.0234,
+      "step": 20284
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 4.9744897959183674e-06,
+      "loss": 0.0969,
+      "step": 20285
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.959183673469388e-06,
+      "loss": 0.0576,
+      "step": 20286
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.943877551020407e-06,
+      "loss": 0.0797,
+      "step": 20287
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.928571428571428e-06,
+      "loss": 0.0677,
+      "step": 20288
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.913265306122448e-06,
+      "loss": 0.1543,
+      "step": 20289
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.897959183673469e-06,
+      "loss": 0.0677,
+      "step": 20290
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.88265306122449e-06,
+      "loss": 0.1028,
+      "step": 20291
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.86734693877551e-06,
+      "loss": 0.0574,
+      "step": 20292
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.8520408163265305e-06,
+      "loss": 0.0754,
+      "step": 20293
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.83673469387755e-06,
+      "loss": 0.0722,
+      "step": 20294
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.82142857142857e-06,
+      "loss": 0.0936,
+      "step": 20295
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.806122448979592e-06,
+      "loss": 0.1617,
+      "step": 20296
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.790816326530612e-06,
+      "loss": 0.0869,
+      "step": 20297
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.775510204081632e-06,
+      "loss": 0.0611,
+      "step": 20298
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.760204081632653e-06,
+      "loss": 0.0191,
+      "step": 20299
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.744897959183672e-06,
+      "loss": 0.1412,
+      "step": 20300
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.729591836734694e-06,
+      "loss": 0.069,
+      "step": 20301
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.714285714285714e-06,
+      "loss": 0.0774,
+      "step": 20302
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.698979591836734e-06,
+      "loss": 0.0598,
+      "step": 20303
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.683673469387755e-06,
+      "loss": 0.0602,
+      "step": 20304
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.668367346938775e-06,
+      "loss": 0.1249,
+      "step": 20305
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.6530612244897954e-06,
+      "loss": 0.0283,
+      "step": 20306
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.637755102040816e-06,
+      "loss": 0.0665,
+      "step": 20307
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.622448979591836e-06,
+      "loss": 0.1999,
+      "step": 20308
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.607142857142857e-06,
+      "loss": 0.0659,
+      "step": 20309
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.591836734693877e-06,
+      "loss": 0.0727,
+      "step": 20310
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.576530612244898e-06,
+      "loss": 0.0632,
+      "step": 20311
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.561224489795918e-06,
+      "loss": 0.0251,
+      "step": 20312
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.545918367346938e-06,
+      "loss": 0.0493,
+      "step": 20313
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.530612244897959e-06,
+      "loss": 0.0906,
+      "step": 20314
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.515306122448979e-06,
+      "loss": 0.1044,
+      "step": 20315
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.499999999999999e-06,
+      "loss": 0.0671,
+      "step": 20316
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.4846938775510205e-06,
+      "loss": 0.1059,
+      "step": 20317
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.46938775510204e-06,
+      "loss": 0.1086,
+      "step": 20318
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.45408163265306e-06,
+      "loss": 0.0272,
+      "step": 20319
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.438775510204082e-06,
+      "loss": 0.0558,
+      "step": 20320
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.423469387755102e-06,
+      "loss": 0.1033,
+      "step": 20321
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.4081632653061216e-06,
+      "loss": 0.108,
+      "step": 20322
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.392857142857143e-06,
+      "loss": 0.0675,
+      "step": 20323
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.377551020408163e-06,
+      "loss": 0.0325,
+      "step": 20324
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.362244897959183e-06,
+      "loss": 0.0681,
+      "step": 20325
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.346938775510204e-06,
+      "loss": 0.0542,
+      "step": 20326
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.331632653061224e-06,
+      "loss": 0.071,
+      "step": 20327
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.316326530612245e-06,
+      "loss": 0.1028,
+      "step": 20328
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.301020408163265e-06,
+      "loss": 0.0308,
+      "step": 20329
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 0.0635,
+      "step": 20330
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.270408163265306e-06,
+      "loss": 0.0277,
+      "step": 20331
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.255102040816326e-06,
+      "loss": 0.0492,
+      "step": 20332
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.239795918367347e-06,
+      "loss": 0.0903,
+      "step": 20333
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.224489795918367e-06,
+      "loss": 0.0628,
+      "step": 20334
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.209183673469387e-06,
+      "loss": 0.028,
+      "step": 20335
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.193877551020408e-06,
+      "loss": 0.0543,
+      "step": 20336
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.178571428571428e-06,
+      "loss": 0.1145,
+      "step": 20337
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.1632653061224485e-06,
+      "loss": 0.0339,
+      "step": 20338
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.147959183673469e-06,
+      "loss": 0.1003,
+      "step": 20339
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.132653061224489e-06,
+      "loss": 0.0241,
+      "step": 20340
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.11734693877551e-06,
+      "loss": 0.0225,
+      "step": 20341
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.10204081632653e-06,
+      "loss": 0.056,
+      "step": 20342
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.0867346938775504e-06,
+      "loss": 0.0612,
+      "step": 20343
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.071428571428571e-06,
+      "loss": 0.0359,
+      "step": 20344
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.056122448979592e-06,
+      "loss": 0.1467,
+      "step": 20345
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.0408163265306116e-06,
+      "loss": 0.1806,
+      "step": 20346
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.025510204081632e-06,
+      "loss": 0.0251,
+      "step": 20347
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 4.010204081632653e-06,
+      "loss": 0.0526,
+      "step": 20348
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.9948979591836736e-06,
+      "loss": 0.0686,
+      "step": 20349
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.979591836734693e-06,
+      "loss": 0.0422,
+      "step": 20350
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.964285714285714e-06,
+      "loss": 0.0296,
+      "step": 20351
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.948979591836735e-06,
+      "loss": 0.0909,
+      "step": 20352
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.933673469387754e-06,
+      "loss": 0.0962,
+      "step": 20353
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.9183673469387755e-06,
+      "loss": 0.1297,
+      "step": 20354
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.903061224489796e-06,
+      "loss": 0.0957,
+      "step": 20355
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.887755102040815e-06,
+      "loss": 0.0752,
+      "step": 20356
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.872448979591837e-06,
+      "loss": 0.0272,
+      "step": 20357
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.857142857142857e-06,
+      "loss": 0.0672,
+      "step": 20358
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.841836734693877e-06,
+      "loss": 0.0171,
+      "step": 20359
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.826530612244898e-06,
+      "loss": 0.0874,
+      "step": 20360
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.811224489795918e-06,
+      "loss": 0.0259,
+      "step": 20361
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.7959183673469385e-06,
+      "loss": 0.0484,
+      "step": 20362
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.780612244897959e-06,
+      "loss": 0.0843,
+      "step": 20363
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.7653061224489793e-06,
+      "loss": 0.0807,
+      "step": 20364
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.7499999999999997e-06,
+      "loss": 0.0952,
+      "step": 20365
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.73469387755102e-06,
+      "loss": 0.0314,
+      "step": 20366
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.7193877551020404e-06,
+      "loss": 0.1349,
+      "step": 20367
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.7040816326530612e-06,
+      "loss": 0.0572,
+      "step": 20368
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.688775510204081e-06,
+      "loss": 0.0432,
+      "step": 20369
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.6734693877551016e-06,
+      "loss": 0.1461,
+      "step": 20370
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.6581632653061224e-06,
+      "loss": 0.0493,
+      "step": 20371
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.6428571428571423e-06,
+      "loss": 0.0284,
+      "step": 20372
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.6275510204081627e-06,
+      "loss": 0.0611,
+      "step": 20373
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.6122448979591835e-06,
+      "loss": 0.0644,
+      "step": 20374
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.5969387755102035e-06,
+      "loss": 0.1068,
+      "step": 20375
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.581632653061224e-06,
+      "loss": 0.0393,
+      "step": 20376
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.5663265306122447e-06,
+      "loss": 0.2399,
+      "step": 20377
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.551020408163265e-06,
+      "loss": 0.1075,
+      "step": 20378
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.5357142857142854e-06,
+      "loss": 0.0701,
+      "step": 20379
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.520408163265306e-06,
+      "loss": 0.0369,
+      "step": 20380
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.505102040816326e-06,
+      "loss": 0.0343,
+      "step": 20381
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.489795918367347e-06,
+      "loss": 0.1211,
+      "step": 20382
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.474489795918367e-06,
+      "loss": 0.0865,
+      "step": 20383
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.4591836734693874e-06,
+      "loss": 0.099,
+      "step": 20384
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.443877551020408e-06,
+      "loss": 0.0596,
+      "step": 20385
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.428571428571428e-06,
+      "loss": 0.0555,
+      "step": 20386
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.4132653061224485e-06,
+      "loss": 0.1009,
+      "step": 20387
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 3.3979591836734693e-06,
+      "loss": 0.0309,
+      "step": 20388
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.3826530612244893e-06,
+      "loss": 0.1207,
+      "step": 20389
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.3673469387755096e-06,
+      "loss": 0.0288,
+      "step": 20390
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.3520408163265304e-06,
+      "loss": 0.0688,
+      "step": 20391
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.336734693877551e-06,
+      "loss": 0.0224,
+      "step": 20392
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.3214285714285708e-06,
+      "loss": 0.0638,
+      "step": 20393
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.3061224489795916e-06,
+      "loss": 0.0523,
+      "step": 20394
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.290816326530612e-06,
+      "loss": 0.168,
+      "step": 20395
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.2755102040816328e-06,
+      "loss": 0.1046,
+      "step": 20396
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.2602040816326527e-06,
+      "loss": 0.0591,
+      "step": 20397
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.244897959183673e-06,
+      "loss": 0.1097,
+      "step": 20398
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.229591836734694e-06,
+      "loss": 0.0204,
+      "step": 20399
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.214285714285714e-06,
+      "loss": 0.0536,
+      "step": 20400
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.1989795918367343e-06,
+      "loss": 0.0278,
+      "step": 20401
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.183673469387755e-06,
+      "loss": 0.054,
+      "step": 20402
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.168367346938775e-06,
+      "loss": 0.0645,
+      "step": 20403
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.1530612244897954e-06,
+      "loss": 0.0659,
+      "step": 20404
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.1377551020408162e-06,
+      "loss": 0.0737,
+      "step": 20405
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.1224489795918366e-06,
+      "loss": 0.0263,
+      "step": 20406
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.1071428571428566e-06,
+      "loss": 0.0478,
+      "step": 20407
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.0918367346938774e-06,
+      "loss": 0.0404,
+      "step": 20408
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.0765306122448977e-06,
+      "loss": 0.0233,
+      "step": 20409
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 3.0612244897959177e-06,
+      "loss": 0.0257,
+      "step": 20410
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 3.0459183673469385e-06,
+      "loss": 0.0195,
+      "step": 20411
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 3.030612244897959e-06,
+      "loss": 0.0802,
+      "step": 20412
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 3.0153061224489797e-06,
+      "loss": 0.0742,
+      "step": 20413
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.9999999999999997e-06,
+      "loss": 0.226,
+      "step": 20414
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.98469387755102e-06,
+      "loss": 0.1448,
+      "step": 20415
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.969387755102041e-06,
+      "loss": 0.0276,
+      "step": 20416
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.954081632653061e-06,
+      "loss": 0.1167,
+      "step": 20417
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.938775510204081e-06,
+      "loss": 0.061,
+      "step": 20418
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.923469387755102e-06,
+      "loss": 0.0407,
+      "step": 20419
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.9081632653061224e-06,
+      "loss": 0.1482,
+      "step": 20420
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.8928571428571423e-06,
+      "loss": 0.0606,
+      "step": 20421
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.877551020408163e-06,
+      "loss": 0.0233,
+      "step": 20422
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.8622448979591835e-06,
+      "loss": 0.0664,
+      "step": 20423
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.8469387755102035e-06,
+      "loss": 0.0193,
+      "step": 20424
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.8316326530612243e-06,
+      "loss": 0.0267,
+      "step": 20425
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.8163265306122447e-06,
+      "loss": 0.0785,
+      "step": 20426
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.8010204081632646e-06,
+      "loss": 0.0883,
+      "step": 20427
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.7857142857142854e-06,
+      "loss": 0.0807,
+      "step": 20428
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.770408163265306e-06,
+      "loss": 0.1049,
+      "step": 20429
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.7551020408163266e-06,
+      "loss": 0.0686,
+      "step": 20430
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.7397959183673466e-06,
+      "loss": 0.0329,
+      "step": 20431
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.724489795918367e-06,
+      "loss": 0.0472,
+      "step": 20432
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.7091836734693878e-06,
+      "loss": 0.1013,
+      "step": 20433
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.693877551020408e-06,
+      "loss": 0.0704,
+      "step": 20434
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.678571428571428e-06,
+      "loss": 0.0226,
+      "step": 20435
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.663265306122449e-06,
+      "loss": 0.0236,
+      "step": 20436
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.6479591836734693e-06,
+      "loss": 0.0268,
+      "step": 20437
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.6326530612244892e-06,
+      "loss": 0.0583,
+      "step": 20438
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.61734693877551e-06,
+      "loss": 0.0665,
+      "step": 20439
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.6020408163265304e-06,
+      "loss": 0.1018,
+      "step": 20440
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.5867346938775504e-06,
+      "loss": 0.1011,
+      "step": 20441
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.571428571428571e-06,
+      "loss": 0.0576,
+      "step": 20442
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.5561224489795916e-06,
+      "loss": 0.0276,
+      "step": 20443
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.540816326530612e-06,
+      "loss": 0.0191,
+      "step": 20444
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.5255102040816323e-06,
+      "loss": 0.1173,
+      "step": 20445
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.5102040816326527e-06,
+      "loss": 0.1014,
+      "step": 20446
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.4948979591836735e-06,
+      "loss": 0.0616,
+      "step": 20447
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.479591836734694e-06,
+      "loss": 0.0208,
+      "step": 20448
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.464285714285714e-06,
+      "loss": 0.0236,
+      "step": 20449
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.4489795918367347e-06,
+      "loss": 0.0702,
+      "step": 20450
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.433673469387755e-06,
+      "loss": 0.0406,
+      "step": 20451
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.418367346938775e-06,
+      "loss": 0.0195,
+      "step": 20452
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.403061224489796e-06,
+      "loss": 0.1012,
+      "step": 20453
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.387755102040816e-06,
+      "loss": 0.0362,
+      "step": 20454
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.372448979591836e-06,
+      "loss": 0.0228,
+      "step": 20455
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.357142857142857e-06,
+      "loss": 0.0727,
+      "step": 20456
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.3418367346938773e-06,
+      "loss": 0.0656,
+      "step": 20457
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.3265306122448977e-06,
+      "loss": 0.1505,
+      "step": 20458
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.311224489795918e-06,
+      "loss": 0.0655,
+      "step": 20459
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.2959183673469385e-06,
+      "loss": 0.0235,
+      "step": 20460
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.280612244897959e-06,
+      "loss": 0.1175,
+      "step": 20461
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.2653061224489797e-06,
+      "loss": 0.0262,
+      "step": 20462
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.2499999999999996e-06,
+      "loss": 0.0493,
+      "step": 20463
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.23469387755102e-06,
+      "loss": 0.1694,
+      "step": 20464
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.219387755102041e-06,
+      "loss": 0.0644,
+      "step": 20465
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.2040816326530608e-06,
+      "loss": 0.0624,
+      "step": 20466
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.1887755102040816e-06,
+      "loss": 0.0754,
+      "step": 20467
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.173469387755102e-06,
+      "loss": 0.1405,
+      "step": 20468
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.1581632653061223e-06,
+      "loss": 0.0237,
+      "step": 20469
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.1428571428571427e-06,
+      "loss": 0.0615,
+      "step": 20470
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.127551020408163e-06,
+      "loss": 0.1395,
+      "step": 20471
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.1122448979591835e-06,
+      "loss": 0.0962,
+      "step": 20472
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.096938775510204e-06,
+      "loss": 0.0384,
+      "step": 20473
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.0816326530612243e-06,
+      "loss": 0.0697,
+      "step": 20474
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.0663265306122446e-06,
+      "loss": 0.0317,
+      "step": 20475
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.051020408163265e-06,
+      "loss": 0.0658,
+      "step": 20476
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.0357142857142854e-06,
+      "loss": 0.0125,
+      "step": 20477
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.0204081632653058e-06,
+      "loss": 0.1029,
+      "step": 20478
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.0051020408163266e-06,
+      "loss": 0.1207,
+      "step": 20479
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.9897959183673465e-06,
+      "loss": 0.0835,
+      "step": 20480
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.9744897959183674e-06,
+      "loss": 0.0665,
+      "step": 20481
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.9591836734693877e-06,
+      "loss": 0.0314,
+      "step": 20482
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.9438775510204077e-06,
+      "loss": 0.1729,
+      "step": 20483
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.9285714285714285e-06,
+      "loss": 0.0588,
+      "step": 20484
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.913265306122449e-06,
+      "loss": 0.0257,
+      "step": 20485
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8979591836734693e-06,
+      "loss": 0.0625,
+      "step": 20486
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8826530612244896e-06,
+      "loss": 0.0213,
+      "step": 20487
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.86734693877551e-06,
+      "loss": 0.0256,
+      "step": 20488
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8520408163265306e-06,
+      "loss": 0.116,
+      "step": 20489
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8367346938775508e-06,
+      "loss": 0.0675,
+      "step": 20490
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8214285714285712e-06,
+      "loss": 0.063,
+      "step": 20491
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.8061224489795918e-06,
+      "loss": 0.0659,
+      "step": 20492
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.790816326530612e-06,
+      "loss": 0.029,
+      "step": 20493
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.7755102040816325e-06,
+      "loss": 0.0175,
+      "step": 20494
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.760204081632653e-06,
+      "loss": 0.0987,
+      "step": 20495
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.7448979591836735e-06,
+      "loss": 0.064,
+      "step": 20496
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.7295918367346937e-06,
+      "loss": 0.0939,
+      "step": 20497
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.714285714285714e-06,
+      "loss": 0.0213,
+      "step": 20498
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.6989795918367347e-06,
+      "loss": 0.0348,
+      "step": 20499
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.6836734693877548e-06,
+      "loss": 0.074,
+      "step": 20500
+    },
+    {
+      "epoch": 19.9,
+      "eval_cer": 0.23937778017603395,
+      "eval_loss": 1.2571896314620972,
+      "eval_runtime": 234.26,
+      "eval_samples_per_second": 37.625,
+      "eval_steps_per_second": 4.704,
+      "eval_wer": 0.6278109173999494,
+      "step": 20500
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.6683673469387754e-06,
+      "loss": 0.0867,
+      "step": 20501
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.6530612244897958e-06,
+      "loss": 0.1549,
+      "step": 20502
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.6377551020408164e-06,
+      "loss": 0.0241,
+      "step": 20503
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.6224489795918366e-06,
+      "loss": 0.0929,
+      "step": 20504
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.607142857142857e-06,
+      "loss": 0.0933,
+      "step": 20505
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.5918367346938775e-06,
+      "loss": 0.0173,
+      "step": 20506
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.5765306122448977e-06,
+      "loss": 0.0211,
+      "step": 20507
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.5612244897959183e-06,
+      "loss": 0.1558,
+      "step": 20508
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.5459183673469387e-06,
+      "loss": 0.0655,
+      "step": 20509
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.5306122448979589e-06,
+      "loss": 0.1013,
+      "step": 20510
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.5153061224489794e-06,
+      "loss": 0.1072,
+      "step": 20511
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.4999999999999998e-06,
+      "loss": 0.099,
+      "step": 20512
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.4846938775510204e-06,
+      "loss": 0.0157,
+      "step": 20513
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.4693877551020406e-06,
+      "loss": 0.1092,
+      "step": 20514
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.4540816326530612e-06,
+      "loss": 0.0329,
+      "step": 20515
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.4387755102040816e-06,
+      "loss": 0.0372,
+      "step": 20516
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.4234693877551017e-06,
+      "loss": 0.0402,
+      "step": 20517
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.4081632653061223e-06,
+      "loss": 0.1488,
+      "step": 20518
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.3928571428571427e-06,
+      "loss": 0.0641,
+      "step": 20519
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.3775510204081633e-06,
+      "loss": 0.0334,
+      "step": 20520
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.3622448979591835e-06,
+      "loss": 0.178,
+      "step": 20521
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.346938775510204e-06,
+      "loss": 0.1062,
+      "step": 20522
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.3316326530612245e-06,
+      "loss": 0.0778,
+      "step": 20523
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.3163265306122446e-06,
+      "loss": 0.0563,
+      "step": 20524
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.3010204081632652e-06,
+      "loss": 0.1069,
+      "step": 20525
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.2857142857142856e-06,
+      "loss": 0.0472,
+      "step": 20526
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.270408163265306e-06,
+      "loss": 0.0688,
+      "step": 20527
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.2551020408163264e-06,
+      "loss": 0.1137,
+      "step": 20528
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.239795918367347e-06,
+      "loss": 0.023,
+      "step": 20529
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.2244897959183673e-06,
+      "loss": 0.109,
+      "step": 20530
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.2091836734693875e-06,
+      "loss": 0.0792,
+      "step": 20531
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.193877551020408e-06,
+      "loss": 0.0162,
+      "step": 20532
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.1785714285714285e-06,
+      "loss": 0.1043,
+      "step": 20533
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.1632653061224489e-06,
+      "loss": 0.0202,
+      "step": 20534
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.1479591836734692e-06,
+      "loss": 0.1033,
+      "step": 20535
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.1326530612244898e-06,
+      "loss": 0.0254,
+      "step": 20536
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.11734693877551e-06,
+      "loss": 0.0684,
+      "step": 20537
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.1020408163265304e-06,
+      "loss": 0.024,
+      "step": 20538
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.086734693877551e-06,
+      "loss": 0.0864,
+      "step": 20539
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.0714285714285714e-06,
+      "loss": 0.0286,
+      "step": 20540
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.0561224489795917e-06,
+      "loss": 0.0945,
+      "step": 20541
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.0408163265306121e-06,
+      "loss": 0.0654,
+      "step": 20542
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.0255102040816325e-06,
+      "loss": 0.034,
+      "step": 20543
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.0102040816326529e-06,
+      "loss": 0.0329,
+      "step": 20544
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.948979591836733e-07,
+      "loss": 0.1213,
+      "step": 20545
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.795918367346939e-07,
+      "loss": 0.0622,
+      "step": 20546
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.642857142857142e-07,
+      "loss": 0.1791,
+      "step": 20547
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.489795918367346e-07,
+      "loss": 0.0972,
+      "step": 20548
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.33673469387755e-07,
+      "loss": 0.0664,
+      "step": 20549
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.183673469387754e-07,
+      "loss": 0.0409,
+      "step": 20550
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.030612244897959e-07,
+      "loss": 0.0415,
+      "step": 20551
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 8.877551020408163e-07,
+      "loss": 0.0751,
+      "step": 20552
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 8.724489795918368e-07,
+      "loss": 0.0544,
+      "step": 20553
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 8.57142857142857e-07,
+      "loss": 0.0625,
+      "step": 20554
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 8.418367346938774e-07,
+      "loss": 0.0236,
+      "step": 20555
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 8.265306122448979e-07,
+      "loss": 0.1091,
+      "step": 20556
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 8.112244897959183e-07,
+      "loss": 0.0304,
+      "step": 20557
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 7.959183673469388e-07,
+      "loss": 0.1634,
+      "step": 20558
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 7.806122448979591e-07,
+      "loss": 0.0261,
+      "step": 20559
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 7.653061224489794e-07,
+      "loss": 0.0567,
+      "step": 20560
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 7.499999999999999e-07,
+      "loss": 0.0202,
+      "step": 20561
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 7.346938775510203e-07,
+      "loss": 0.0324,
+      "step": 20562
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 7.193877551020408e-07,
+      "loss": 0.0308,
+      "step": 20563
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 7.040816326530612e-07,
+      "loss": 0.1149,
+      "step": 20564
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.887755102040817e-07,
+      "loss": 0.0235,
+      "step": 20565
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.73469387755102e-07,
+      "loss": 0.0424,
+      "step": 20566
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.581632653061223e-07,
+      "loss": 0.019,
+      "step": 20567
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.428571428571428e-07,
+      "loss": 0.068,
+      "step": 20568
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.275510204081632e-07,
+      "loss": 0.1119,
+      "step": 20569
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.122448979591837e-07,
+      "loss": 0.0807,
+      "step": 20570
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 5.96938775510204e-07,
+      "loss": 0.1864,
+      "step": 20571
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 5.816326530612244e-07,
+      "loss": 0.0227,
+      "step": 20572
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 5.663265306122449e-07,
+      "loss": 0.024,
+      "step": 20573
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 5.510204081632652e-07,
+      "loss": 0.1704,
+      "step": 20574
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 5.357142857142857e-07,
+      "loss": 0.0451,
+      "step": 20575
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 5.204081632653061e-07,
+      "loss": 0.0453,
+      "step": 20576
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 5.051020408163264e-07,
+      "loss": 0.2606,
+      "step": 20577
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.897959183673469e-07,
+      "loss": 0.0716,
+      "step": 20578
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.744897959183673e-07,
+      "loss": 0.1092,
+      "step": 20579
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.591836734693877e-07,
+      "loss": 0.0235,
+      "step": 20580
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.4387755102040813e-07,
+      "loss": 0.0173,
+      "step": 20581
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.285714285714285e-07,
+      "loss": 0.119,
+      "step": 20582
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.1326530612244895e-07,
+      "loss": 0.1251,
+      "step": 20583
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 3.979591836734694e-07,
+      "loss": 0.0251,
+      "step": 20584
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.826530612244897e-07,
+      "loss": 0.0605,
+      "step": 20585
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.6734693877551015e-07,
+      "loss": 0.092,
+      "step": 20586
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.520408163265306e-07,
+      "loss": 0.1118,
+      "step": 20587
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.36734693877551e-07,
+      "loss": 0.0789,
+      "step": 20588
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.214285714285714e-07,
+      "loss": 0.1586,
+      "step": 20589
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.0612244897959183e-07,
+      "loss": 0.0228,
+      "step": 20590
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.908163265306122e-07,
+      "loss": 0.0191,
+      "step": 20591
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.755102040816326e-07,
+      "loss": 0.0739,
+      "step": 20592
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.6020408163265303e-07,
+      "loss": 0.0265,
+      "step": 20593
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.4489795918367347e-07,
+      "loss": 0.0275,
+      "step": 20594
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 2.2959183673469385e-07,
+      "loss": 0.1247,
+      "step": 20595
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 2.1428571428571426e-07,
+      "loss": 0.062,
+      "step": 20596
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.989795918367347e-07,
+      "loss": 0.0713,
+      "step": 20597
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.8367346938775507e-07,
+      "loss": 0.0297,
+      "step": 20598
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.683673469387755e-07,
+      "loss": 0.083,
+      "step": 20599
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.5306122448979592e-07,
+      "loss": 0.0833,
+      "step": 20600
+    },
+    {
+      "epoch": 20.0,
+      "step": 20600,
+      "total_flos": 5.2126802777376974e+20,
+      "train_loss": 0.3304263302687447,
+      "train_runtime": 43278.5158,
+      "train_samples_per_second": 30.463,
+      "train_steps_per_second": 0.476
+    }
+  ],
+  "max_steps": 20600,
+  "num_train_epochs": 20,
+  "total_flos": 5.2126802777376974e+20,
+  "trial_name": null,
+  "trial_params": null
+}